Online Communities and Social Computing - OCSC 2011

Citation preview

Lecture Notes in Computer Science Commenced Publication in 1973 Founding and Former Series Editors: Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board David Hutchison Lancaster University, UK Takeo Kanade Carnegie Mellon University, Pittsburgh, PA, USA Josef Kittler University of Surrey, Guildford, UK Jon M. Kleinberg Cornell University, Ithaca, NY, USA Alfred Kobsa University of California, Irvine, CA, USA Friedemann Mattern ETH Zurich, Switzerland John C. Mitchell Stanford University, CA, USA Moni Naor Weizmann Institute of Science, Rehovot, Israel Oscar Nierstrasz University of Bern, Switzerland C. Pandu Rangan Indian Institute of Technology, Madras, India Bernhard Steffen TU Dortmund University, Germany Madhu Sudan Microsoft Research, Cambridge, MA, USA Demetri Terzopoulos University of California, Los Angeles, CA, USA Doug Tygar University of California, Berkeley, CA, USA Gerhard Weikum Max Planck Institute for Informatics, Saarbruecken, Germany

6778

A. Ant Ozok Panayiotis Zaphiris (Eds.)

Online Communities and Social Computing 4th International Conference, OCSC 2011 Held as Part of HCI International 2011 Orlando, FL, USA, July 9-14, 2011 Proceedings

13

Volume Editors A. Ant Ozok UMBC, Information Systems Department 1000 Hilltop Circle Baltimore, MD 21250, USA E-mail: [email protected] Panayiotis Zaphiris Cyprus University of Technology Department of Multimedia and Graphic Arts 3036 Limassol, Cyprus E-mail: [email protected]

ISSN 0302-9743 e-ISSN 1611-3349 e-ISBN 978-3-642-21796-8 ISBN 978-3-642-21795-1 DOI 10.1007/978-3-642-21796-8 Springer Heidelberg Dordrecht London New York Library of Congress Control Number: 2011929699 CR Subject Classification (1998): K.4, K.6, H.3-5, C.2, J.1, J.3 LNCS Sublibrary: SL 3 – Information Systems and Application, incl. Internet/Web and HCI

© Springer-Verlag Berlin Heidelberg 2011 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting: Camera-ready by author, data conversion by Scientific Publishing Services, Chennai, India Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Foreword

The 14th International Conference on Human–Computer Interaction, HCI International 2011, was held in Orlando, Florida, USA, July 9–14, 2011, jointly with the Symposium on Human Interface (Japan) 2011, the 9th International Conference on Engineering Psychology and Cognitive Ergonomics, the 6th International Conference on Universal Access in Human–Computer Interaction, the 4th International Conference on Virtual and Mixed Reality, the 4th International Conference on Internationalization, Design and Global Development, the 4th International Conference on Online Communities and Social Computing, the 6th International Conference on Augmented Cognition, the Third International Conference on Digital Human Modeling, the Second International Conference on Human-Centered Design, and the First International Conference on Design, User Experience, and Usability. A total of 4,039 individuals from academia, research institutes, industry and governmental agencies from 67 countries submitted contributions, and 1,318 papers that were judged to be of high scientific quality were included in the program. These papers address the latest research and development efforts and highlight the human aspects of design and use of computing systems. The papers accepted for presentation thoroughly cover the entire field of human–computer interaction, addressing major advances in knowledge and effective use of computers in a variety of application areas. This volume, edited by A. Ant Ozok and Panayiotis Zaphiris, contains papers in the thematic area of online communities and social computing (OCSC), addressing the following major topics: • • • • •

On-line communities and intelligent agents in education and research Blogs, wikis and twitters Social computing in business and the enterprise Social computing in everyday life Information management in social computing

The remaining volumes of the HCI International 2011 Proceedings are: • Volume 1, LNCS 6761, Human–Computer Interaction—Design and Development Approaches (Part I), edited by Julie A. Jacko • Volume 2, LNCS 6762, Human–Computer Interaction—Interaction Techniques and Environments (Part II), edited by Julie A. Jacko • Volume 3, LNCS 6763, Human–Computer Interaction—Towards Mobile and Intelligent Interaction Environments (Part III), edited by Julie A. Jacko • Volume 4, LNCS 6764, Human–Computer Interaction—Users and Applications (Part IV), edited by Julie A. Jacko • Volume 5, LNCS 6765, Universal Access in Human–Computer Interaction— Design for All and eInclusion (Part I), edited by Constantine Stephanidis

VI

Foreword

• Volume 6, LNCS 6766, Universal Access in Human–Computer Interaction— Users Diversity (Part II), edited by Constantine Stephanidis • Volume 7, LNCS 6767, Universal Access in Human–Computer Interaction— Context Diversity (Part III), edited by Constantine Stephanidis • Volume 8, LNCS 6768, Universal Access in Human–Computer Interaction— Applications and Services (Part IV), edited by Constantine Stephanidis • Volume 9, LNCS 6769, Design, User Experience, and Usability—Theory, Methods, Tools and Practice (Part I), edited by Aaron Marcus • Volume 10, LNCS 6770, Design, User Experience, and Usability— Understanding the User Experience (Part II), edited by Aaron Marcus • Volume 11, LNCS 6771, Human Interface and the Management of Information—Design and Interaction (Part I), edited by Michael J. Smith and Gavriel Salvendy • Volume 12, LNCS 6772, Human Interface and the Management of Information—Interacting with Information (Part II), edited by Gavriel Salvendy and Michael J. Smith • Volume 13, LNCS 6773, Virtual and Mixed Reality—New Trends (Part I), edited by Randall Shumaker • Volume 14, LNCS 6774, Virtual and Mixed Reality—Systems and Applications (Part II), edited by Randall Shumaker • Volume 15, LNCS 6775, Internationalization, Design and Global Development, edited by P.L. Patrick Rau • Volume 16, LNCS 6776, Human-Centered Design, edited by Masaaki Kurosu • Volume 17, LNCS 6777, Digital Human Modeling, edited by Vincent G. Duffy • Volume 19, LNCS 6779, Ergonomics and Health Aspects of Work with Computers, edited by Michelle M. Robertson • Volume 20, LNAI 6780, Foundations of Augmented Cognition: Directing the Future of Adaptive Systems, edited by Dylan D. Schmorrow and Cali M. Fidopiastis • Volume 21, LNAI 6781, Engineering Psychology and Cognitive Ergonomics, edited by Don Harris • Volume 22, CCIS 173, HCI International 2011 Posters Proceedings (Part I), edited by Constantine Stephanidis • Volume 23, CCIS 174, HCI International 2011 Posters Proceedings (Part II), edited by Constantine Stephanidis I would like to thank the Program Chairs and the members of the Program Boards of all Thematic Areas, listed herein, for their contribution to the highest scientific quality and the overall success of the HCI International 2011 Conference. In addition to the members of the Program Boards, I also wish to thank the following volunteer external reviewers: Roman Vilimek from Germany, Ramalingam Ponnusamy from India, Si Jung “Jun” Kim from the USA, and Ilia Adami, Iosif Klironomos, Vassilis Kouroumalis, George Margetis, and Stavroula Ntoa from Greece.

Foreword

VII

This conference would not have been possible without the continuous support and advice of the Conference Scientific Advisor, Gavriel Salvendy, as well as the dedicated work and outstanding efforts of the Communications and Exhibition Chair and Editor of HCI International News, Abbas Moallem. I would also like to thank for their contribution toward the organization of the HCI International 2011 Conference the members of the Human–Computer Interaction Laboratory of ICS-FORTH, and in particular Margherita Antona, George Paparoulis, Maria Pitsoulaki, Stavroula Ntoa, Maria Bouhli and George Kapnas. July 2011

Constantine Stephanidis

Organization

Ergonomics and Health Aspects of Work with Computers Program Chair: Michelle M. Robertson Arne Aar˚ as, Norway Pascale Carayon, USA Jason Devereux, UK Wolfgang Friesdorf, Germany Martin Helander, Singapore Ed Israelski, USA Ben-Tzion Karsh, USA Waldemar Karwowski, USA Peter Kern, Germany Danuta Koradecka, Poland Nancy Larson, USA Kari Lindstr¨om, Finland

Brenda Lobb, New Zealand Holger Luczak, Germany William S. Marras, USA Aura C. Matias, Philippines Matthias R¨ otting, Germany Michelle L. Rogers, USA Dominique L. Scapin, France Lawrence M. Schleifer, USA Michael J. Smith, USA Naomi Swanson, USA Peter Vink, The Netherlands John Wilson, UK

Human Interface and the Management of Information Program Chair: Michael J. Smith Hans-J¨ org Bullinger, Germany Alan Chan, Hong Kong Shin’ichi Fukuzumi, Japan Jon R. Gunderson, USA Michitaka Hirose, Japan Jhilmil Jain, USA Yasufumi Kume, Japan Mark Lehto, USA Hirohiko Mori, Japan Fiona Fui-Hoon Nah, USA Shogo Nishida, Japan Robert Proctor, USA

Youngho Rhee, Korea Anxo Cereijo Roib´ as, UK Katsunori Shimohara, Japan Dieter Spath, Germany Tsutomu Tabe, Japan Alvaro D. Taveira, USA Kim-Phuong L. Vu, USA Tomio Watanabe, Japan Sakae Yamamoto, Japan Hidekazu Yoshikawa, Japan Li Zheng, P. R. China

X

Organization

Human–Computer Interaction Program Chair: Julie A. Jacko Sebastiano Bagnara, Italy Sherry Y. Chen, UK Marvin J. Dainoff, USA Jianming Dong, USA John Eklund, Australia Xiaowen Fang, USA Ayse Gurses, USA Vicki L. Hanson, UK Sheue-Ling Hwang, Taiwan Wonil Hwang, Korea Yong Gu Ji, Korea Steven A. Landry, USA

Gitte Lindgaard, Canada Chen Ling, USA Yan Liu, USA Chang S. Nam, USA Celestine A. Ntuen, USA Philippe Palanque, France P.L. Patrick Rau, P.R. China Ling Rothrock, USA Guangfeng Song, USA Steffen Staab, Germany Wan Chul Yoon, Korea Wenli Zhu, P.R. China

Engineering Psychology and Cognitive Ergonomics Program Chair: Don Harris Guy A. Boy, USA Pietro Carlo Cacciabue, Italy John Huddlestone, UK Kenji Itoh, Japan Hung-Sying Jing, Taiwan Wen-Chin Li, Taiwan James T. Luxhøj, USA Nicolas Marmaras, Greece Sundaram Narayanan, USA Mark A. Neerincx, The Netherlands

Jan M. Noyes, UK Kjell Ohlsson, Sweden Axel Schulte, Germany Sarah C. Sharples, UK Neville A. Stanton, UK Xianghong Sun, P.R. China Andrew Thatcher, South Africa Matthew J.W. Thomas, Australia Mark Young, UK Rolf Zon, The Netherlands

Universal Access in Human–Computer Interaction Program Chair: Constantine Stephanidis Julio Abascal, Spain Ray Adams, UK Elisabeth Andr´e, Germany Margherita Antona, Greece Chieko Asakawa, Japan Christian B¨ uhler, Germany Jerzy Charytonowicz, Poland Pier Luigi Emiliani, Italy

Michael Fairhurst, UK Dimitris Grammenos, Greece Andreas Holzinger, Austria Simeon Keates, Denmark Georgios Kouroupetroglou, Greece Sri Kurniawan, USA Patrick M. Langdon, UK Seongil Lee, Korea

Organization

Zhengjie Liu, P.R. China Klaus Miesenberger, Austria Helen Petrie, UK Michael Pieper, Germany Anthony Savidis, Greece Andrew Sears, USA Christian Stary, Austria

Hirotada Ueda, Japan Jean Vanderdonckt, Belgium Gregg C. Vanderheiden, USA Gerhard Weber, Germany Harald Weber, Germany Panayiotis Zaphiris, Cyprus

Virtual and Mixed Reality Program Chair: Randall Shumaker Pat Banerjee, USA Mark Billinghurst, New Zealand Charles E. Hughes, USA Simon Julier, UK David Kaber, USA Hirokazu Kato, Japan Robert S. Kennedy, USA Young J. Kim, Korea Ben Lawson, USA Gordon McK Mair, UK

David Pratt, UK Albert “Skip” Rizzo, USA Lawrence Rosenblum, USA Jose San Martin, Spain Dieter Schmalstieg, Austria Dylan Schmorrow, USA Kay Stanney, USA Janet Weisenford, USA Mark Wiederhold, USA

Internationalization, Design and Global Development Program Chair: P.L. Patrick Rau Michael L. Best, USA Alan Chan, Hong Kong Lin-Lin Chen, Taiwan Andy M. Dearden, UK Susan M. Dray, USA Henry Been-Lirn Duh, Singapore Vanessa Evers, The Netherlands Paul Fu, USA Emilie Gould, USA Sung H. Han, Korea Veikko Ikonen, Finland Toshikazu Kato, Japan Esin Kiris, USA Apala Lahiri Chavan, India

James R. Lewis, USA James J.W. Lin, USA Rungtai Lin, Taiwan Zhengjie Liu, P.R. China Aaron Marcus, USA Allen E. Milewski, USA Katsuhiko Ogawa, Japan Oguzhan Ozcan, Turkey Girish Prabhu, India Kerstin R¨ ose, Germany Supriya Singh, Australia Alvin W. Yeo, Malaysia Hsiu-Ping Yueh, Taiwan

XI

XII

Organization

Online Communities and Social Computing Program Chairs: A. Ant Ozok, Panayiotis Zaphiris Chadia N. Abras, USA Chee Siang Ang, UK Peter Day, UK Fiorella De Cindio, Italy Heidi Feng, USA Anita Komlodi, USA Piet A.M. Kommers, The Netherlands Andrew Laghos, Cyprus Stefanie Lindstaedt, Austria Gabriele Meiselwitz, USA Hideyuki Nakanishi, Japan

Anthony F. Norcio, USA Ulrike Pfeil, UK Elaine M. Raybourn, USA Douglas Schuler, USA Gilson Schwartz, Brazil Laura Slaughter, Norway Sergei Stafeev, Russia Asimina Vasalou, UK June Wei, USA Haibin Zhu, Canada

Augmented Cognition Program Chairs: Dylan D. Schmorrow, Cali M. Fidopiastis Monique Beaudoin, USA Chris Berka, USA Joseph Cohn, USA Martha E. Crosby, USA Julie Drexler, USA Ivy Estabrooke, USA Chris Forsythe, USA Wai Tat Fu, USA Marc Grootjen, The Netherlands Jefferson Grubb, USA Santosh Mathan, USA

Rob Matthews, Australia Dennis McBride, USA Eric Muth, USA Mark A. Neerincx, The Netherlands Denise Nicholson, USA Banu Onaral, USA Kay Stanney, USA Roy Stripling, USA Rob Taylor, UK Karl van Orden, USA

Digital Human Modeling Program Chair: Vincent G. Duffy Karim Abdel-Malek, USA Giuseppe Andreoni, Italy Thomas J. Armstrong, USA Norman I. Badler, USA Fethi Calisir, Turkey Daniel Carruth, USA Keith Case, UK Julie Charland, Canada

Yaobin Chen, USA Kathryn Cormican, Ireland Daniel A. DeLaurentis, USA Yingzi Du, USA Okan Ersoy, USA Enda Fallon, Ireland Yan Fu, P.R. China Afzal Godil, USA

Organization

Ravindra Goonetilleke, Hong Kong Anand Gramopadhye, USA Lars Hanson, Sweden Pheng Ann Heng, Hong Kong Bo Hoege, Germany Hongwei Hsiao, USA Tianzi Jiang, P.R. China Nan Kong, USA Steven A. Landry, USA Kang Li, USA Zhizhong Li, P.R. China Tim Marler, USA

XIII

Ahmet F. Ozok, Turkey Srinivas Peeta, USA Sudhakar Rajulu, USA Matthias R¨ otting, Germany Matthew Reed, USA Johan Stahre, Sweden Mao-Jiun Wang, Taiwan Xuguang Wang, France Jingzhou (James) Yang, USA Gulcin Yucel, Turkey Tingshao Zhu, P.R. China

Human-Centered Design Program Chair: Masaaki Kurosu Julio Abascal, Spain Simone Barbosa, Brazil Tomas Berns, Sweden Nigel Bevan, UK Torkil Clemmensen, Denmark Susan M. Dray, USA Vanessa Evers, The Netherlands Xiaolan Fu, P.R. China Yasuhiro Horibe, Japan Jason Huang, P.R. China Minna Isomursu, Finland Timo Jokela, Finland Mitsuhiko Karashima, Japan Tadashi Kobayashi, Japan Seongil Lee, Korea Kee Yong Lim, Singapore

Zhengjie Liu, P.R. China Lo¨ıc Mart´ınez-Normand, Spain Monique Noirhomme-Fraiture, Belgium Philippe Palanque, France Annelise Mark Pejtersen, Denmark Kerstin R¨ ose, Germany Dominique L. Scapin, France Haruhiko Urokohara, Japan Gerrit C. van der Veer, The Netherlands Janet Wesson, South Africa Toshiki Yamaoka, Japan Kazuhiko Yamazaki, Japan Silvia Zimmermann, Switzerland

Design, User Experience, and Usability Program Chair: Aaron Marcus Ronald Baecker, Canada Barbara Ballard, USA Konrad Baumann, Austria Arne Berger, Germany Randolph Bias, USA Jamie Blustein, Canada

Ana Boa-Ventura, USA Lorenzo Cantoni, Switzerland Sameer Chavan, Korea Wei Ding, USA Maximilian Eibl, Germany Zelda Harrison, USA

XIV

Organization

R¨ udiger Heimg¨artner, Germany Brigitte Herrmann, Germany Sabine Kabel-Eckes, USA Kaleem Khan, Canada Jonathan Kies, USA Jon Kolko, USA Helga Letowt-Vorbek, South Africa James Lin, USA Frazer McKimm, Ireland Michael Renner, Switzerland

Christine Ronnewinkel, Germany Elizabeth Rosenzweig, USA Paul Sherman, USA Ben Shneiderman, USA Christian Sturm, Germany Brian Sullivan, USA Jaakko Villa, Finland Michele Visciola, Italy Susan Weinschenk, USA

HCI International 2013

The 15th International Conference on Human–Computer Interaction, HCI International 2013, will be held jointly with the affiliated conferences in the summer of 2013. It will cover a broad spectrum of themes related to human–computer interaction (HCI), including theoretical issues, methods, tools, processes and case studies in HCI design, as well as novel interaction techniques, interfaces and applications. The proceedings will be published by Springer. More information about the topics, as well as the venue and dates of the conference, will be announced through the HCI International Conference series website: http://www.hci-international.org/ General Chair Professor Constantine Stephanidis University of Crete and ICS-FORTH Heraklion, Crete, Greece Email: [email protected]

Table of Contents

Part I: On-Line Communities and Intelligent Agents in Education and Research Promoting Reflective Learning: The Role of Blogs in the Classroom . . . . Rahayu Ahmad and Wayne G. Lutters

3

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sung-Wook Baek, Jong-Hoon Seo, and Tack-Don Han

12

Gaining Insight into the Application of Second Life in a Computing Course: Students’ Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . James Braman, Alfreda Dudley, Kidd Colt, Giovanni Vincenti, and Yuanqiong Wang

20

Third Agers and Social Networking in Higher Education . . . . . . . . . . . . . . Christopher M. Connolly and Gabriele Meiselwitz

30

A Real-time Disaster Situation Mapping System for University Campuses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kayo Iizuka, Yasuki Iizuka, and Kyoko Yoshida

40

Recent Trends in Software Support for Online Communities for Teaching and Research Projects in Higher Education . . . . . . . . . . . . . . . . . Daniel Kadenbach and Carsten Kleiner

50

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments: Recommendations for Educators and Wiki Designers . . . . Portia Pusey and Gabriele Meiselwitz

60

Teacher Agents: The Current State, Future Trends, and Many Roles of Intelligent Agents in Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kevin Reed and Gabriele Meiselwitz

69

Part II: Blogs, Wikis and Twitters Interpreting User-Generated Content: What Makes a Blog Believeable? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Rahayu Ahmad and Wayne G. Lutters Extracting Events from Wikipedia as RDF Triples Linked to Widespread Semantic Web Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Carlo Aliprandi, Francesco Ronzano, Andrea Marchetti, Maurizio Tesconi, and Salvatore Minutoli

81

90

XVIII

Table of Contents

Collaborative Sensemaking during Admin Permission Granting in Wikipedia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Katie Derthick, Patrick Tsao, Travis Kriplean, Alan Borning, Mark Zachry, and David W. McDonald Mining Social Relationships in Micro-blogging Systems . . . . . . . . . . . . . . . Qin Gao, Qu Qu, and Xuhui Zhang Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nirupama Dharmavaram Sreenivasan, Chei Sian Lee, and Dion Hoe-Lian Goh

100

110

120

Impact of Blog Design Features on Blogging Satisfaction: An Impression Management Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wee-Kek Tan and Hock-Hai Teo

130

An Exploratory Study of Navigating Wikipedia Semantically: Model and Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I.-Chin Wu, Yi-Sheng Lin, and Che-Hung Liu

140

Part III: Social Computing in Business and the Enterprise Productivity Improvement by Using Social-Annotations about Design Intent in CAD Modelling Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gerardo Alducin-Quintero, Manuel Contero, Jorge Mart´ın-Guti´errez, David A. Guerra-Zubiaga, and Michael D. Johnson Handshake: A Case Study for Exploring Business Networking for the Enterprise, Inside and Out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Laurie E. Damianos, Donna L. Cuomo, and Stan Drozdetski “Your Team Cohesion is Low”: A Systematic Study of the Effects of Social Network Feedback on Mediated Activity . . . . . . . . . . . . . . . . . . . . . . Luciano Gamberini, Francesco Martino, Anna Spagnolli, Roberto Ba` u, and Michela Ferron

153

162

172

E-Business Solutions in the Cable TV Industry . . . . . . . . . . . . . . . . . . . . . . Viorica Harrison and June Wei

182

Measuring the Success of On-line Communities in an Enterprise Networking Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lester J. Holtzblatt and Laurie E. Damianos

188

Business-to-Business Solutions for the Cosmetic Industry . . . . . . . . . . . . . Manami Koblentz and June Wei

197

Table of Contents

XIX

Online Crowdsourcing in the Public Sector: How to Design Open Government Platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Giordano Koch, Johann F¨ uller, and Sabine Brunswicker

203

An M-Pill Framework in the Electronic Healthcare . . . . . . . . . . . . . . . . . . . Nien-Chieh Lee, Hi Tran, Albert Yin, and June Wei

213

Online Design Discussion Sites: Emerging Resource for Creative Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Moushumi Sharmin and Brian P. Bailey

219

Social Computing for Home Energy Efficiency: Technological and Stakeholder Ecosystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marc Torrent-Moreno, Rafael Gim´enez, Regina Enrich, Jos´e Javier Garc´ıa, and Mar´ıa P´erez

229

Part IV: Social Computing in Everyday Life Conversational Lives: Visualizing Interpersonal Online Social Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Heng Chen and Aisling Kelliher

241

CICRO: An Interactive Visual Interface for Crowd Communication Online . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Masao Ohira, Hitoshi Masaki, and Ken-ichi Matsumoto

251

Features to Support Persistent Chat Conversation . . . . . . . . . . . . . . . . . . . . Mohsen Rahimifar and Siti Salwah Salim

261

Design and Evaluation of an Online Social Support Application for Family Caregivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Matthieu Tixier and Myriam Lewkowicz

267

An Investigation into the Social Network between Three Generations in a Household: Bridging the Interrogational Gaps between the Senior and the Youth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tsai-Hsuan Tsai, Yi-Lun Ho, and Kevin C. Tseng

277

Preschoolers as Video Gamers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alexander Voiskounsky

287

GAZO GAZO KUN: Photo-Sharing System Using an Anthropomorphic Photo Frame for Communication Support . . . . . . . . . . . . . . . . . . . . . . . . . . . Takashi Yoshino and Tomoya Matsuo

297

XX

Table of Contents

Part V: Information Management in Social Computing Is Your Social Networking Privacy Reliant on Intuitive Interfaces? . . . . . Deborah S. Carstens and Veronica Giguere

309

A Study on Social Network Services Visualization Based on User Needs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Young Suk Han, Jong Kyu Choi, and Yong Gu Ji

319

A Virtual World Prototype for Interacting with a Music Collection . . . . . Jukka Holm and Arto Lehtiniemi

326

A Supervised Machine Learning Link Prediction Approach for Tag Recommendation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Manisha Pujari and Rushed Kanawati

336

Beyond the Social Search: Personalizing the Semantic Search in Social Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J´ ulio Cesar dos Reis, Rodrigo Bonacin, and M. Cec´ılia C. Baranauskas

345

Factors Influencing Online Social Interactions . . . . . . . . . . . . . . . . . . . . . . . . Shankar Ravi

355

Effects of a Synchronized Scoring Interface on Musical Quality . . . . . . . . . Yuji Takai, Masao Ohira, and Ken-ichi Matsumoto

363

Measurement of Tagging Behavior Differences . . . . . . . . . . . . . . . . . . . . . . . Li-Chen Tsai, Sheue-Ling Hwang, and Kuo-Hao Tang

373

Community Search: A Collaborative Searching Web Application with a User Ranking System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Papon Yongpisanpop, Masao Ohira, and Ken-ichi Matsumoto

378

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

387

Promoting Reflective Learning: The Role of Blogs in the Classroom Rahayu Ahmad and Wayne G. Lutters Department of Information Systems, University of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250, USA {arahayu1,lutters}@umbc.edu

Abstract. The enthusiasm for adopting social media technologies should be tempered by a critical, empirical understanding of how they facilitate an effective learning envionrment. This study analyzed blog use in two graduatelevel university courses, specifically identifying reflective learning markers in 279 blog entries. This analysis was deepened with follow-up interviews of several top bloggers. The reflective bloggers were characterized as being comfortable with their audience, having a deeper thought process, enjoying the interaction, and progressing well throughout the semester. Our results suggest guidlines for effective use of blogging in the classroom to support reflective learning.

1 Introduction Web 2.0 technologies like web blogs (blogs) and wikis have received great interest as a means of innovation in the classroom. The enthusiasm for adopting these technologies should be tempered by a critical, empirical understanding of how they facilitate a conducive learning environment. Ideally, students should take active roles in constructing new knowledge and have skills to critically examine their assumptions and beliefs. These are known as reflective learning skills, which have been recognized as a essential component of life long learning [6]. Some researchers have claimed that blogs are “effective at supporting reflective learning more so than other technologies” [26, p. 1656]. Despite the attractiveness of these arguments, many of the studies claiming the appropriateness of blogs for reflective learning fail to provide evidence from student’s blog entries to support their assertions [23][27]. There is a dearth of empirical work that can thoughtfully guide educators on the correct use of blogs to support reflective learning. For example, in a study involving 56 students in a teacher education program, only five expressed any interest in incorporating blogs in their own classrooms after experimenting with blogs. The primary reason cited was that they were unable to see how the blogs could facilitate reflective learning [11]. Contributing to our understanding of the impact blogging can have on reflectice learning, this study examines actual students’ blogs entrees for identifying markers of reflective thinking, and does not rely merely on self-report perceptions of blogging. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 3–11, 2011. © Springer-Verlag Berlin Heidelberg 2011

4

R. Ahmad and W.G. Lutters

2 Background 2.1 Reflective Learning Reflective learning has been recognized as one of the most important lifelong learnin skills as it empowers students to wrestle effectively with complex ideas [6]. Reflective learning can be summarized as a process of conscious thought as to what is being learned, which means that the learner is not simply a passive recipient but rather an active contributor within the learning process [22]. Through reflective thinking, the learner becomes intellectually responsible and can make informed, accountable decisions [9]. One means of fostering reflective learning among students is through writing reflective journals. Journaling has been recognized as effective in developing students’ meta cognitive and reflective skills [18]. This approach has been widely adopted in nursing and teacher education as it has been found to be helpful in promoting deeper understanding [15]. Through journaling students become aware of their thought process and the consequences of their experiences. However, students need to be encouraged to be critical in their reflection, in which mere description of events and experiences in journals is not sufficient; students should be able to synthesize the new experience or knowledge and able to relate with previous knowledge, forming a coherent perspective towards an issue or phenomenon [10]. 2.2 Blogs as Tool for Reflective Learning Traditionally hard-copy learning logs have been used to document experiences and reflective thoughts. The rise of blog publishing gives new aspiration for reflection learning. Unlike traditional journals, which only serve the instructor, blogs have the ability to engage a larger audience, forming a close knit community of learners. The public nature of blogs within the class encourages the students to view each other work, hence promoting collaborative writing and reflection among students [1]. Researchers have emphasized the needs for collaborative reflection in enhancing the development of professional knowledge [17]. Furthermore, a blog is interactive in the sense that readers can respond with comments which will be beneficial in promoting a collaborative learning environment [7]. Initial studies showed that blogging can support reflective thinking, critical thinking, and collaboration, allowing students to take ownership of their learning [2][20]. However, there is very limited evidence showing the levels of reflection that are demonstrated in the students’ blogs. Some papers made substantial contributions providing arguments of the appropriateness of blog features for reflective learning, but did not demonstrate the real implementation in a class [12][25]. The majority of these studies relied on students’ self-report data on their perceptions of blog usefulness, not their actual blogging activity or content of their entries [27]. Although there have been a few papers demonstrating initial indicators of reflective learning in blogs, these studies were mainly conducted in a very small classes, ranging from five to twenty students [19][21]. Thus, we still do not know whether similar findings can be reached in a larger population of learners. This study attempts to address call for more in-field studies of the informal nature of blogging to foster reflective learning [1][16][19].

Promoting Reflective Learning: The Role of Blogs in the Classroom

5

3 Study Design This study is based on the blog implementation in two sections of a graduate course within the Information Systems Department at UMBC. This course, called Online Communities, employed a hybrid approach in which there were a face-to-face academic lectures and discussions every week for 3 hours supplemented with significant online activities. Prior to every week’s class, students were given an assignment to be explored and they were required to blog about their experiences. These personal blogs were managed using the Blackboard course content management software’s blogging tool. These posts were private within a “neighborhood,” and accessible only by participants and instructional staff for the two classes (and researchers after the conclusion of the semester). Basic sets of reflective guideline questions were included in each week’s activity Guided by these questions students were free to explore the assignments and be reflective in their blogs. These blogs were graded and contributed to student’s overall course grade. In total there were 279 distinctive blog entries posted by 31 students across the two classes. The majority of the students had not created a blog before, and none had ever been asked to produce a blog for a course assessment.

4 Analysis We conducted a content analysis on the weekly blog posts to classify them into reflective and non-reflective posts. We used a deductive approach in analyzng the data, guided by an established reflective coding framework by Kember et al. [13]. We used the following definition of reflective learning to guide our coding process: Reflective learning is the process of internally examining and exploring an issue of concern, triggered by an experience, which creates and clarifies meaning in terms of self, and which results in a changed conceptual perspective [3, pp.100]. We also examined the interaction among the students through the commenting capabilities supported by the Blackboard blog system. Following this analysis, we conducted contextual interviews with seven individuals, talking through their blog entries from the prior semester.

5 Results and Discussion We start off this section by describing the characteristics of the reflective bloggers. This is followed by an examination of their reflective posts. 5.1 The Reflective Bloggers For classifying the reflective and non-reflective bloggers, we used the total number of entrees classified as reflective, using the method desribed above, as our indicator. Students who had more than 6 reflective posts (out of 91) were classified as reflective 1

For this study, we excluded the blog entries from weeks 1 and 13 as they were largely introductory. Posts from weeks 10, 11 and 12 were also excluded as the focused on structured debates.

6

R. Ahmad and W.G. Lutters

bloggers. In total, there were 8 (26%) students classified as reflective bloggers. Among the non reflective bloggers, 12 of them had 4 to 5 entrees classified as reflective. The rest of 11 students had lower than 3 entriess classified as reflective posts. Figure 1 illustrates the total coded reflective posts for each student. Comfortable with Peer Audience. In our interviews with the top reflective bloggers, they explained how they were comfortable in blogging their thoughts and experiences. They enjoyed blogging for known audience, in this case, their peers. Blogging in a close knit community that shared the same purpose motivated them to be more reflective: “I am comfortable blogging in that environment because it was a small community and I knew the people there.” Josh2 “The audience was interested in this stuff from the getgo, so that makes it easier.” Jay

Fig. 1. Total Reflective Posts by Student

Deeper Thoughts. The students appreciated the process of blogging which motivates them to critically review their learning experience [8]. Being required to blog encouraged them to systematically think about their experiences: “It does help my thinking process. In thinking about posting, it triggered some thoughts that I wouldn't have thought about or looked over if I was just thinking it and not writing it down.” Josh “I think you always learn more and get more out of an experience if you have to explain it or share it with others.” Sara

2

Our stylistic convention is to identify quotations from the interviews in italics and direct excerpts from student blog entries in Courier font.

Promoting Reflective Learning: The Role of Blogs in the Classroom

7

“Dr. Z would give us certain things to do each week and since we had to blog about it, it allowed us to really think about our experiences and what we were observing rather than just going through the motions.” Tom Interaction is Important. In reference to Stahl’s social theory of computer-supported collaborative learning [22], learning is argued as more a process of knowledge creation process through interactions or conversations. In blogs, conversations can take place through commenting features which allows for personalized feedback and allow subsequent referencing and revisions. All our reflective bloggers were enthusiastic in receiving comments and feedback from their peers especially constructive feedback that allow extended discussions in their blogs as described: “Yeah. I recall really looking forward to when people would ask engaging questions within the comments. I always enjoy a compliment, but it was fun to "extend" discussion into the comments” Jay “If someone just posted "I agree" that's not as much value. I prefer feedback that contains content and adds value to the topic.” Josh Positive Evolution. Students thought their blogs were progressing well when they learned to understand more about the audience and the dynamics of the class mainly from the interactions within blogs. The topic of the assignment plays significant role, in which the latter assignments allow them to be more personalized in writing something they can relate more and having real experiences: “It's gotten longer and I'm a bit more comfortable voicing more opinions after spending time with the other classmates.” Sara “There was a feeling of "improvement," for sure. At the start I didn't really bother with a lot of style (as I can recall), but I like to think I was a lot more free. I got more comfortable asking research questions, adding formatted text, pictures. I also like to think I got better with the research questions.” Jay 5.2 Reflective Posts As demonstrated by Figure 2 there was no uniform distribution of reflective posts although there was a slight positive trend toward more reflective posts over the course of the semester. Most students were reflective in their week 2 entrees followed by week 10 entrees. We will futher describe the potential factors that influenced their reflective process. Guiding Questions. We argued that blogs alone might not be sufficient for promoting reflective thinking. We echoed Chen’s [5] recommendation that students should be aware that reflection should be demonstrated in their blog posts. Students need to be provided with a common ground topic or issue to initiate their reflective writing process [16]. Anecdotal evidence in previous studies suggested that students faced difficulty writing an online diary whose inputs they were (relatively) free to determine [8][21]. Among the problems faced by the students were difficulty in finding the topics to talk about and attracting others to talk on the particular topic. Our students shared the importance of the guided questions as quoted:

8

R. Ahmad and W.G. Lutters

Fig. 2. Percentage of Reflective Posts by Week

I think one of my main issues with blogging is having something to blog about. So when we had assignments and topics that my fellow classmates might be interested in reading about, it was fairly easy to blog. Tom The exercises were very helpful since it gave us a topic to discuss and for people like me who haven't blogged, it was a starting point. Sara Most of the time, especially in second week, the guiding questions stimulated the students to be more reflective in their blog posts. The instructor would supplement the assignments with guiding reflective questions for example: We have all had experiences with online communities and social computing over the years. Think back over your history and provide a retrospective on your blog.In which communities have you participated? What tools have you used? Think critically.What has changed over time? What has remained the same? What general questions arise from these specific experiences? In respond to these guiding reflective questions, students were able to reflect on own their experience and formed perspective on the evolution of soial computing. Following is one exceprt that demonstrates the kind of reflective response: It is interesting to look at how social computing has evolved over time. The core method of communication (text line chat) hasn’t changed much, the environment in which the communication occurs has. Individual representation has evolved from a screen name, to a buddy icon, to 3D avatars. It also appears that there is a wider age gap in the social computing scene. Ben Similarly in another activity from week 8, the guiding questions required users to explore behind the scenes of the Wikipedia editing process and provoked them to think on the stability and reliability of Wikipedia. This resulted numerous reflective posts such as:

Promoting Reflective Learning: The Role of Blogs in the Classroom

9

I was skeptical of Wikipedia.org at first because of the validity of information, because Wikipedia.org was not a credible source and if you think logically it is not a credible source until information in it is cited with a credible source.I got over my paranoia and became more of an avid user. Overall the user base is what Wikipedia.org's strength is people; who have passion for information and not only information but correct information. User's put some good quality time in their posts and are passionate for sharing information. There is definitely an overly more positive user base then negative which helps Wikipedia.org to remain accurate and free of ambiguity. Matthew Immersion and Learning. We proposed that the type of learning activities may influence the reflective learning process. In week 10, for example, the students were required to blog on their experiences in massive multiplayer online role playing games. Students were able to articulate and analyze the components for an engaging virtual world as demonstrated by this excerpt: While both Secondlife and MMORPGs have social interaction, they are very different. The MMORPG games give users a common starting point for communication. They are all in the game trying to complete a set of goals. These games also have guilds which are tiny communities within the game that players belong. I believe these guilds give the user a sense of belonging and therefore foster more in game communication/interaction. In my experience, Second Life is too open and doesn't give players a sense of belonging Ben The immersion in virtual worlds triggered some of the deepest reflective moments of the course which led to individualized knowledge construction demonstrated by most of the students.

6 Recommendations This study examined the role of blogs in promoting reflective learning. Some practical guidelines emerged that would be useful for anyone seeking to incorporate blogs for this purpose or to fine-tune their existing implementations. • Define and bound the audience. The greatest utility came from having the blogs open to more than just the instructor, but limited to the students enrolled in the class. By knowing their audience students were able to tailor their blog entries and felt more comfortable than posting for world-wide readers on the Internet. • Provide thoughtful prompts. The exercises and guiding questions were essential to provide a common ground among the students and make reflection a clear goal of the learning experience. • Encourage active peer feedback. While the blog entries were evaluated by the instructor, the most valuable feedback came from the peer discussions that emerged in the comments surrounding student posts. • Make blogging a routine, ongoing activity. It is critical for students to be able to see development in their own thinking and that of their peers. This is not possible with a one-off blogging activity. Further research is required to understand the effective design of reflective prompts and peer commenting processes.

10

R. Ahmad and W.G. Lutters

References 1. Armstrong, L., Berry, M., Lamshed, R.: Reflective Learning and Blogs. In: International Conference on Computers in Education (2004) 2. Bouldin, A.S., Holmes, E.R., Fortenberry, M.L.: Blogging About Course Concepts: Using Technology For Reflective Journaling In A Communications Class. American Journal of Pharmaceutical Education 70(4) (2006) 3. Boyd, E.M., Fales, A.W.: Reflecctive Learning: Key To Learning From Experience. Journal of Humanistic Psychology 23(2), 99–117 (1983) 4. Chen, H.L., Cannon, D., Gabrio, J., Leifer, L., Toye, G., Bailey, T.: Using wikis and weblogs to support reflective learning in an introductory engineering design course. In: Proceedings: American Society for Engineering Education Annual Conference & Exposition, Portland, Oregon (2005) 5. Costa, A.L., Kallik, B.: Getting Into The Habit Of Reflection. Educational Leadership 57(7) (2000) 6. Dickey, M.D.: The Impact Of Web-Logs (Blogs) On Student Perceptions Of Isolation And Alienation In A Web-Based Distance-Learning Environment. Open Learning 19(3), 279– 291 (2004) 7. Du, H.S., Wagner, C.: Learning With Weblogs: An Empirical Investigation. In: Proceedings: 38th Hawaii International Conference on System Science, Hawaii (2005) 8. Dewey, J.: How we think. DC Heath, New York (1993) 9. Fulwiler, T.: The Journal Book. Boynton/Cook, Portsmouth (1987) 10. Hernández-Ramos, P.: Web Logs And Online Discussions As Tools To Promote Reflective Practice. The Journal of Interactive Online Learning 3(1) (2004) 11. Ke, F., Xie, Y.: Blogging for Reflective Learning in an Introductory Political Science Course. In: 29th Annual Proceeding of the Accociations for Educational Communications and Technology (2006) 12. Kember, D., Jones, A., Loke, A., McKay, J., Sinclair, K., Tse, H., et al.: Determining The Level Of Reflective Thinking From Students’ Written Journals Using A Coding Scheme Based On The Work Of Mezirow. International Journal of Lifelong Education 18(1), 18– 30 (1999) 13. Knapp, C.E.: Lasting lessons: A Teacher’s Guide To Reflecting On Experience. ERIC Clearinghouse on Rural Education and Small Schools, Charleston (1993) 14. Loo, R., Thorpe, K.: Using Reflective Learning Journals to Improve Individual and Team Performance. Team Performance Management: An International Journal 8(5/6), 134–139 (2002) 15. Macduff, I.: Using Blogs as a Teaching Tool. Negotiation Journal, 107–124 (2009) 16. Manouchehri, A.: Developing Teaching Knowledge Through Peer Discourse. Teaching and Teacher Education 18, 715–737 (2002) 17. Morrison, K.: Developing Reflective Practice In Higher Degree Students Through A Learning Journal. Studies in Higher Education 21(3), 317–332 (1996) 18. O’Keeffe, M., Arnedillo-Sánchez, I., Flanagan, A.: An Investigation into the Use Of Weblogs For Reflection In Learning. International Conference e-Society (2006) 19. Oravec, J.: Blending By Blogging: Weblogs In Blended Learning Environments. Journal of Educational Media 28, 225–233 (2003) 20. Paulus, T.M., Payne, R., Jahns, L.: Am I Making Sense Here?: What Blogging Reveals About Undergraduate student Understanding. Journal of Interactive Online Learning 8(1) (2009)

Promoting Reflective Learning: The Role of Blogs in the Classroom

11

21. Stahl, G.: Building Collaborative Knowing: Elements Of A Social Theory Of Learning. In: Strijbos, J.W., Kirschner, P., Martens, R. (eds.) What we know about CSCL in higher education. Kluwer, Amsterdam (2003) 22. Stiler, G., Philleo, T.: Blogging And Blogspots: An Alternative Format For Encouraging Reflective Practice Among Preservice Teachers. Education 123(4), 789 (2003) 23. Thorpe, K.: Reflective Learning Journals: From Concept To Practice. Reflective Practice 5(93), 327–343 (2004) 24. Wagner, C.: Put Another (B)Log On The Wire: Publishing Learning Logs As Weblogs. Journal of Information Systems Education 14(2), 131–132 (2003) 25. West, R.E., Wright, G.A., Graham, C.R.: Blogs, Wikis, And Aggregators: A New Vocabulary For Promoting Reflection And Collaboration In A Preservice Technology Integration Course. In: Crawford, C., et al. (eds.) Proceedings of the Society for Information Technology and Teacher Education, pp. 1653–1658. Association for the Advancement of Computing in Education, Chesapeake (2005) 26. Williams, J.B., Jacobs, J.: Exploring the use of blogs as learning spaces in the higher education sector. Australasian Journal of Educational Technology 20(2), 232–247 (2004)

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher Sung-Wook Baek, Jong-Hoon Seo, and Tack-Don Han Dept. of Computer Science, Yonsei University, 134, Seodaemun-Gu, Seoul, 120-749, Republic of Korea {yoshiboarder,jonghoon.seo,hantack}@msl.yonsei.ac.kr

Abstract. In this paper, we focused on a research on a Location-Based Social Network Service to increase social interactions among people who are engaging in research and development. Since existing many of location-based social network services have only provided information of regions and social relations of users have targeted an unspecific majority of groups, they could not have active interactions, compared to users of online communities who have particular use purposes. ConAR what we designed to get over these weaknesses, aims at researches to academically exchange with others and it enables them to have social interactions with people of the same interests by helping them select discussion topics and decide assignment places simply and on impulse. Keywords: Social Computing, Social Interaction, Location-Based Service, Context-Aware Service, Communities, Human Computer Interaction.

1 Introduction Social Networking Service is useful to not only keep friendships with acquaintances, but also make new relationships. Because it is an online service, it is also possible to keep friendships with friends who live far away using this. This service is gradually developing into a variety of forms. For example, Twitter is now used to get news or information useful for life and Academia or Research Gate provides services like theses or projects of researchers for users for an academic exchange. In addition, Location-Based Social Networking Services using location information have been developed with an invention of Smart-phone. In particular, Location-Based Social Networking Service provides not only information of regions like restaurants and transportation, but social networking functions among people who live in the same regions. In spite of these services, however, it is not easy to make social relations with people who have common interests and social networking doesn't work as actively as online community services, because it still aims at an unspecific majority of people. The following scenarios suggest problems of the current Social Networking Services: Sung-Wook in a graduate school has a lot of interests in HCI. He opens research achievements and projects of named researchers in the HCI field and exchanges messages with them using an online Social Networking Service provided A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 12–19, 2011. © Springer-Verlag Berlin Heidelberg 2011

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher

13

on Academia. One day, Sung-Wook participated in an academic conference held in COEX, Seoul to make a presentation of his thesis. He wanted to discuss with people who are interested in SNS like him, so tried to find them through his Location-Based Social Networking Service, but it was not easy for him to go forward and speak to them, because only a minority of people had interests in it and he had no idea of their research achievements and participation purposes in the conference. Here, we can extract two main problems from the above examples. Firstly, many of the current social networking services including Academia cannot be practically used for discussion in a conference or office or laboratory. Secondly, they don't provide social networking services for an academic exchange of users. ConAR has three characteristics as following. Firstly, it shows profiles and research performances of researchers around users based on their locations. Here, we designed it to help users search information what they really want by extracting information required for cooperation of researchers through a survey targeting researchers of companies and schools. Secondly, users can do online or offline social interactions, because it provides them with Contextual Information of participation possibility in offline meetings or meeting places where they prefer. Lastly, using 'ConAR', users can choose discussion topics by themselves and decide assignment places simply and promptly.

2 Methods To induce necessary information and functions for ConAR service, we performed a survey regarding 17 professional R&D researchers in graduate school and company. Based on such result, we induced necessary personal information and service to have scholarly exchange, and by using rapid prototyping tool, we designed UX of ConAR. Finally, to estimate the usefulness and usability of ConAR, we realized using Android SDK. For the estimation, we invited 12 researchers, and they belong to graduate school and company R&D center. The reason for limiting job like this is because the purpose of using ConAR is scholarly exchange in academy, school, or company, and actually, as a result of analyzing the job of academy participants, it was classified mainly as student, professor, and company researcher. 12 researchers were divided into 2 groups in 6 people, and those with no daily friendly relationship were chosen to estimate whether social interaction using ConAR is held well. 2.1 Personal Information Regarding personal information related to users, survey was made based on the information opened in Korea's graduate school R&D room, and 17 professionals participated in it. Table 1 is a survey investigation result regarding personal information of researchers needed for scholarly exchange. Through Table 1 we could largely check two kinds of information. Firstly, the answer that Social Networking Service is needed in scholarly exchange was 3.7 in average, so it could be checked that it was positive. Secondly, regarding the most important thing in scholarly exchange, F. Consider the area of research was 4.2 in average and it was the highest, and I. Consider the project was 3.9 in average, and J. Consider the experience of field was 3.8 in average in the greatest order.

14

S.-W. Baek, J.-H. Seo, and T.-D. Han Table 1. The result of survey about researcher's information for academical interaction

Number 1 2

Attribute

Evaluation Tool

Do you consider using SNS for the academic interaction? What kind of researcher's information do you consider when you do the academic interaction?

Negative 1 2 3 4 5 Positive

A

Consider the degree

B

Consider the personality

C

Consider the advisor

D

Consider the paper (Journal)

E

Consider the sex

F

Consider the area of research

G

Consider the social position

H

Consider the lab

I

Consider the project

J

Consider the experience of field

Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive Negative 1 2 3 4 5 Positive

Fig. 1. Personal Information of ConAR

Average

Response Rate

3.7

93.7%

3.1

100%

3.5

100%

3.5

100%

3.5

100%

1.8

100%

4.2

100%

2.5

100%

3.0

100%

3.9

100%

3.8

100%

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher

15

Contrarily, regarding scholarly exchange, what is not important was E. Consider the sex with 1.8 in average which is the lowest, and G. Consider the social position was 2.5 in average which the second lowest. We, through this result, made the Personal Information which will be used in ConAR as in Fig 1. Looking at Fig 1, not only personal information of investigated researcher, but also location of researcher, current status information, and meeting places they liked were also included. Researchers' personal information is data which is very much needed when searching for the people one wants have scholarly exchange through ConAR, and meeting place they like or current status of the self (in the paper, we will write this information as Contextual Information) and location information are data which is needed for meeting to be held in the real world. In the information above, information like Contextual Information can be a problem related to privacy invasion. So, we provide Privacy protection function which can protect location of the self and status information for researchers. In case this function is activated, Contextual Information of the self can be opened, and in case of not activated status, no such Contextual Information is opened. 2.2 Service To analyze necessary service for scholarly exchange, we also held survey regarding service during the survey in the above. We classified the items regarding service into 5 kinds, and we made the survey so that necessary 3 kinds in the greatest order could be answered. Table 2 is the following result. Table 2. The result of survey about service for academical interaction Number

Attribute Researcher management

Preference Service Researcher search

Researcher registration

4

Communication ability

Sending a note

Conversation request

5

Individual information protection function

6 7

3

Conversation rejection

Profile management

Research material management function

Paper search

File sending

Location Based Service function

Share the location

Map service

Researcher management Opening the Community Share my status Cooperation service Recommendation for place of appointment

As shown in Table 2, service which is needed for scholarly exchange is investigated by dividing it into 5 items. For example, the necessary function for researcher management service was preferred in the order of searching researcher, registering researcher, and managing researcher. Based on such result, we composed the service which will be provided in ConAR and realized necessary functions for these services.

16

S.-W. Baek, J.-H. Seo, and T.-D. Han

2.3 Prototyping Based on the investigated survey result above, we designed UX of ConAR in the method of Paper Prototyping. To use ConAR at the first time, user registration course should be held, and the course is expressed in Fig 2.

Fig. 2. User Registration Course Scenario

Because ConAR is a service based on location, personal privacy problem can occur. So, to previously prevent the problem, we composed a scenario as follows. Firstly, in case of using it for the first time, after user registration is held, Privacy function can be checked, and if it is Off, it will show an alert message so that no one would subconsciously open location information and status information. Even if a user is already registered, before starting ConAR service, the Privacy function is always checked and alert message is shown. Next is the scenario regarding searching researcher, MAP View, registering meeting place which is preferred, researcher management, personal profile management.

Fig. 3. ConAR's Scenario regarding Service

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher

17

To express each 5 kind of ConAR service in mobile environment, we designed a scenario as in Fig 3. Using tab controller, we designed User Interface, and explanation regarding each tab as follows. Firstly, in case of Research Tab, it provides searching for researcher to have scholarly exchange. For example, it shows related researcher by searching Paper, Project, or studying field. In case of Map Tab, it shows researchers around using ConAR service through Google Map. If Privacy mode in the Off state, it is not signed in the Map. In case of Place Tab, it searches preferred meeting place and provides registration function. Meeting place such as library, coffe shop, vending machine, and etc can be chosen, and based on the current location of the self, it searches the location by searching close place. In case of Coworker Tab, it shows the list of people who went through scholarly exchange (for example, when giving and taking message was held or conversation was held by real meeting) through ConAR. Also, if researcher is chosen, functions which can make scholarly exchange are provided. Finally, in case of Profile Tab, personal information of the self can be changed, and also, On/Off setting regarding Privacy can be done. 2.4 Procedure In order to assess whether ConAR is helpful for an academic exchange of researchers, we conducted an experiment by dividing 12 participants into Group A and B consisting of six members separately. First of all, A group was organized with 4 of introvert people and 2 of active people, and it was observed if there was a social interaction among researchers by setting them to join a meeting actually in Contextual Information. On the contrary, B group was composed of 2 of introvert people and 4 of active people, and it was observed as well if there was a social interaction among them by setting them to be busy in Contextual Information.

Fig. 4. Experimenter estimating usefulness and usability of ConAR service

2.5 Results Fig 5 is a task process of ConAR that participants carried out. Although, Group A and B had equal processes in contacting researchers who they wanted to work with, using the Matching service of ConAR, the places for their social interaction were different between Group A and B depending on Contextual Information of researchers. For A group that was set to join the actual meeting in Contextual Information, participants

18

S.-W. Baek, J.-H. Seo, and T.-D. Han

Virtual World

Virtual World B Group Searching the researcher

Get recommendation of places for the appointment

Real World A Group

Check the users current status

Fig. 5. Social interactions in the virtual world and the real world

had 5 total meetings actually out of 8 attempts, but for B group that was set to be busy in Contextual Information had 9 attempts, they had a total of 8 social interactions using a message delivery function online.

3 Conclusion and Future Work We made participants have an academic exchange by carrying out a research on a location-based social networking service. ConAR connected researchers to other people who have the same interests to them through the Matching Function and made their academic exchange work smoothly in a conference or school or office by simplifying the process of the actual meeting. We also devised a social interaction considering the current conditions of researchers using Contextual Information. Lastly, we provided a location-based social networking service which is very useful for researchers based on online community, instant message delivery and User Management functions to keep their academic exchange continual. In addition to this experiment, we will carry out a study on improvement of cooperation of researchers in the actual meeting. Although, existing Matching services have been developed to enable researchers that are motivated to meet the opponents, to cooperate with each other in the actual meeting, other services should be developed for an online social interaction. Therefore, we will develop services providing functions like brain storming, voice recording or video filming and white board used in discussion places.

References 1. Wang, F.-Y., Zeng, D., Carley, K.M., Mao, W.: Social Computing: From Social Informatics to Social Intelligence. IEEE Computer Society, Los Alamitos (2007) 2. Yamamoto, D., Takumi, I., Matsuo, H.: Location-Based Social Network Services Employing Student Cards for University. ACMGIS (2009)

Meet Researcher in the Real World Using the ConAR: Context-Aware Researcher

19

3. You, C.-W., Chen, Y.-L., Cheng, W.-H., Chen, M.-S., Sinica, A., Tsai, S.-A.: SocialCRC: A Social-and Context-Aware Rendezvous Coordination System. In: SIGCHI 2010 (2010) 4. Cummings, J.N., Butler, B., Kraut, R.: The Quality Of Online Social Relationships. Communications of the ACM (2002) 5. Zhao, D., Rosson, M.B.: How and Why People Twitter: The Role that Micro-blogging Plays in Informal Communication at Work. In: ACM 2009 International Conference on Supporting Group Work (2009) 6. Karimi, H.A., Zimmerman, B., Ozcelik, A., Roongpiboonsopit, D.: SoNavNet: A Framework for Social Navigation Networks. In: ACMGIS 2009 (2009)

Gaining Insight into the Application of Second Life in a Computing Course: Students’ Perspectives James Braman1, Alfreda Dudley1, Kidd Colt2, Giovanni Vincenti1, and Yuanqiong Wang1 1 Towson University Dept. of Computer and Information Sciences 8000 York Rd. Towson, MD, USA {jbraman,adudley,gvincenti,ywang}@towson.edu 2 Independent Consultant Sarasota, Florida, USA [email protected]

Abstract. As educational delivery methods have changed over the years, often through enhancement by technology, new and innovative strategies have been introduced to enhance learning. Virtual worlds such as Second Life are being used as an educational tool in many domains and for a variety of purposes. To gain a further understanding of the use of virtual worlds in the classroom from a student’s perspective, the authors discuss the feedback gained from students about their participation utilizing Second Life in a computer course. Pre and post surveys were distributed to the class to get student’s feedback. The responses from these surveys are reported in this paper. Keywords: Virtual Worlds, Second Life, Education, Perception, Educational Tools.

1 Introduction As virtual worlds and their technological counterparts (i.e., video games, interactive multimedia and other forms of virtual and augmented realities) become more common place, they can be used for much more than entertainment purposes alone. These technologies can be used to extend the capabilities of the traditional classroom to create a more immersive and interactive educational environment. As we begin to increase the use of virtual worlds as part of instruction, it is important to assess the perception of effectiveness as a learning tool by the students. Perception of a tool such as virtual worlds can often affect its use and adoption in an educational context [1]. The application of this particular technology in the classroom can have an impact on how students and educators react to it, and their willingness to use it for courserelated activities. In this paper, the authors outline their preliminary work and discuss feedback collected from students regarding their participation in Second Life© in a computer course. Second Life (SL), like many other virtual worlds, is a 3-dimensional internetbased interactive space where users can communicate in real-time through visual A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 20–29, 2011. © Springer-Verlag Berlin Heidelberg 2011

Gaining Insight into the Application of Second Life in a Computing Course

21

representations of themselves, called avatars. It is through these avatars that users can build objects, socialize, interact within the environment and “view” the world. Avatars can be created in many different forms and representations, allowing for a diverse variety of visual representations [2]. Unlike many variants of virtual reality, Second Life does not require expensive gear to interact within its environment; one can use the platform by using traditional hardware (i.e. keyboard, mouse, monitor, headsets) increasing access to average users. Students can use Second life in the classroom or on their home computer for class activities or at-home projects. The content within Second Life is almost entirely created by the users, where user creativity is encouraged. Built-in modeling tools can be used to create detailed objects and shapes enhancing many ideas and concepts. One can even use many different types of external programs like Blender, Plopp or 3D Studio Max, to create textures and elaborate graphics to enhance designs. Virtual goods and services can also be bought and sold using Second Life’s virtual currency called the Linden dollar which can be exchanged for real US currency on the Lindex Market exchange, enabling users to link real accounts to their virtual accounts. Second Life and similar platforms have often been referred to as a “3D-Web” and pointed out as a possibility of the future of online interaction. Teaching inside a virtual world not only provides students the means to be immersed in real-time scenarios that can be customized, but also allocates to them opportunities to interact with an online society that emerged from the technology they study in classes. Virtual classrooms are engaging to students as “Students engaged in educational games and simulations are interpreting, analyzing, discovering, evaluating, acting, and problem solving” [3, p. 116]. These virtual spaces provide a means to model complex systems and facilitate role-play simulations at a low cost [4]. Second Life has been used in many educational contexts across various disciplines in order to support and enhance learning [5]. In previous studies, students generally agreed that they did see potential educational benefits of using Second Life to enhance learning [6]. This led the authors to conduct further studies on the impact of virtual worlds/reality in educational contexts. In a related study, the professors used Second Life as a tool in several computer courses to teach students about decision making through virtual role playing activities with encouraging results [7]. In addition, Second Life was used to demonstrate the use of virtual worlds in real life case scenarios [8]. The authors wanted to gain insight on students’ responses on virtual worlds to begin investigations revealing concerns students have about using these technologies in the classroom. In the next section the surveys are discussed.

2 Methods In order to begin to assess students’ perceptions related to potential concerns of course related use of Second Life, two surveys were administered to students enrolled in a course titled Computers and Creativity during the Fall of 2009. The main goal of this course is to teach students basic computer skills while emphasizing multimedia and the creative nature of technology (i.e. flash animation, web design and basics of multimedia applications). The general structure of the class contains both lab and lecture components. The course content and creative goals correlate with the

22

J. Braman et al.

expressive power and creative nature of Second Life. Activities for this class included in-world student projects, virtual field trips, basic building and scripting. Second Life is also the basis for many class discussions. Second Life based Machinima (filming in 3D worlds) has also been incorporated into class discussion topics to illustrate filming techniques and video editing [9]. The typical audience of this course includes first and second year students from all fields of study. Three sections of the Computers and Creativity course were given the pre and post surveys, with a total enrollment of 92 students. The purpose of pre-Second Life survey was to assess students’ preconceived ideas about virtual worlds. At this point in the class, no information, discussions or labs had been related to virtual worlds or Second life. After this initial survey, the class spent two sessions discussing various aspects of virtual worlds and their impact on the real world. In addition, the class spent two lab sessions in Second Life learning to build objects, visiting areas of the environment and learning the features of the platform. After these Second Life class activities, the post-SL survey was administered to the same class sections to see if there was an overall change in student perception. The questions were the same as in the pre-SL survey, except for a few additional questions that were asked related to avatar appearance and their overall thoughts on Second Life. The pre-Second Life survey consisted of five questions and the post-Second Life survey consisted of nine questions.

3 Pre-Second Life (SL) Survey From the enrolled students, 67 students responded to the survey for a 72.8% response rate. The average age for those responding was 19.8, with a male to female ratio of 28 male (41.8%) and 39 female (58.2%). Students had a variety of backgrounds when it came to declared majors and areas of study. Question #1 asked “Do you have any prior experience with any 3D virtual worlds? (For Instance, Second Life, Sims online, WoW...etc) If so, which ones.” From the responses, 9% answered “Yes” (6) and 91% (61) reported “No”. Those that did note some experience, students noted that they had experience in virtual worlds such as: The Sims, Guildwars, World of WarCraft, Maplestory, IMVU, Gates of Andaron, and through Xbox Live. Question #2 asked “Have you ever taken a class that has used a virtual world as part of classroom instruction or as a learning tool?”. The majority responded “No”, 97% (65) and 3% (2) responded “Yes”. Of the two students who responded yes; one reported that they had participated in a class exercises for a business class in high school and the other noted they had been involved in an activity for an accounting class, but did not mention the actual environment. Question #3 asked “List any ethical concerns you have about Second Life or other virtual worlds”. In order to assess initial thoughts on the subject, no information was given up to this point on these types of issues or anything related to computer ethics. From the respondents, 38 students (56.7%) gave no answer or wrote “no” or “none” on the survey. Table 1 lists student concerns and its respective frequency. Some students responded to more than one item. A post-SL frequency column has also been included to compare pre and post SL results. More detail about the post-SL questionnaire is documented in the later section.

Gaining Insight into the Application of Second Life in a Computing Course

23

Table 1. Concerns about Second Life or other Virtual Worlds Concern

Privacy Concerns Detriment to Real Life Social Skills Identity Theft Social Issues/ Social problems (General) Virtual Crime Security Addiction Copyright infringement Not being taken seriously Online Stalkers Encountering Adult content “It scares me” Encountering Users with Malicious Intent Speaking with Strangers False Appearances General Safety Online

Frequency (Pre) 29 respondents 43.3% 13 7 4 2 2 2 2 2 1 1 1 1 0 0 0 0

Frequency (Post) 22 respondents 31.4% 8 3 2 3 0 1 3 1 1 0 0 0 1 1 1 1

Table 2. General Student Comments on Online Behavior General Student Comments on Online Behavior “Anonymity allows people to do and say anything” “Might be for those shy in RL” [Real Life] “Yes, it’s more informal” “Yes, you can be what you want online” “Maybe because no one knows me” “No, being online is just another way to be yourself” “Possibly I would be more social” “Not real, so I may act out differently” “Maybe because I could do anything without people knowing me”

We asked students in Question # 4 “Do you think your behavior online would be different than your behavior in Real Life? Why”. From the respondents 34 students (50.7%) answered “Yes” while the 23 students (34.3%) said “No” and the remaining 10 (15%) answered “Maybe”. Table 2 above highlights the general comments associated with answers for this question from those who gave a brief explanation. As a last question (Question #5), in order to understand what types of issues the students generally thought were important, five general concerns were listed randomly, which included: “Social Issues”, “Privacy Concerns”, “Network and Security Issues”, “Intellectual Property/ Copyright Issues” , and “Computer Crime Issues”. From the list of choices, students were asked to order these concerns from 1 (highest) to 5 (lowest). Table 3 lists the concerns and what the average ranking score was for all students. All respondents answered this question.

24

J. Braman et al. Table 3. Average Ranking of Listed Concerns for Pre-SL Listed Concern Privacy Concerns Networks and Security Issues Computer Crime Issues Social Issues Intellectual Property/ Copyright Issues

Average Ranking 2.06 2.54 3.27 3.45 3.63

4 Post-Second Life (SL) Survey After these Second Life class activities, a secondary post-SL survey was administered to the same class sections to see if there was an overall change in student perception. The questions were the same as in the pretest, except for a few additional questions that were asked related to avatar appearance and their overall thoughts on Second Life. For the post Second Life questionnaire, nine main questions were asked of the students. There were 70 student respondents yielding a 76% response rate from the total 92 possible students. The Average age from these students was 19.95 years old, with 35 students identified as “Male” (50%) and 35 students as “Female” (50%). Question #1 was in the form of a table, where students could select their choice for each question as presented in Table 4. The results of the student’s responses are summarized in Table 4. Table 4. General Reactions to Second Life for Class Activity Question Would you be interested in using Second Life for a class Lab / activity in the future? Do you think virtual environments could have an educational benefit? Rather use another tool / virtual world other than SL Enjoyed using SL in class Intend to use SL in the future (outside of class)

YES

NO

Maybe

41

15

14

37

0

29

8 59 10

31 5 34

31 6 26

Table 5. Avatar Appearance Question How important is your avatar's appearance? How important is the appearance of your friend's avatar? How important is the appearance of a stranger's avatar? How important is the appearance of a significant others/ partner' avatar?

Very Important 4

Important 15

Somewhat Important 32

Not Important 19

1

3

35

31

1

7

30

32

3

14

32

21

Gaining Insight into the Application of Second Life in a Computing Course

25

Question #2 was also presented in the form of a table where students could select the appropriate response. The authors were interested in the perceptions students had on the appearances of other avatars. Specifically, we wanted to know if students felt threatened by the appearances of anonymous users, or if they had an affinity towards higher quality avatars for making friends [10]. Additional research is needed in this area. Table 5 presents the results of the general questions of the importance of appearance. In Question #3, students were also asked “Do you think you could feel threatened by the appearance of another avatar?” 9 students responded “Yes” (13%), 53 responded “No” (76%) and the remaining reported “Maybe” (11%). Some students commented on appearance stating that “Some appearance[s] could be offensive” and that appearance could relate to “stereotypes of how people appear”. One student commented that avatar appearance was something they did not feel was important because interaction in SL was “No different than an IM or chat session”. In Question #4, Students were asked about their general feeling of safety while using Second Life. “How would you rate your level of safety while in Second Life? (1-Unsafe/threatened: 10-Very Safe)”. It was mentioned in class about safety concerns of virtual worlds (as with other online environments) not to reveal personal information and also to choose avatar names that were dissimilar to their real-life name and not to choose words that could identify them or their location. When asked to rate their level of safety on a scale of 1 to 10 (1-Unsafe/threatened: 10-Very Safe), from the 70 respondents the average rating score was a 7.74. In Question #5, students were asked to list any difficulties they had with the SL platform. Many students left this question blank. For those that did have difficultly, Table 6 describes the problems noted. A frequency is noted for difficulties mentioned in some cases. These described difficulties are similar to previous studies [6]. Question #6 asked students “What did you like the most about Second Life?”. These items and frequency of responses are described in Table 7 below. Students were then asked in Question #7 what they generally disliked about Second Life. These items and frequency of responses are described in Table 8 below. Students were then asked in Question #7 what they generally disliked about Second Life. These items and frequency of responses are described in Table 8 below. Table 6. Difficulties in Second Life Difficulty Learning the controls Confusing for new users at first Appearance difficult to change/edit Didn’t know where to explore Flying Building things Needed more time to learn Teleporting Difficulty getting the concept Lag Awkward to control

Frequency 9 6 6 3 2 2 1 1 1 1 1

26

J. Braman et al. Table 7. Liked about Second Life General Comments Flying / Ability to Fly Being able to visit/ explore new places It was a different/ unique experience Freedom to be anyone / do anything Social aspects Graphics Fun Editing appearance Other avatar’s appearances Building Making a avatar Easy to use It’s more fun than being in a real class Whole other world online Like a game but with more interaction

Frequency 11 10 8 8 6 4 3 1 1 1 1 1 1 1 1

Table 8. Disliked about Second Life General Comments Controls are difficult Not into virtual worlds Lag Trying to figure out where to go / explore Building Too much to understand Hard to focus on the task I was bored Graphics Rather create things in real life Security could be an issue Different feel safe or protected Creating an avatar/account Movement Complicated Hard to understand Wanted more time in class Everyone in class can see what you are doing Seems overwhelming Couldn’t change my clothes Rude people Characters look unreal “It scared me a little bit”

Frequency 5 4 3 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

The authors again wanted to see what types of concerns students had about Second Life or using virtual worlds and posed it as another question in the post-SL survey. Question #8 asked “List any ethical concerns you have about Second Life or other virtual worlds:” The results from the pre-SL questionnaire and post-SL questionnaire were different. To compare both results refer to Table 1.

Gaining Insight into the Application of Second Life in a Computing Course

27

In Question #9, the authors wanted to see what types of concerns from the same list as mentioned previously in question #5 in the pre-SL survey, was most important to the students after using Second Life. Again the same five general issues were listed randomly. Students were asked to order these concerns from 1 (highest) to 5 (lowest). Table 9 lists the concerns and what the average ranking score was for all students. All respondents answered this question. The only main change in the average ranking of these general concerns regarding differences were between “Social Issues” and “Intellectual Property/ Copyright Issues”. Previously Social Issues were initially ranked higher, but in the second assessment, it was listed as the least issue of concern. The other issues remained in the same order. The importance of students’ perceived concerns generally remained the same before and after interacting within Second Life. Table 3 compares the results from pre and post-SL questionnaire. Table 9. Average Ranking of Listed Concerns for Post-SL Listed Concern Privacy Concerns Network and Security Issues Computer Crime Issues Intellectual Property/ Copyright Issues Social Issues

Frequency 2.11 2.97 3.09 3.16 3.66

5 Summary of Survey Results As indicated in the above reported results, the feedback from the pre and post survey questionnaires provided basic insight on the impact of using Second Life as an instructional tool. In addition, the responses of the students indicated their perceptions on the application and use of this software. The purpose of this study was for instructors to gain insight on how effective the inclusion of Second Life technology impacted their computing course. However, the instructors were aware of some limitations surrounding this study. For example, the questions on the pre and post surveys can be more in-depth to cover more pedagogical issues; inclusion of analyses and testing of the data would have yielded stronger results; and a reporting of larger population (different types of computing courses) may have also yielded interesting results. In future investigations the authors propose by means of a pre and post SL survey to compare two separate groups of students with differences in class activities to see if there is a significant difference in responses. A major strength of this study is the feedback from students’ perception of using this technology. It is the opinions of the authors that while this study has limitations, it adds value for further examination in this area. The feedback did yield interesting results which demonstrated that there were concerns related to the perception of virtual words as reported in Table 1. These main concerns were still present after interacting with Second Life, but reporting of these concerns was much less in the post-SL survey. In addition, as outlined by Table 3 and Table 9, the ranking of importance of the randomly selected issues, remained similar with both the pre and post surveys. Using this preliminary data as a starting point, we plan to conduct more in-depth studies on concerns related to virtual world usage.

28

J. Braman et al.

6 Conclusions As technology-enhanced learning increases in the classroom coupled with virtual worlds and other computer mediated environments, it is important to gain insight on the best practices using these types of mediums for the benefit of both students and educators. Cultivating on our experiences with implementing Second Life activities in the classroom from several types of activities has encouraged experimentation and potential improvement with these types of technologies. Not only has feedback from students been generally positive, usage in the classroom has generally been encouraging as students are interested in being engaged in learning from within a 3D space. Using the results of this study as a preliminary baseline on the inclusion of Second Life in classroom instruction, we can begin to enhance our in-world activities by trying to overcome potential barriers; such as, misconceptions and concerns or technical / usability problems. The feedback from this study suggests that there are still many concerns that are held by students about interacting in a 3D world and it may take some time to dispel some of the concerns. While there is still much needed research in this area, from both a student and instructor point of view regarding overall successfulness in long term studies. Many of the questions posed in our surveys simply serve as a starting point for other research initiatives by the authors, as it is our intention to explore other areas of education and social interaction through virtual worlds. It is out hope that other educators will explore virtual worlds.

References 1. Laughlin, D.: Overcoming Objections to MUVEs in Education. In: Vincenti, G., Braman, J. (eds.) Teaching through Multi-User Virtual Environments: Applying Dynamic Elements to the Modern Classroom. Information Science Reference, Hershey (2010) 2. McArthur, A.: Real Ethics in a Virtual World. In: Proceedings of CHI 2008 (2008) 3. Antonacci, D.M., Modress, N.: Envisioning the Educational Possibilities of User-Created Virtual Worlds. AACE Journal 16(2), 115–126 (2008) 4. CMP Media LLC.: Second Life Tries for a Second Act The online virtual world has weathered a boom-and-bust cycle. Now it has a New CEO. Can this Social Networking Phenom Attract a New Wave of Consumers and Enterprise Users? (2008), http://www.commonwebnews.com (retrieved) 5. Vincenti, G., Braman, J.: Teaching through Multi-User Virtual Environments: Applying Dynamic Elements to the Modern Classroom. Information Science Reference, Hershey (2010) 6. Wang, Y., Braman, J.: Extending the Classroom through Second Life. Journal of Information Systems Education. Special Issue: Impacts of Web 2.0 and Virtual World Technologies on IS Education 20(2) (2009) 7. Wang, Y., Dudley, A., Braman, J., Vincenti, G.: Simulating Ethical Dilemmas: Teaching Ethics through Immersive Virtual Environments. In: Proceedings of the 12th IASTED International Conference on Computers and Advanced Technology in Education (2009)

Gaining Insight into the Application of Second Life in a Computing Course

29

8. Dudley, A., Braman, J., Wang, Y., Vincenti, G., Tupper, D.: Security, Legal, and Ethical Implications of using Virtual Worlds. In: Proceedings of the 14th World Multi-Conference on Systemics, Cybernetics and Informatics: WMSCI 2010 (2010) 9. Filimon, S.: Machinima. In: Braman, J., Vincenti, G., Trajkovski, G. (eds.) The Handbook of Research on Computational Arts and Creative Informatics, Information Science Reference, Hershey (2009) 10. Yee, N., Bailenson, J.: The Proteus Effect: The Effect of Transformed Self-Representation on Behavior. Human Communication Research 33(3), 271–290 (2007)

Third Agers and Social Networking in Higher Education Christopher M. Connolly and Gabriele Meiselwitz Applied Information Technology Computer and Information Science Towson University 8000 York Road, Towson, MD 410-704-2835 {cconno4,gmeiselwitz}@towson.edu

Abstract. With an aging population and the social networking boom significant research has been performed in three areas. The first is the challenge higher education institutions face integrating social networking sites in class offerings. Students are embracing this medium at an accelerated rate; however, the benefits of social networking for students are not always clear. Secondly, this paper will look into the trend of seniors re-entering institutions of higher education; especially the challenges relating to Web 2.0 technologies such as social networking sites. Although a large number of seniors are online today; many are not comfortable using social networking [20, 14]. Finally, studies relating to the use of social networking sites by third agers outside the classroom will lead to a better understanding of potential challenges seniors and instructors may experience when seniors decide to return to the higher education roles. Keywords: Social networks, computer science education, higher education, third age.

1 Introduction The population of the United States is aging. Recent demographic research has shown that within the next ten years the populace of third agers will become a full 20% of the American population [1]. Due to seniors being in better health today and staying active long after retirement; many decide to engage in occupational pursuits or return to higher education [2]. But as third agers re-enter the workforce as retirees, start their own business, or enroll in a university; they face new challenges that are going to transform how they perceive themselves and how they are perceived by others [2]. These challenges can include barriers linked to age such as limited mobility or cognitive problems; they can be attitudinal like age based discrimination and their own perceptions of themselves, and they may be structural like lack of transportation or financial considerations [2]. The PEW Internet and the American Life Project indicates that while the technology skills of many seniors may initially be lacking, once they are introduced, many become avid e-mailers, internet searchers, and gamers [2]. Social media and A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 30–39, 2011. © Springer-Verlag Berlin Heidelberg 2011

Third Agers and Social Networking in Higher Education

31

networking websites have seen the number of users aged fifty and over nearly double in the last year [2]. Although there is the potential for great benefit; the number of seniors who actually utilize technology solutions remains low when compared to other groups [2]. As of January 2010 38% of people age 65 and over use the internet; a relatively low percentage considering the national average of 74% for other age groups [2]. Another study performed by PEW investigated this increase in internet use by seniors. This particular survey showed that while all age groups experienced an increase; the percentage experienced by users 55 and over was more significant. The study measured the percentage of seniors that used the internet in 2000 and then again in 2010. The result was a dramatic increase, people aged 50 to 64 increased from about 48% in 2000 to 78% in 2010. Further, the study showed that the number of those over 64 accessing the internet nearly tripled from about 15% in 2000 to 42% in 2010 [2]. With numbers increasing at these rates they become a demographic segment that is available, and often has disposable income to make purchases, pursue educational goals, and subscribe to services. As third agers are making efforts to improve their lives; they are turning to their local community colleges and universities for support. They are doing this in record numbers causing the educational community to take notice and pursue new and better ways to support this age group [3]. In fact, the number of students who fall into the 50 and over group has increased by 144% [3]. This trend shows that these learners are going to be a wonderful resource, as both a student pool and even a group from which to recruit professors, for educational institutions in the future.

2 Social Networking in Higher Education A Social networking site is defined as a site that allows individuals to construct public or semi public profiles, articulate a list of other users to share with, and view and traverse their list and those made by others [4, 5, 6, 8]. What separates social networking sites is that they are built on these connections [8]. According to Social Networking and Education: Emerging Research within CSCL; the study of how students are using this software is still in its early stages and the results of these surveys show the practice of using social networking sites in higher education is developing and relatively informal [6]. In looking at how social networking sites should be used, a few suggestions were put forth by Ryan Alexander in Social Networking in Higher Education [5]. The first is that institutions should identify which social networking architectures are most successful and look into whether or not a particular architecture is suitable for their environment. One feature that has made these sites so successful is the ease with which users can post their content. There is no need for web design expertise anymore; users simply fill out a form. This is true for posting pictures and graphics, submitting a blog update, and sending an online message to a fellow student. A second suggestion is to examine other social networking sites that have experienced a high degree of success. Other suggestions are to encourage professors to work on their own Wikis, and have students contribute and create entries [5]. While many campuses have utilized

32

C.M. Connolly and G. Meiselwitz

these technologies others have not; but that has not stopped creative students from building their own micro-versions of these sites specifically for their class or group. There are also some downsides for universities not participating in this trend; information and research that the University has created may end up on one of these informal sites without that University’s review and/or approval. Universities also run the risk of having proprietary information appear to an unrestricted audience. Universities may look into the uses and benefits of participating in Web 2.0 technologies and find the rewards will far outweigh the costs and difficulties [5, 8]. In looking at how social networking sites can be used in higher education there are two schools of thought that are being investigated. One is that students have already garnered how to get the most out of social networking sites and that the study of how they are currently being used is most important [6]. This is informal research where a portal is created for students; then researchers study how students utilize it. The main point of these studies is to explore the opportunities that are being created by the explosion of use among students; with a focus on using social networking as a collaborative tool for improving student achievement [6]. Students received many benefits in the process of making these connections. Students are shown as using the sites to create and test versions of their own personality, create relationships, and as creative outlets; aside from the simple sharing of information and knowledge. Other benefits of social networking sites to learners include creating a social learning environment, creating a practicing community, and creating team building environments. These activities contribute to the learning in the traditional classroom by bringing it outside the classroom and class-time. One can not underestimate the importance of the social aspects of learning [7]. There are other factors that will influence the creation of these social networks. One is the institution itself and its policy toward social networking. Another would be the technology policy of the institution. Are there Terms of Use policies that create a sense of Big Brother over the shoulder? This will limit social interaction and the depth of the relationships and collaborative efforts being forged as students are reluctant to break rules or to be exposed to online monitors that, while university sanctioned, are not the intended recipient. The type and quality of the network at the institution can be a factor as network speeds may influence the performance of a social networking endeavor. If the speeds are adequate; there are other factors like network congestion and down time for maintenance or repairs users can expect to experience that may also affect use. Aside from the physical network; things like the current social climate, political atmosphere, and student emotions can create barriers to social networking. The other position that researchers take when looking at social networking sites in higher education is that the best method for using social networking sites has yet to be discovered [7,8]. The role that social network sites play in education is the focus of the paper by Liccardi [8]. One point that was noted was that social networking group size should be a focus of study as the optimal number has not been calculated [8]. Another point that should receive attention was the importance of the group skill-set when creating groups, since much of the work is done online and groups have their own online dynamics. [8]. Hamasaki and Takeda studied the phenomenon of the “matchmaker” or the “friend of a friend” [9]. In this method, a friend of a friend introduces two people that are unfamiliar with each other. The article states that the manner in which social

Third Agers and Social Networking in Higher Education

33

networking sites facilitate the creation of relationships can be helpful in understanding their application to the student learning process. Another challenge this paper discusses is the community of practice [10]. This research looks into the dynamics that occur when a group forms for the completion of a task. This paper poses the question: “What will be the effect of social software on the formation of these groups?” Cho investigated the way advanced computer users were using social networking sites to develop a collaborative learning environment [11]. The project was with a homogeneous group of students working on the design of an aerospace system. The researchers took two groups of geographically dispersed students and emphasized the importance of collaboration in the assignment; the researchers then created and delivered a social collaboration web site for them to work with. A survey was distributed at the beginning, and at the end of the study to measure any pre-existing networks. There were two hypotheses being tested, the first was those with a better social status, or closer to the center of the friend circle would perform better. The second was that those in the center would not be as inclined to make new connections. The result showed that both hypotheses of the research were supported. But for this paper the results are not as important as the amount of study that went into the research and the scientific method that was used in the definition of a social network friend group, the positions of the individuals in the group, and the effect that can be observed from this. Indeed, the true result of this study was the observance that those who take advantage of their social contacts tend to achieve better results. How they use the sites was not covered.

3 Seniors in Higher Education As seniors become more attuned to the possibilities of internet activities many experience a desire for enhancing their knowledge. The desire for education is a natural response once third agers are exposed to the opportunities for learning that permeate the internet. Education is not strictly about the traditional subjects such as reading and mathematics, it is also learning about entertainment, local and national news, causes and charitable works, sporting events, and informal discussion of people’s lives. Especially in the recent economic downturn; where the stock market lost 27% of its value and job losses are at record levels, many older Americans are experiencing difficult situations. Indeed, as of November 2010 more than one million people aged 55 and over are jobless. Although this seems like a difficult situation; this same population of people are experiencing better health and have had more, and better educational opportunities [12]. This study has also shown an increase in the participation of older adults in educational activities. In 1998, only 3.6% of the population of Slovenia was active in educational pursuits, whereas in 2004 that had risen to 12.1%. This trend is true elsewhere and America is no exception; third agers are returning to the roles. In Education: a Possibility for Empowering Older Adults, the authors examine the environment older adults experience as they re-enter higher education [13]. The actual

34

C.M. Connolly and G. Meiselwitz

study was related to the motivations and successes that older adults were experiencing in higher educational institutions. There were three of these studies performed to measure trends. The first was in 1987, and second in 1998, and the third and last in 2004. When seniors in 1987 and 1998 were surveyed about their motivations for more education they listed the “joy of learning” as their primary motivator. Other choices that had a significant response were “enable self-development” and “improve my situation”. The newer evidence does point to the fact that overwhelmingly older learners are returning to the educational roles from an intrinsic motivation. In 2004, the primary reason (64%) listed for pursuing education was a need for companionship. They were satisfying this need by becoming a classmate and teammember. The second most popular reason (61%) was the “joy of learning”, again indicating the value this demographic places on education. This paper goes on to state that it can be assumed from these results that these learners have had a lifelong love of learning and that they are simply continuing to indulge it. This is attributed to another result of the study that indicated those that generally listed “love of learning” as a motivator had higher educational levels. Another area that was questioned were the barriers that seniors experienced when trying to re-enter school. The important data here was older learners did not list a lack of programs as a barrier. The choice that was most common not having time, 36% in 1987, 35% in 1998 and 45% in 2004. Interestingly, the researchers did not trust this result and stated it was more likely a lack of income is the true barrier as the older population is at a higher risk of living below the poverty line and/or alone [13]. Other studies that have been performed look not to older people’s perceptions about activities or institutions but at their perception of themselves. In Forging New Identities: Older Adults in Higher Education the researchers looked first at the terms used to address third agers and found that most viewed being called “older” or “seniors” negatively [14]. These people have their wealth of experience, vast knowledge, and understanding of the human condition. The challenge for younger generations is to view older learners through a different lens, it means to do away with many of the old perceptions and see the aging population as living intentionally, learning, adapting and growing [13, 14]. Unfortunately, these adults are also influenced by societal perceptions of what they can and cannot do. This is closely tied to learning. If these older adults perceive themselves as “done learning” it may be because they feel they are too old to learn. Overcoming this invisible barrier has more to do with changing the way older adults think of themselves than dealing with their actual abilities. For these adults a solution is remaining engaged, whether through work, learning, or community. For many of them it means forging a new identity, or a new perception of themselves. This is an important concept because many people’s personal identity is established by the kind of work they do [14]. As a society we look at each other in terms of what we each contribute. We are not simply asking adults to take a few classes but really we are asking them to be open to the possibility of changing the way they see themselves at a fundamental level [14].

4 Seniors and Social Networking Many social networking sites are experiencing the explosion of users that are age 50 and over [2]. In fact, during the period from 2009 to 2010 social network use by

Third Agers and Social Networking in Higher Education

35

Americans over 50 increased from 22% to 42%. Their motivation can be traced to any number of personal causes, such as combating loneliness, trying to keep in touch with loved ones in other parts of the country, or just staying active and using these as tools to stay involved in their culture [15]. Still there is a great opportunity; for although the numbers are growing rapidly, there are a significant number of older people not getting involved. Social networking sites are fighting a battle of perception among many older people. Many perceive them as a little silly and empty headed [16]. A study by Lehtinen looked at the perceptions older adults have of social networking sites and after conducting surveys found that: Older adults find the internet unwelcoming, and that social networking is not important. Many seniors lack a reference community, that is, they would be more interested in joining a social group that communicated online if they were familiar with the people in it. This study suggests third agers are not inclined to simply go online and find a group of people to talk to. This study used Netlog, a social networking site that provided space for a profile, file sharing like pictures and video, friend lists, and blogs. The subjects were instructed how to use the site; but not specifically why to use the site. The findings indicated that the users largely did not use the services; there were some that experimented with the site and made some postings. Some of the reasons cited for not using the site were: • A simple lack of interest – users may have used computers for many years but stated they view them as tools, not sources of entertainment. • Some users only felt comfortable using computer systems when someone else was maintaining them – they used machines at work but felt they were more trustworthy because they were maintained by professionals. • Many other people their age do not use it – older users have trouble finding each other; and when they do they find they would rather chat on the phone than online. • It is “cold” – many older users responded that the interactions were impersonal and less engaging than other mediums of communications. When understanding older adults entering higher education it cannot be assumed that they will immediately adopt all the technological communications mediums that are routinely used by students in their late teens and twenties. While older users may be more inclined to use a social networking site to meet with a study group or share notes, once the class task or assignment is complete it cannot be assumed they will maintain their account or remain active on that site [16]. It is important to realize that social networking in education is not just occurring through the classic internet site. An important social networking tool group is highlighted by Twitter and other mobile based applications. When looking at the senior population and taking into account the physical limitations they experience, texting (a life-skill for younger users) can be very difficult. The application of SMS by older people lags behind other generational groups [17]. Research is being performed into how to increase the adoption of texting by older people. It should result in some changes to the existing technology or new technologies that will make SMS (Simple Message System) easier to use. When the current twenty-something population ages, they may find texting has become an antiquated method of communication; but that does not necessarily mean they will wish to abandon it. They may simply need a modification to the way it is performed.

36

C.M. Connolly and G. Meiselwitz

Relating to this topic, social sites like Twitter require knowledge of the mobile device’s SMS system and how to send and receive messages. These are two options that have yet to be truly adapted for the physical limitations of older users.

5 Conclusion and Further Research Educators are working to find ways to engage third agers, and are dealing with a series of steps that can be taken to improve accessibility and acceptance [13, 14]. The first task is to identify the factors that older adults list as their objections to using social networking sites. A few were described as: They are simply not interested. The answer is motivation; however, it is important to reach out to your audience correctly with a sense of enthusiasm and excitement. At the same time it is equally important not to become pushy. As DeMicco discusses in Motivation for Social Networking at Work; “There has been an emergence of social networking sites that are targeted towards specific groups” [18]. The proper focus can be used as a motivating point for older users. Many listed a reason for their lack of interest was there was no one on the site they knew. Focusing on this audience will allow more users to feel the site was created for them, incline them to give it a try, and perhaps create regular users. Other lacks of motivation are simply seeing the sites a frivolous or not meaningful. However, another article in Seniors Online states that seniors most often go online to engage in social activities [19]. It then states that those seniors that have active social calendars and lives say that they have a higher quality of life. Education is part of this social calendar and therefore part of that quality of life. Studies performed by the PEW research institute state that when seniors go online they are also looking for information about health care, news, and events [3]. Online communities that target that population present an excellent opportunity to accomplish these activities all in one place. Many seniors may find great benefit from the social interaction they would experience if they could be convinced to access, and had the ability to use, social networking sites. They simply require an introduction to the information available. Many social networking sites do not have enough accessibility options for those with physical impairments [20]. In early November 2010 the social networking giant Facebook reduced the size of its font for status updates and names. This is easily corrected if the user knows to hold the Control key while hitting the plus sign (+); but that is not indicated anywhere on the site. There are many online resources detailing the process but how are older users to know to go there? Research into making sites easier to access for older adults is not new, but newer technologies like touch-screens and tabletop computers can make things easier to use by expanding the screen and removing the need for the double-click [21]. What can be encouraged, that may increase use of social networking sites among third agers, is the publicizing of these aids so seniors may find more and better solutions to accessibility problems. Privacy is a concern for older adults. Many consider the internet a frightening and unfriendly place. They are concerned about their information falling into the wrong hands and doubly concerned about how the information they put on the internet may be used. Consequently they are reluctant to join a site and provide information over a connection regardless of the privacy controls and security in place [22]. While the information and the privacy issues are well addressed by the creators of the big social

Third Agers and Social Networking in Higher Education

37

networking sites, it is important to educate the users, in layman terms, how to secure their profiles. Then inform them of policies that are in place and software engaged to make sure that the information they post to their friends is not going to be taken and used against them. It is also important to make sure that the company providing the social networking site lives up to these promises. Establishing and cultivating trust is the most important ingredient when asking people who are naturally mistrustful of the medium to embrace it. Once the barriers to using the medium have been addressed the second step is to introduce the benefits of the medium to these users. One of the benefits is how the collaborative nature of the medium can create new communities for seniors to join and enjoy. As third agers have expressed they are “not done yet”. They may be up for new challenges and embrace education as a terrific and rewarding undertaking [13, 14]. When they engage they will meet new people and expand their knowledge. They will develop new social structures and relationships as well as engage and form new identities in a new environment [14]. Many enjoy a love of learning and can turn what some see as a great challenge into an enjoyable pastime that is its own reward [13]. The possible benefit here is the enhancement of their lives which, although it is a more intrinsic and less tangible benefit, can be a powerful motivator. Many third agers enjoy feeling useful and providing important contributions to society after they have retired. Education, and the social networking involved, gives them a chance to work toward their goals. An important benefit experienced by their younger student peers will be the sharing of their considerable life experience in the classroom [14]. Further research can be done into the terminology that these sites use and the conflicts with pre-existing notions those seniors have for what each word means and what the site is asking them to do. These double meanings create needless barrier to access. Current research also suggests that interface methods for adjusting sites for the physical limitations that seniors experience should be given more attention, for example magnifiers for increasing text size, options for single click over double click on links, and quick and easy navigation back to their home or profile that include the action words like “Click here to return to your profile”. Investigation and progress into these particular areas may make these social environments more inviting to seniors. This research will be used in a larger study to create an interface that third agers can easily use to access Internet content. The goal is to encourage Internet use by creating an interface that makes it easier to navigate. The information in this paper about how and why third agers are currently using the Internet will be used to create a survey measuring seniors current internet use. This survey will be used to identify a group of seniors, age 70 to 75 with little or no internet experience. These seniors will be asked to volunteer their assistance in creating this interface. The interface design and function will be created using some of the research from this paper regarding design and accessibility features; and with information garnered from the volunteers for ease of use and adoptability. This application will run as a proxy server opening the Internet inside a window. The interface will surround the internet window but will still run in a traditional Java-enabled browser; making it accessible on most internet capable devices.

38

C.M. Connolly and G. Meiselwitz

The interface will eliminate double-clicking as a navigation tool by allowing users to associate single click buttons with links existing in the website. These large buttons will have clearly defined functions to simplify the interface and contain action words telling the user to “click here” and what will happen when they do. The interface will also provide “always present” tools for increasing text size and screen contrast that are clearly labeled. Finally, the input from the volunteer group will help in the organization of these features for screen location, size, and labeling. Our volunteers will then be asked to perform a series of internet tasks using Internet Explorer. The users will then perform a closely related but different set of tasks using the new interface. For both task sets the users will be timed and measured for accuracy, with accuracy being a measure of errant clicks and times the user asked for help. The research question being endeavored is whether users with limited internet experience can find the internet easier to use with an interface that provides action based buttons for navigation rather than hyperlinked text.

References 1. The Demographics of an Aging Population, http://www.ageworks.com ROI, http://www.ageworks.com/course_demo/200/module2/module2.htm (accessed on May 12, 2011) 2. Lakin, M., Mullane, L., et al.: Framing New Terrain: Older Adults & Higher Education. Reinvesting in the Third Age: Older Adults and Higher Education. American Council on Education, Washington, DC (2007) 3. Pew Internet and the American Life Project, http://www.pewinternet.org/ (accessed on January 12, 2011) 4. Boyd, D., Ellison, N.: Social Network Sites: Definition, History, and Scholarship. Journal of Computer Mediated Communication 13(1) Article 11 5. Alexander, B.: Social Networking in Higher Education. Selection from: The Tower and the Cloud. Copyright, Educause (2008) 6. Greenhow, C.: Social Networking and Education: Emerging Research within CSCL. CSCL 2009 ROI (2009) 7. Brown, J., Duguid, P.: Organizational Learning and Communities-of-Practice: Toward a Unified View of Working, Learning, and Innovation. Organization Science 2(1), 40–57 (1991) 8. Liccardi, I., Ounnas, A., Pau, R., Massey, E., Kinnunen, P., Lewthwaite, S., Midy, M., Sarkar, C.: The Role of Social Networks in Students’ Learning Experiences. In: ITiCSE – WGR 2009 (2009) 9. Hamasaki, M., Takeda, H.: Find Better Friends? Re-Configuration of Personal Networks by Neighborhood Matchmaker Method. In: SWAFT, pp. 73–76 (2003) 10. Wenger, E.: Communities of Practice: The Organizational Frontier. Harvard Business Review (2000) 11. Cho, H., Gay, G., Davidson, B., Ingraffea, A.: Social Networks, Communication Styles, and Learning Performance in a CSCL Community. Journal of Computers in Education 49(2) (September 2007) 12. Rix, S.: The Employment Picture, October 2008 – Mostly Grim News for Older and Younger Workers. Research Report (November). AARP Public Policy Institute, Washington D.C (2008)

Third Agers and Social Networking in Higher Education

39

13. Kump, S., Krasovec, S.: Education: a Possibility for Empowering Older Adults. International Journal of Life Long Education 26(6), 635–649 (2007) 14. Lankin, M.: Forging New Identities: Older Adults in Higher Education. Internationsl Journal of Continuing Education and Lifelong Learning 2(1) (2009) 15. Cross, D.: Older users Flocking to Facebook, Twitter. CNN ROI, http://www.cnn.com/2010/TECH/social.media/08/27/older.users. social.networks/ (accessed October, 2010) 16. Lehtinen, V., Nasanen, J., Sarvas, R.: A Little Silly and Empty Headed – Older Adults Understandings of Social Networking Sites. British Computer Society (2009) 17. Weilenmann, A.: Learning to Text: an Analytic Study of How Seniors Learn to Enter Text on Mobile Phones. In: CHI 2010. ACM, New York (2010) 18. DiMicco, J., Millen, D., Geyer, W., Dugan, C., Brownholtz, B., Muller, M.: Motivations for Social Networking at Work. In: CSCW 2008. ACM, New York (2008) 19. Seniors Online, Seniors online increase article (2004), http://www.seniorjournal.com/NEWS/SeniorStats/03-02-04 SnrOnline.htm (accessed on January 11, 2010) 20. Holmes, J., Powell-Griner, E.: Lethbridge –Cejku, M. Heyman, K.: Aging Differently: Physical Limitations Among Adults Aged 50 Years and Over. Center For Disease Control, http://www.cdc.gov/nchs/data/databriefs/db20.htm 21. Piper, A., Campbell, R., Hollan, J.: Exploring the Accessibility and Appeal of Surface Computing for Older Adult Health Care Support CHI. ACM, New York (2010) 22. Pfeil, U., Arjan, R., Zaphiris, P.: Age Differences in Online Social Networking – A Study of User Profiles and the Social Capital Divide among Teenagers and Older Users in MySpace. Computers in Human Behavior 25(3), 643–654 (2009) 23. Pfeil, U., Arjan, R., Zaphiris, P.: Age Differences in Social Networking – A study of user Profiles and the Social Capital Divide among Teenagers and Older MySpace Users. Center for HCI Design London, UK (2008)

A Real-time Disaster Situation Mapping System for University Campuses Kayo Iizuka1, Yasuki Iizuka2, and Kyoko Yoshida1 1

School of Network and Information, Senshu University, Kanagawa, Japan 2 School of Science, Tokai University, Kanagawa, Japan {iizuka,k-yoshida}@isc.senshu-u.ac.jp, [email protected]

Abstract. This paper proposes a real-time disaster situation mapping system for campuses. University campuses have various features and issues (e.g., difficult to determine the number of people on campus at any one time as there are many students studying at the library or in their rooms, visitors can also use the open spaces, building layouts are often very complicated etc.), and therefore unique disaster prevention efforts are required. In order to address these prevention issues, a system that facilitates disaster situation information transmission by users will be effective. PlaceEngine is implemented for this system to allow users to estimate the current location easily by utilizing Wi-Fi devices. Keywords: disaster prevention, campus map, situation information gathering, PlaceEngine, social media.

1 Introduction There has been great improvement in various disaster (e.g., capital earthquakes directly above their epicenters) prevention countermeasures and their researches for more than a decade. However, most of the countermeasures that have been implemented are classified as “Kojo” which are implemented by public sectors (in Japanese), therefore, it is considered that the importance of both “Jijo” (countermeasures implemented by the individual) and “Kyojo” (countermeasures implemented by mutual assistance) is increasing [1]. University campuses are one of the prime examples of places that require Kyojo (mutual assistance). In order to assist better evacuation, current situation information of each site is useful, and in most cases, the latest information of each site is grasped by people who were just there. The current situation of each site is necessary not only for evacuation, but also for administering first aid or relief operations. However, not all buildings are closely packed together on campuses, and that may create difficulty in gathering situation information inside the campuses. Under these conditions, a system that facilitates disaster situation information transmission by users (students, staff, faculties, visitors) will be effective. It facilitates supporting people to evade danger, and facilitates support of disaster countermeasures offices to grasp the situation of the many places on campuses. The system proposed in this paper was designed and A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 40–49, 2011. © Springer-Verlag Berlin Heidelberg 2011

A Real-time Disaster Situation Mapping System for University Campuses

41

prototypes were developed for such needs for real-time information sharing during disasters. In order to stimulate mutual assistance, information infrastructures are required. “Recent disasters highlight the importance of social media supporting critical information gathering and dissemination efforts by members of the public” [2]. Recently, social media such as twitter have been effective for disaster evacuation or rescue [3], but these need to be even more effective to gather disaster situation information, categorize them and visually show them on maps. The real-time disaster situation mapping (RDSM) system proposed in this paper can function as one of the social media.

2 Background As mentioned previously, the need for mutual assistance seems to be increasing, because in actual fact many people survive in times of disaster by mutual assistance. In this chapter, disaster prevention works are reviewed and classified from several perspectives (Table 1). Related papers were extracted that matched the specified criteria: those that examined disaster prevention, those that developed an information system, and those whose publication date is post the Great Hanshin Earthquake (Jan.17, 1995). Twenty six papers were extracted using this process [4]-[30]. Public sectors are the overwhelming majority for target users of the system, one paper targets citizens, and two papers target schools or companies (Table 1 (a)). For assistance type, Kojo (assistance by public sectors) is the overwhelming majority (72.1%), and only five papers dealt with Kyojo (mutual assistance) (Table 1 (b)). More than half of the papers dealt with disaster a priori (Table 1 (c)). For the area scope, “broad” (area includes public sector) has the highest percentage (84.6%), and “medium” (schools, shopping center etc.) occupies only 15.4% (Table 1 (d)). As seen above, disaster prevention systems are mainly targeted at public assistance from the public sectors across broad areas. The RDSM system proposed in this paper targets university campuses whose area scope is medium sized, and Kyojo (mutual assistance) because mutual assistance is effective for this scope. Considering the accessibility and role of social media, common devices (mobile phones, PCs) are adapted to the RDSM system.

3 Requirement for Campuses Disaster Prevention University campuses have various unique features and issues when considering disaster prevention. For example, it is difficult to determine how many people there are on campuses. There will be many students studying at libraries, spare rooms, or cafeterias. Visitors can also use the open spaces. Issues with places also exist. The layouts of buildings are often very complicated. They differ depending on when they were built, and construction of buildings differs by research field. Sometimes public roads cut through the campuses. Some campuses are located on inclined ground. Layout of the rooms (e.g., classrooms, experimental laboratories, practical rooms etc.) is not always familiar, therefore, arriving at one's destination is sometimes not easy. In addition, administrative structural issues exist, which are also different from those

42

K. Iizuka, Y. Iizuka, and K. Yoshida

in companies. On-campus disaster countermeasures offices will be set up in times of emergency. However, not all buildings are closely packed together in campuses, and that may cause difficulty in gathering situation information inside the campuses. Under these conditions, a system that facilitates disaster situation information transmission by users (students, staff, faculties, visitors) will be effective. It facilitates support of people to evade danger, and it facilitates the support of disaster countermeasures offices to grasp the situation of many places on the campuses. This system compensates for the area not covered by the public sector (national response system) (Figure 1). Table 1. Classification of Disaster Prevention Related Works (a) Target users Number of papers Public sectors 23 Citizens (individual) 1 Others (*1) 2 Total 26 (b) Assistance type Substances Number of papers Kojo 19 (Assistance by Public Sector) Kyojo (Mutual 5 assistance) Jijo (Self assistance) 2 Total 26 (c) System operating point Substances Number of papers A priori 15 Substances

(e) Point of use Number of papers Outdoor only 19 Outdoor / Indoor 5 Indoor only 2 Total 26 (f) Equipment used Ratio Substances Number of (%) papers 71.2 PC・mobile-phone 15 (via Web browser)

Ratio (%) 88.5 3.8 7.7 100.0

Substances

19.2 Special-purpose 11 terminal 7.7 100.0 Total 26 (g) Adapted technology Ratio Substances Number of (%) papers 57.7 GIS・web-GIS 17

Ratio (%) 73.1 19.2 7.7 100.0 Ratio (%) 57.7

42.3

100.0 Ratio (%) －

(first response)

A priori A priori and a posteriori A posteriori

7 4 0

26.9 Wireless sensor 15.4 GPS 0.0 Others (QR Code, etc.)

6 4 1

－

28

－

－ －

(reconstruction support)

Total Substances Broad

26 (d) Area scope Number of papers 22

100.0

Ratio (*1) Companies, Schools, etc. (%) (*2) Ratio is not applicable because some papers deal with more than one technology. 84.6

(Area includes public sector)

Medium (Schools, Shopping center etc.) Narrow

4

15.4

0

0

26

100.0

(Home, Private concern)

Total

Total

A Real-time Disaster Situation Mapping System for University Campuses Formally included In National Response System

Formal response organizations Affiliated Volunteers Spontaneous Volunteers

43

Incident management system Victims

Community members

Private Sector No defined role In National Response System

Media

Fig. 1. Positioning of Disaster Prevention Systems [31]

4 Real-time Disaster Situation Mapping (RDSM) System 4.1 Overview The Real-time Disaster Situation Mapping (RDSM) system is a web based system that can handle disaster situation information on the campuses by the level of detail, such as classrooms or laboratories in the school buildings. The RDSM system can aggregate the information that is sent by users (informers), store it in a situation database on the server and display a disaster situation map to users on request. Assumed users are people staying on the campus and disaster staff of countermeasures offices (Figure 2, Figure 3). 4.2 Service Conditions of RDSM System The RDSM system is developed based on the following service conditions: • Disaster type: The RDSM system mainly targets the countermeasures for earthquakes. The quake damage assumed that buildings on campus were not fully but may partially destroyed, and mobile devices (PC or phone) are in use and connected to servers inside the campuses or other campuses located remotely. Disaster Information Map

Real-time Disaster Situation Mapping system PlaceEngine General Users -Students -Administration Stuffs -Faculties -Visitors

Updated on request

Disaster Countermeasures Offices

Stored in DB (server)

Situation information by location

Get entry screen

Store

・ situation text voice image

on request Need map

It seems to be safe here .

Situation mapping subsystem General Users （informers） -Students -Administration Staff -Faculties -Visitors

Fig. 2. Overview of System Use

Situation Database ・location

（SGS） Send situation data Use recent information for better decision making

That person who has fallen over needs to be rescued

(RDSM) Situation gathering subsystem

Get location data

mapping on request

show to users

(SMS) information request

Fig. 3. System Flow of RDSM

44

K. Iizuka, Y. Iizuka, and K. Yoshida

• Place of use: People can access the RDSM system from inside the building on campuses, and they can also access it from outside of campuses such as from evacuation centers. Although the RDSM system is developed for campuses, this system can be used for facilities that have similar features to campuses. • Targeted users: People such as students, administration staff, faculties, and visitors are assumed to be informers. People who are assumed to use maps are disaster countermeasures office staff, in addition to people listed as informers above. 4.3 Subsystems of RDSM and Functions The RDSM system consists of two subsystems: the “situation gathering subsystem (SGS)” and the “situation mapping subsystem (SMS)”. Situation gathering subsystem (SGS). The situation gathering subsystem (SGS) facilitates the support of people on the campus to send situation information (e.g., safety or danger, damaged situation) of the location where they are staying at that time, using mobile devices (PC or phone). Informers can access the disaster situation entry screen (Figure 4). Their location information (degree of latitude and longitude) can be specified by PlaceEngine [32]. PlaceEngine is a service that allows users to easily estimate the current location by utilizing Wi-Fi devices. Users can send situation information from the “situation data entry screen” by selecting an answer from check boxes or radio buttons. A free description column is prepared if required. Speech to text transformation is feasible using w3voice [33]. "w3voice skeleton" is a development kit for voice-enabled Web applications. Voice data and image data as well as movie data are also available for sending (Figure 5).

5. Identify location by intensity of Wi-Fi signals

PlaceEngine server

Real-time Disaster Situation Mapping system (RDSM)

Situation Database

Situation gathering subsystem (SGS) 4.Access to PlaceEngine Server

6.Get location data from PlaceEngine 1.Get access to SGS (entry screen)

Situation mapping subsystem (SMS)

3.Receive Wi-Fi signals 2.Activate client software of PlaceEngine

Fig. 4. Entry Screen (SGS)

7． Access to SMS (Situation Map)

Fig. 5. Location Data Acquisition Scheme by PlaceEngine

Location information (degrees of latitude and longitude, or name of the place at which the informer is staying) can be identified by manual entry; in addition, automatic acquisition of location information is feasible by using PlaceEngine. By this function, users (informers) can send location information, even if they do not recognize their location. Entered situation data are stored in databases.

A Real-time Disaster Situation Mapping System for University Campuses

45

Fig. 6. Screens of SMS

In order to set up PlaceEngine for automatic acquisition of location information (degrees of latitude and longitude) the following requirements must be met: • • • • •

Access point data of Wireless LAN are saved in the database of PlaceEngine. At least one access point is reachable from the place where the informer is staying. Access to the PlaceEngine server is feasible. Mobile devices (PC or phone) are equipped with Wireless LAN (Wi-Fi) equipment Client software is installed on the mobile devices (PC or phone) of the informer.

Location information (degrees of latitude and longitude or place name) can be registered by each access point, so identifying places such as classroom or hallways is feasible because there are many access points of wireless LAN on campuses. Situation mapping subsystem (SMS). The situation mapping subsystem (SMS) reflects situation data that are transmitted by users onto maps. It reads data of the situation database that is updated by SGS on request from users. Maps are composed of three levels: overall campus map level (it presents a birds-eye view), building and floor map level, and detailed information of disaster situation by room type (e.g., classrooms, offices, laboratories) level. The situation information amount is displayed at the overall campus level (Figure 6). By clicking a certain building on the overall campus map, a building and floor map appears. Disaster situation information such as fire, rescue required, building damaged, property destruction, and flood damage is shown by icons. The time when those situations were informed is also shown on this

46

K. Iizuka, Y. Iizuka, and K. Yoshida

level map. By clicking a certain room on the buildings and floors map, detailed information of the disaster situation is shown. Detailed information is displayed by time-series on this screen, and voice data or image data are available for access here. The map displayed in the situation mapping subsystem (SMS) can be created from existing map image data that are used for university brochures. The existing map data would be transformed to png format. Situation data that are described or depicted by text data or icons are rendered on this png map. In order to insert disaster-affected points on the map, obtained degrees of latitude and longitude of two different points (place) from Googlemaps are used. By calculating the distance between users’ location and the point obtained from Googlemaps, users’ location is identified from the reduced scale distance that is started from two different points. 4.4 Significance and Features of RDSM System The significance and features of RDSM system are as follows: • Assists users (students, staff, faculties, visitors) in evacuation or rescue: Since the RDSM system is a web based system and uses common mobile devices (PC, phone), users can find out the latest information of a disaster situation (e.g., safety or danger, damaged) according to location easily. PlaceEngine was adapted to assist users in locating and entering their position. Users can choose disaster situation information from options on the screen, instead of having to type everything about the situation they want to convey; however, users can send text data, voice data, and graphic data to the server, if necessary. • Assists the disaster countermeasures office inside/outside the campuses: Although countermeasures office staff inside the university need to gather situation information, it is difficult to keep tabs on the situation overall because of the limited human resources. The RDSM system can cope with this problem by gathering information on the places from the people who are actually there. • Available for use in everyday situations (not just in times of disaster): It is often said that disaster prevention systems do not work in times disaster because the users are not used to operating them. The framework of the RDSM system is effective for everyday situations in the universities, for posting event notifications, maintaining buildings, and so forth (this idea was obtained through feedback from administration staff in the university). In addition to the features mentioned above, disaster situations (fire, rescue required, building damaged, property destruction, and flood damage) are indicated by icons, in order to improve usability. The latest situation data is presented on the map face; however access to time series data is also available for cases where information of the degree of situation change is required. This feature is practical for assisting with rapid rescue, or preventing the damage from spreading. To ensure the communication environment in times of disaster, a backup system at remotely-situated campuses of universities is recommended.

A Real-time Disaster Situation Mapping System for University Campuses

47

5 Conclusion and Future Works As mentioned in previous chapters, the RDSM system facilitates assistance of general users (students, staff, faculties, visitors) and disaster countermeasures office on campuses for effective evacuation and rescue. Thus, the RDSM system facilitates Kyojo (mutual assistance) that is expected to reduce disaster damage, and it can play the role of a social media. The RDSM system is an updated version of a disaster mapping system that was developed by the Yoshida-Iizuka Project 2009 in Senshu University. Comments from attendees of the project exhibition were as follows: The majority of comments were “It makes us feel safe should a disaster occur”, and “It must be available for various sites on campuses” was also heard often. For agita factors, comments such as, “feel anxious about the communication environment in times of disaster”, “worry about mischief and malevolent literature”, “information security must be considered”, were heard. The RDSM system is also premised for use in everyday situations, so this answers the comment “It must be available for various situations on campuses”. RDSM is not supposed to save personally identifiable information. However, if the requirement for dealing with personally identifiable information appears in the future for reasons such as corresponding to the need for urgency, authority setting for accessing data must be carefully considered. Countermeasures for mischief and malevolent literature are not considered now, however, the analysis function of informed data by comparing time series data or data of the surrounding area, is recognized as one of the items to be considered. When considering the “feel anxious about the communication environment in times of disaster” comment, the communication environment is not always completely destroyed by disaster as shown when the communication environment was restored on the day following that of the Great Hanshin-Awaji Earthquake (Magnitude:7.3) in 1995 [6]. By setting up a backup system at remotely-situated campuses of universities, ensuring the communication environment becomes more feasible. Crowd computing would be effective for the universities that do not have remotely-situated campuses. Customizing for each university, upgrading the control panel of RDSM system is planned in order to allow flexibility of controlling system parameters such as disaster emergency level. The input screen of the building layout can be improved to facilitate use. Even though the operating systems that are supported by PlaceEngine are currently limited, they will be widespread in the future. By improving these factors, the role of the RDSM system as social media will be more firmly established. Acknowledgement. Part of this work is supported by a research grant from the Institute of Information Science, Senshu University. We thank Hayato Tsuchiya, Masahiro Urabe, Toshiaki Tsunashima, Tokinori Suzuki, and Masayuki Ishikawa for their contribution to the development of the RDSM system. We would also like to thank members of Yoshida-Iizuka Project 2009 in Senshu University.

References 1. White Paper on Disaster management, Cabinet Office, Government of Japan, http://www.bousai.go.jp/hakusho/h20/index.htm 2. Guido, L., Raquel, B.-F.: The Use of Social Media in Disaster Situations: Framework and Cases. International Journal of Information Systems for Crisis Response and Management (IJISCRAM) 2(1), 11–23 (2010)

48

K. Iizuka, Y. Iizuka, and K. Yoshida

3. Emergency Services embrace Social Networking in Queensland Floods, http://socialmediadaily.com.au/emergency-services-embracesocial-media/ 4. A large-scale disaster and the Internet - How did the Internet correspond to the Great Hanshin Earthquake?, All-in-One INTERNET magazine 2.0, http://i.impressrd.jp/e/2008/01/17/343 5. Nanbu, S., Takahashi, I.: Personal Safety Confirmation System Utilizing Emergency Earthquake Warning and Position Information System by Local Mobile Phone Network. In: Proceedings of the Annual Conference of the Institute of Social Safety Science, vol. (14), pp. 13–14 (2004) 6. Kuwata, Y., Shinjo, A., Ohtani, H., Inoue, U.: A GIS-based Real-time Disaster Information Sharing System. Journal of Information Processing Society of Japan 43(11), 3419–3428 (2002) 7. Terada, M., Sano, Y., Inoue, A., Kaneda, S.: A Web-GIS for Initial Stage of Disasters. Doshisha University Policy & Management Review 8(2), 257–266 (2006) 8. Asakawa, K., Hirano, K., Tsukada, T., Hayashi, Y., In, Y., Omiya, Y., Hamai, T., Murakami, H.: Collection and presentation system of safety information in disaster: An application of GPS mobile phone. ITE Technical Report 33(11), 123–126 (2009) 9. Ichihara, T., Ito, S., Mase, K., Kunifuji, S.: Information gathering system using IR tag in environment with difficult construction of network, IPSJ SIG Technical Reports. Groupware and Network services (30), 37–42 (2005) 10. Hada, Y., Takizawa, O., Shibayama, A., Gyoda, K., Suzuki, T., Kawabata, K., Kaetsu, H., Asama, H.: Research on Ubiquitous Network Technology for Information Acquiring in Disaster. Journal of The Japanese Society for Artificial Intelligence 23(4), 480–485 (2008) 11. Jeong, B., Zama, S., Endo, M., Takizawa, O.: Development of a Prototype Disaster Damage Information Collecting System Based on Cellular Phone. In: Proceedings of the Annual Conference of The Institute of Social Safety Science, vol. (21), pp. 15–16 (2007) 12. Aoki, M., Yonemura, S., Shimokura, K.: Information presentation based on cellular phone in a disaster. Correspondences on Human Interface: Human Interface 9(2), 169–174 (2007) 13. Hirabayashi, Y., Hasegawa, A., Hasegawa, S.: Campus Evacuation Routes Information to be delivered to Mobile Phones. Journal of Nagoya Bunri University 7, 57–64 (2007) 14. Shibayama, A., Takizawa, O., Hosokawa, M., Ichii, T., Hisada, Y., Murakami, M.: A Study on the Information System Using Radio Frequency Identification. Journal of Social Safety Science (8), 135–144 (2006) 15. Ichii, T., Shibayama, A., Murakami, M., Sato, T., Hisada, Y., Namai, C.: Development of the Disaster Information Mutual Support WebGIS as purposes of use in advance or disaster situation. AIJ Journal of Technology and Design (22), 553–558 (2005) 16. Ohno, K., Furuhashi, T., Ishimaru, S., Hata, I.: Basic Research on Earthquake Disaster Preventing Bulletin System using Web Sites: The Outline and Usage of The System, Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, B-2, pp. 799–800 (2008) 17. Kurata, K., Fukuwa, N., Tobita, J.: Development of Web-Based Geographic Information System for Supporting Disaster Mitigation Activities. Journal of Social Safety Science (10), 293–300 (2008) 18. Nishimura, T., Nakada, Y.: Design of an Open Disaster Information System Using GIS. The 57th National Convention of IPSJ (3), 607 (1998) 19. Ishida, E., Fukuwa, N.: Development of GIS for Seismic Disaster Mitigation on Internet Using JAVA. AIJ Journal of Technology and Design (5), 287–291 (1997)

A Real-time Disaster Situation Mapping System for University Campuses

49

20. Tani, S., Fukuhara, M.: Disaster Prevention System by Web GIS in Rural Area. Geoinformatics 17(2), 146–147 (2006) 21. Takabayashi, S.: Information acquisition in disaster crisis management: implementation of image and location data. Bulletin of Saitama Women’s Junior College (20), 1–10 (2009) 22. Narita, S., Shibata, Y.: Disaster Safety Support System Using both Indoor and Outdoor Location Information. The 70th National Convention of IPSJ (4), 875–876 (2008) 23. Nozaki, K., Fukui, Y., Shibata, M., Tanaka, H., Matsuda, Y., Yoshimoto, N., Tsukada, K.: A Proposal for Dynamic Emergency Navigation System in a Time of Disaster. IPSJ SIG Technical Reports DBS (141), 185–190 (2007) 24. Shiwaku, K., Sasaki, M., Kakumoto, S.: Introduction Process of Information System for Community Based Disaster Management. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, F-1, pp. 371–372 (2008) 25. Park, Y., Wang, J., Furuya, T., Sadohara, S.: Development of The Crisis Management System for Improvement on the Emergency Ability Part1: Construction of Emergency Information Database system for the Disaster Support Using GIS. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, D-1, pp. 605–606 (2005) 26. Aoyama, T., Ichii, T., Murakami, M., Hisada, Y.: WebGIS used in normal and disaster situations for sightseeing spot: Application for the ITO in IZU peninsula. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, A-2, pp. 491–492 (2006) 27. Tsuchiya, M.: “TSUJIDO Area” Disaster Prevention Map using GIS (Geographic Information System). Annual Design Review Japanese Society for the Science of Design 11(11), 12–15 (2006) 28. Kurata, K., Murakami, H.: Development of Simple GIS software for Community Disaster Preparedness: A Case Study for Ube City. In: Proceedings of the Annual Conference of The Institute of Social Safety Science, vol. (17), pp. 29–32 (2005) 29. Sato, T., Ichii, T., Murakami, M., Shibayama, A.: Development of the Disaster Information Mutual Support WebGIS: Part 1: Construction of the Date Base on Disaster Information. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, A-2, pp. 501–502 (2005) 30. Imahashi, Y., Kawaguchi, J., Morino, S.: Study on the Earthquake-proof Hazard Map of Mie University Campus by Using M-GIS. Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, B-2 pp. 1027–1028 (2004) 31. http://www.gwu.edu/~icdrm/publications/PDF/Harrald%20HSGAC%2 0testimony_rev.pdf 32. PlaceEngine, http://www.placeengine.com/PlaceEngine 33. w3voice, http://w3voice.jp/skeleton/

Recent Trends in Software Support for Online Communities for Teaching and Research Projects in Higher Education Daniel Kadenbach and Carsten Kleiner University of Applied Sciences and Arts Hannover, Germany {daniel.kadenbach,carsten.kleiner}@fh-hannover.de Abstract. This paper investigates the evolution of software project support for teaching and research projects in higher education. It therefore analyses the results of two surveys conducted with students of the department of computer science of the University of Applied Sciences and Arts Hannover in February 2009 and November 2010. The surveys focused on examining the experiences and requirements of the students. Beneath mere technical requirements the surveys also investigate the acceptance of social software elements in project support. Since containing a core of equal questions the answers to the two surveys allow to deduce trends. These trends towards more sophisticated supportive software tools and environments are further investigated not only to allow a better support for projects, but also to make projects more visible and valuable, preserving their achievements and making them accessible for others. Keywords: software support, projects in higher education, online collaboration, social software.

1 Introduction and Motivation The University of Applied Sciences and Arts Hannover offers bachelor and master’s degree programs in applied computer science. Within these programs a multitude of different projects is conducted each semester and beyond. Similar to many institutions, in addition research projects of longer durations are also executed. These projects differ greatly in their type, contents, lifetime, the number of people involved and their goals. They vary from one person projects such as seminar works, comprehensive homework or final theses to those with many project members like capstone projects or scientific projects led by the faculty or staff. In particular in the scope of software development projects, but also beyond that, a trend becomes more and more apparent to make use of sophisticated supporting software tools, which can provide an enormous benefit if a suitable set of tools is chosen for a project, accepted by the project team and used correctly. This paper analyses how project support by software in the department has changed over time and which possibilities this change offers, but also which challenges are involved with it. By this other institutions shall be given the opportunity to have an insight into the gathered experiences and findings, especially by drafting trends in the usage of supportive software tools for projects, and to encourage an information exchange leading to an improved situation for projects in all higher education institutions. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 50–59, 2011. c Springer-Verlag Berlin Heidelberg 2011 

Recent Trends in Software Support for Online Communities

51

1.1 Example: A “Traditional” Project A software-development project for students without supportive software-systems shall serve as an example for this evolution. Such a project could begin with a few meetings of all project members in which the requirement analysis is conducted, the softwaresystem is designed and further tasks and responsibilities are assigned to the participants. The students henceforth will work on their own on their particular tasks between further team meetings; or they could use frequent meetings to work together in a room on the development of the software. The latter approach, although offering far better communication, will be more expensive in terms of time effort, because the students need time to reach the team’s working-room and because the meeting place usually lacks some comfort, so probably students will try to avoid the additional effort if possible. In addition it is sometimes difficult for the institutes to provide each project team with an individual lab room due to space constraints. Finally the students will put together the modular developed parts of their application and in the best case will test the complete system extensively, and maybe even document it properly if enough time is left. The results of their work will be delivered and if applicable also presented in front of other students, but in most cases they will not be trackable even after a short time. Thus the value of the software developed itself as well as lessons learned during project execution will soon be lost. No sustainable project improvement process is achievable for student’s projects on one hand. On the other hand projects cannot build on previously implemented software in order to achieve higher goals. 1.2 Example: A Project with Software-Support This process can be improved in many ways by supportive software systems for software development projects, which appeared mainly within the development of Open Source Software, such as version control systems, wikis, trackers and others. Because they are used in nearly every real software development project nowadays, the knowledge of these different tools has become an absolute necessity in an up-todate study program in applied computer science to improve student employability. In addition these tools also support the development project itself and thus may improve the results. That implies the knowledge of their operating mode, their advantages and disadvantages in different projects and their handling. In order to make the right choice in the upcoming projects, a student has to know about the qualities and features of a variety of supportive software systems. In using the adequate supporting functions, totally new possibilities could arise in this exemplary software development project. To just name a few: • The use of a distributed version control system like Subversion offers the possibility to work together on the whole source-code of the application from any place. Additionally, one can be informed about the results of each member of the team, avoid conflicts of different versions and make the results of the projects accessible and useful even after the run-time of the project. • A wiki provides a central, up to date, documentation which matches the specific requirements of the project. On one hand the created content can also be accessed and

52

D. Kadenbach and C. Kleiner

preserved beyond the project life-time. On the other hand it can always be changed and updated easily at all times. It is therefore remarkable how simple it is for all team members to manage and find project related information which increases the likelihood that a useful documentation is created. The wiki can provide an external online presentation of the project for other students or even on the Internet as required. It increases the project’s visibility and, if desired, can be a way for the project to receive feedback. • Issue-tracking systems can be used to keep track of bugs, feature requests, or even the planned tasks, forming milestones of packages of this tasks. Team members can be attached to every issue as responsible persons for it. Also the progress of the issues and milestones and therefore of the project itself can easily be tracked by other developers, instructors or even customers. The motivation of the students to reach a high quality in every aspect of the project can be increased through a greater visibility of the project and therefore a much higher possible value and benefit of the project outcomes. This effect increases the more the project generates a value for any type of customer, be it a research project at the school, a company or some non-profit organization. Especially the last aspect allows to use the mechanisms and the potential of social software. In social software system the users are often inherently rewarded by gaining reputation when they create content, which is useful for others (see [12]). Thus the motivation of project participants for providing and preserving quality information and improving existing content can be increased. In section 2 we will be reviewing related publications on software support for student projects. In section 3 the analysis of the conducted surveys follows, starting with a general description of the surveys, describing their methodology and summing up results from which trends are deduced. Also further observations of the daily project support practise will be enclosed at the end of this section. Section 4 sums up the results and deduces conclusions from them. Finally section 5 will draft future work and the focus of future investigations.

2 Related Work Stanfill et al. document their experiences with version control systems and project management tools like Trac [3] for the use in capstone projects for product design teams in [8]. Also the use of wikis such as the MediaWiki is investigated. Problems and solutions with these tools are described. The acceptance of these tools and their benefit is analysed. The importance of learning effective collaboration techniques with these tools is stated, which will also be the case in this paper. Also Radermacher et al. investigate the benefits of using Subversion and Trac to improve capstone projects [9]. Munson investigates the use of wikis to create an organizational memory of projects which is also a subgoal of our efforts [7]. Grudin et al. explore the challenges in using wikis in scientific and engineering organizations, and at least a part of the results also apply in our environment [4]. Denzer describes the importance of teaching students communication skills and tools, which also applies for the supportive software systems described within this paper [1].

Recent Trends in Software Support for Online Communities

53

Another interesting aspect in which the use of collaborative, supportive online tools can greatly improve the results is their use in distributed student projects across mulitple universities, which is investigated in [2] [10]. Also this paper partially continues to investigate and substantiate some directions of our previous one in [6], which focused more on the aspect of social software.

3 Analysis The following considerations are quantitatively based on two conducted surveys, but also qualitative experiences which where gained in two years of daily project support practise are finally shared. In these two surveys (the first one carried out in February 2009, the second one in November 2010) all students of the department were invited to give their answers to an online questionnaire1 . The aim was to gather the experiences of the students with supportive systems for their projects and their requirements for the support they desire for their future projects to meet their needs as good as possible. Of course it had to be considered, that a probably significant part of the students will not be used to many supportive systems and therefore will not be able to estimate their possible value. Nevertheless both surveys allow to take a look at the level of awareness of the students in this matter, and at their demanded requirements for project support. The first insight allows to further optimize the project offerings of the department, the big advantage of the second one is that knowing which systems the students want to use themselves means to know which system will reach the highest acceptance among them. Therefore they will be used with the greatest motivation. But also this is highly dependent on the knowledge of the students about the advantages and disadvantages of different system components, so they are able to make a good choice at all. The surveys however exceeded mere technical requirements. So they also investigated the form of administration the students demanded for their project support. They were asked if some administrators should set up functions for a project on inquiry or if there should be a system where the students could do that autonomously. The acceptance of social software was also analyzed. The students were asked if they are willing to use social functions in their project support – thus generating content for other students through their project work and offering more ways of interaction between projects – and also how deeply they are interested in accessing the contents of other projects. Furthermore because the core of the questions in both surveys was identical, it is possible to conclude developments and trends in the answers by comparing the results. So the summed up results will be discussed, right after a short look at the methodology of the surveys has been taken. 3.1 Methodology Both surveys were carried out online with the open-source tool limesurvey [11]. Therefore the students where invited through mailing-lists to take part in the particular survey. 1

The results can be found at https://proanvil.inform.fh-hannover.de/surveys

54

D. Kadenbach and C. Kleiner

The first survey consisted of 30, the second of 32 questions, each within the categories: Experiences of the students, desired supportive systems, acceptance of social functions, administration of the software support and miscellaneous questions (like for example the usage of virtualisation to provide server instances for projects). Most questions of the survey could be answered with yes, no or abstention. Sometimes the students could choose between a set of predefined answers (for example when asked which version control system they prefer), but were also able to add further answers themselves. Of the about 250 active students in the two programs 60 participated in the first survey and 94 in the second one. This return rate is pretty good as the majority of students is enrolled in the first semesters of the program where no projects are executed. Thus among the students active in projects the return rate is significantly above 50%. 3.2 Results of the Surveys Detailed results for all of the questions cannot be presented due to space constraints. Thus we focus on a summary of the most interesting results here: Basic Acceptance. The surveys showed that 100% of the participants desire supportive software systems for their projects, so that such systems can expect a high acceptance. However it can be assumed that some students who would refuse supportive systems for their projects have not taken part in the surveys in the first place but their number is probably rather small compared to the ones who just did not care about it enough. But this is not really a drawback of the surveys, because supportive systems should always only be an offering for the students, they can choose to use them if they want. They should never be forced to use them in their projects, because this would very likely far decrease the acceptance of the students and render the supportive systems useless. Students should be convinced with arguments to use the provided systems autonomously. Version Control Systems. In both surveys all students voted for the use of version control systems (VCS). In the second survey the questions regarding the usage of VCS were more precise which showed that 96% of the participants even suggested the use of VCS for small projects with only one person like final theses. Furthermore it was investigated which VCS the students preferred, which is shown in table 1. The acceptance of Subversion increased while there is a slight trend towards the more sophisticated Git. Table 1. Survey: Preferred VCS Type First Survey Second Survey Subversion 56% 74% CVS 23% – Git 19% 22%

Recent Trends in Software Support for Online Communities

55

Wikis. The students where also questioned, if they would recommend the use of wikis for the documentation of projects. The acceptance rate in the first survey was at 48%, in the second at 50%. In both cases the other half of the participants abstained from voting. Again the second survey questioned deeper and the students were asked about their experiences with wikis in their previous project work. It could be seen that just the 50% who suggested the use of wikis had already worked with them in their projects, and that the other half which abstained had not gained any experiences with their use. This also illustrates how important it is that students get to know the different supportive systems, so that they can choose the right ones for their future projects. Tracker. The use of issue-tracking systems like the open-source Mantis or Bugzilla, or integrated as parts of other systems like in Trac or Redmine [5], was suggested by 66% of the students in the first survey. The acceptance increased to 85% in the second survey. Mailing-Lists. Automatically generated mailing-lists for projects were favoured by 72% in the first and 80% of the participants in the second survey. Personal Homepages. Personal homepages are the only point in which the demand was decreasing between the two surveys, it sank from 53% to 30%, which shows the lower importance of this more traditional function. Administration. The results of both surveys indicated (71% and 90%), that the students would prefer a system where they can independently choose and establish supportive systems for their projects – like creating Subversion repositories or wiki instances on demand. Usage for Personal Projects. A growing part of the students (from 69% to 93%) would appreciate it if they could use supportive tools for their own small projects, for example using a VCS and a wiki to support them in their final thesis. Social Functions. An Internet portal page, where all projects are listed, and thus which offers the possibility to find, look at and learn from projects of other students and which therefore increases the visibility and sustainability of their own projects alike, was favoured by 71% of the participants of the first and 80% of the second survey. The surveys additionally not only investigated if the students would like to see the projects of others but also if they are willing to share the contents of their own projects with other students. The positive answers increased from 61% up to 82% in the second survey. Similar high acceptance rates where reached by functions such as the following: • The use of instant messengers to support team communication. • The commenting of content of other projects and the receiving of comments for own projects in return. • The creation of user-profiles with interests and abilities to find project-partners and help. • The creation of a central knowledge base with solutions to frequently arising problems. All in all the surveys show a trend towards more experiences of the students and a greater acceptance of and demand for supportive systems as can be seen in figure 1.

56

D. Kadenbach and C. Kleiner

Fig. 1. Trends of the Surveys

Also an increasing willingness and wish to create a sustainable overvalue with their project work for other students and to profit from other students projects likewise was identified. 3.3 Further Observations Other observations were made next to the surveys in the daily project support work. So figure 2 shows the number of projects which where supported with software systems (most of them with VCS, wikis and issue tracker, or integrated systems like Trac or Redmine). A slightly more than proportional increase can be investigated over the years. Figure 3 shows the accumulated frequency of commits in the version control systems of all projects, which fluctuates depending on the current projects and project phases.

Fig. 2. Number of Projects

Recent Trends in Software Support for Online Communities

57

Fig. 3. Overview of Monthly Subversion Commits

It can be seen that project work takes place continuously, even in semester breaks, and that the systems are used extensively. A similar trend can be seen in the wiki usage. In the last year there was a stronger demand for supportive systems which integrate multiple functions like VCS, wiki, tracker, shared file storage and more in one userinterface like Trac and Redmine. Also an increased interest in Git from students and staff members has been observed, which also showed slightly in the surveys.

4 Summary and Conclusions All in all, there is a trend towards more sophisticated software support for projects. This can not only be proved by the increased number of participants in the second survey and their increased level of awareness of available tools and their handling. Also the higher acceptance of these tools by students and staff members in their daily project work demonstrates this trend. The results of the survey point out how important it is to make the students familiar with the possibilities of recent supportive tools. They do not only have to be able to handle these tools efficiently but also, which may be even more important, have to have the knowledge and experience to decide what kind of tool is the best to use in a certain situation. The survey also showed a trend towards tools such as Trac and Redmine for example. Both integrate multiple supportive functions in a single user-interface. The acceptance of Subversion is enhanced within the version control systems. Surprisingly, there was a slight increase in the interest in Git, which offers a far more sophisticated use as Subversion for example with different workflow models. Therefore, its utilization should be absolutely encouraged. Also important and carrying a great potential is the higher acceptance of social functions in project support and the increased demand for a central portal page for projects.

58

D. Kadenbach and C. Kleiner

If the projects can be presented there in an easy way to browse and search, their value will be increased beyond their development time because other students might be inspired. They could learn from conducted projects or even interact with currently carried out projects, giving comments, advice, or criticism. Therefore, through the higher visibility and possible higher benefit of a project for others, the students would be more encouraged to aim for high quality and reusability of their project artifacts, especially the documentation. In using these tools students are supported furthermore in this effort. The overall sustainability of both the project results as well as the process to achieve them is significantly increased. Finally the results of the survey show, that supporting software systems for projects can expect the highest acceptance when the students have the freedom to choose them themselves and are not forced to use a predefined set of tools or to use them at all. This increased responsibility certainly requires that the students are aware of a profound knowledge of the different types of tools available. They need to know about their use, possibilities, strengths and weaknesses. Then they can choose the right tools for their work. This implies that these tools, their attributes and handling early have to be a part of the curriculum in courses and exercises, in particular because the students can profit from their usage already very soon in their studies.

5 Future Work With these results the project support at our department shall be improved to better fit the requirements of the students and staff members. This includes giving them the facility to easily create supportive software systems like wikis, VCS, trackers and others by themselves using a web interface based on a central user-management. This also means building a portal for the projects and establishing functions of social software. All this will cause great efforts to be accomplished. Therefore it is planed to further investigate how other institutes are satisfying this needs and to work together with them striving for solutions which will be generally applicable and an enrichment for all. Additionally future investigation will be made to keep track of the always changing requirements and to gain a finer understanding which types of projects can benefit the most from which type of supportive systems. Finally an eye has to be kept open to look at other collaborative, supportive tools, new versions of used tools and newly developed ones.

Acknowledgments This paper owes various persons a debt of gratitude, particularly the students and staff of the department of computer science at the University of Applied Sciences and Arts Hannover, who participated in the surveys, stated their requirements and wishes, evaluated and used supportive functions and gave valuable feedback. Especially Oliver Bruns and Noomi M¨uller helped to improve this paper. Also the department itself which invests into the challenging research and development of a preferably optimal support for its projects earns gratitude.

Recent Trends in Software Support for Online Communities

59

References 1. Denzer, A.: Long-Distance Multidisciplinary Collaboration: Some Lessons Learned. In: Proceedings of the National Capstone Design Conference (2010), http://capstoneconf. org/resources/2010%20Proceedings/index.html 2. Brereton, O.P., Lees, S., Bedson, R., Boldgreff, C., Drummond, S., Layzell, P., Macaulay, L., Young, R.: Student group working across universities: a case study in software engineering. In: IEEE Transactions on Education, p. 394. IEEE Education Society, Los Alamitos (2000) 3. Edgewall Software: Trac, http://trac.edgewall.org/ (visited, December 2010) 4. Grudin, J., Poole, E.S.: Wikis at work: success factors and challenges for sustainability of enterprise Wikis. In: Proceedings of the 6th International Symposium on Wikis and Open Collaboration, WikiSym 2010, pp. 5:1–5:8. ACM, New York (2010), http://doi.acm. org/10.1145/1832772.1832780 5. Lang, J.-P.: Redmine, http://www.redmine.org/projects/redmine (visited, December 2010) 6. Kadenbach, D., Kleiner, C.: Benefits and Challenges of Using Collaborative Development Environments with Social Software in Higher Computer Science Education. In: Ozok, A.A., Zaphiris, P. (eds.) OCSC 2009. LNCS, vol. 5621, pp. 479–487. Springer, Heidelberg (2009) 7. Munson, S.A.: Motivating and enabling organizational memory with a workgroup wiki. In: Proceedings of the 4th International Symposium on Wikis, WikiSym 2008, pp. 18:1–18:5. ACM, New York (2008), http://doi.acm.org/10.1145/1822258.1822283 8. Keith Stanfill, R., Blackwelder, E.I.: Adapting Lightweight Source Control and Project Management Software for Use by Multidisciplinary Product Design Teams. In: Proceedings of the National Capstone Design Conference (2010), http://capstoneconf.org/ resources/2010%20Proceedings/index.html 9. Radermacher, A., Helsene, A., Knudson, D.: Improving Capstone Courses with Content Management Systems and Virtualization. In: Proceedings of the National Capstone Design Conference (2010), http://capstoneconf.org/resources/ 2010%20Proceedings/index.html 10. Tabrizi, M.H., Collins, C.B., Kalamkar, V.: An international collaboration in software engineering. In: Proceedings of the 40th ACM Technical Symposium on Computer Science Education, SIGCSE 2009, pp. 306–310. ACM Press, New York (2009), http://doi. acm.org/10.1145/1508865.1508976 11. The Lime Survey Project Team: Limesurvey, http://www.limesurvey.org (visited December, 2010) 12. Tørning, K.: Position Paper: CSCW and the Web 2.0. In: Workshop of the Tenth European Conference on Computer-Supported Cooperative Work (2007)

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments: Recommendations for Educators and Wiki Designers Portia Pusey1 and Gabriele Meiselwitz2 1

Department of Educational Technology and Literacy Department of Computer and Information Sciences, Towson University 8000 York Road Towson, MD, 21252 USA {ppusey,gmeiselwitz}@towson.edu

2

Abstract. This paper discusses assessment practice when wikis are used as learning environments in higher education. Wikis are simple online information systems which often serve user communities. In higher education, wikis have been used in a supporting function to traditional courses; however, there is little research on wikis taking on a larger role as learning environments and even less research on assessment practice for these learning environments. This paper reports on the assessment techniques for large- and small scale- learning environments. It explores the barriers to assessment described in the studies. The paper concludes with a proposal of five improvements to the wiki engine which could facilitate assessment when wikis are used as learning environments in higher education. Keywords: Wiki, Wiki Learning, Wiki Learning Environment, Assessment.

1 Introduction Theoretically, wikis have the potential to harness the benefits of Web 2.0 technology to support collaborative learning. While not specifically designed as a learning environment, a previous review of the literature revealed that wikis have many features which enable collaborative learning to occur [1]. However, since wikis are not designed as an educational tool, a repeated weakness reported in the literature is the inadequacy of the wiki to assist assessment. Assessment is a critical part of the wiki learning process. Angelo and Cross [2] write that “the central purpose of Classroom Assessment is to empower both teachers and their students to improve the quality of learning in the classroom” (p. 4). This is because Classroom Assessment “provides faculty with feedback about their effectiveness as teachers, and it gives students a measure of their progress as learners” (p. xiv). The assessment process plays the same significant roles in wiki learning environments as in traditional learning environments – informing and improving A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 60–68, 2011. © Springer-Verlag Berlin Heidelberg 2011

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments

61

instruction and evaluating student growth. However, in wiki learning environments, assessment also serves to ensure students participate. This is because there is evidence that indicates that students will not contribute fully to the wiki learning experience if they are not being graded for their participation [3, 4, 5, 6, 7, 8, 9]. The authors of this work have taught five semesters in a mid-Atlantic college of education using wikis – in classes as small as 12 and combined groups as large as 180 students. They have used best practice for implementing wikis for instruction but continue to find assessment a time consuming process – this limitation of using wikis for instruction is an often reported barrier to implementation [5, 10, 11, 12, 13]. This work represents a synthesis of examples and recommendations from the literature of assessments used for wiki. Assessment has not been the focus of many research studies; however, many reports and studies mention the methods they used to evaluate student progress. After identifying these assessment methods, the barriers to effective assessments in wiki learning environments will be listed. Many of these barriers can be overcome with improvements to the wiki engine. Therefore, this work will conclude with recommendations to improve wiki engines to better support assessment practice.

2 Wiki Assessment in the Literature Wikis are well suited to fulfill the needs of instructors looking for a collaborative, student-centered learning environment. They provide dynamic repositories that allow students to learn by constructing knowledge based on their experiences in a selfdirecting manner; becoming active contributors to each other’s learning; and developing a sense of community ownership of this content [14, 15, 16]. However, wikis were not designed as learning environments for formal education. Therefore, instructors and wiki researchers have used traditional method of assessment to evaluate student work. Assessment methods have rarely been the focus of a research study so this list represents the types of assessment referred to in the studies on wikis published in peer review journals over the past 5 years. Many of the studies found in the body of literature on instructional wikis are qualitative; many of the journal articles included a reference to the type of assessment used. In the descriptive studies, traditional assessments using quizzes and rubrics were used; peer assessment also was frequently used by instructors. Another method that was evaluated quantitatively as well as qualitatively was the use of the wiki usage data. The qualitative studies explained methods instructors used to attribute an individual’s contribution to group work. One of the qualitative studies attempted to automate the assessment process using the portfolio record kept in the wiki [11]. While there was no consensus in the literature about the best method to use, there was wide agreement that assessment in wiki learning environment remains a time consuming process. 2.1 Quizzes Quizzes have also been used to assess knowledge gained from the creation and use of the wiki. Formats suggested by Hazipanagos & Warburton [17] include multiple choice, short answer, true/false, and matching. In many cases students created the quiz questions and answers [17, 18, 19]. In these studies, and in research by the

62

P. Pusey and G. Meiselwitz

authors of this work, students write the questions and answers. They submit the questions and answers to the instructors who create an automated quiz using a quiz tool or course management system such as Blackboard. Quizzes can be taken without the aid of the wiki; or, when students can use the information in the wiki, we have called the “quiz” a scavenger hunt. Quizzes and scavenger hunts are tools that can be used formatively to improve or drive instruction. They are used summatively when the goal is to improve knowledge or awareness of the wiki content. The benefit to using a quiz, especially an online quiz, is that the feedback to the student can be immediate. However, wikis can be used for more than generating knowledge and awareness and at least one wiki researcher called for a change in practice away from summative, norm referenced tests to meet the demands of the authentic project-based learning that occurs online [20]. 2.2 Rubrics Several of the studies used rubrics to assess their students [4, 18, 21]. The use of rubrics to assess authentic activities reflects the evidence in the body of literature that rubrics are effective tools to assess complex authentic student work [22, 23,24 25, 26, 27].The use of rubrics to assess student work was recommended by one research because rubrics allow the instructor to set clear expectations up front [28]. Best practice on the use of rubrics underscore the rubrics’ ability to inform student work as it is being created when the rubric is given to students prior to instruction [29]. When rubrics are given to students when learning activities are assigned, students are able to review their own work for errors prior to submission for feedback or grading [30]. The authors have used rubrics to create an all-in-one assessment that not only assessed the content and quality of the student’s contribution but also the quantity and quality of their collaborative posts. Rubrics are used both summatively and formatively to assess multiple facets of a student project. 2.3 Peer Assessment Since wiki is a social learning experience a social evaluation tool has been often used. Several studies reported using peer assessment [11, 18]. One study, which used several semesters and repeated measures, reported peer assessments to be equivalent to instructor assessments [19]. Irons [31] recommend using peer feedback as a way to improve student learning and reduce the time commitment on the part of the instructor. In one account, students completed the wiki learning activity over several class periods [18]. This study reported that students anonymously evaluated each team member’s, as well as their own, contribution to the work at the end of each day. The author reported that this self and peer rating system valued each students’ contribution. Peer assessment also can contribute to interaction between students. In a study that reported on 5 semesters of teaching with wikis in a college of education [11] this study suggested that peer assessment encouraged interaction between the author and the other students. In this study, feedback was provided to the authors by the other students in the class using the threaded discussion area of the page. By participating in the conversations occurring in the threaded feedback the instructor has an opportunity to monitor and assess the learning as well as the

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments

63

collaborative experience [12]. The authors also use the threaded discussion area as a formative feedback forum for students. In addition to the improved interaction/ collaboration and improved student learning (evidenced in other studies) the authors also found the threaded feedback discussion a wealth of information about student learning and understandings about the content. Studies suggest that students may not be able to provide quality feedback without being trained to do so [22, 23, 24, 25, 26, 27]. So, a tutorial or practice is beneficial to be sure the peer assessment will reflect the instructor’s expectations. Formative peer assessment when conducted in the threaded discussion, is more time consuming than if the instructor were to complete this evaluation herself. This is because the instructor not only needs to read and provide feedback to the author; the instructor also is assessing the student feedback to the author. However, the threaded discussions also allow the instructor the ability to participate in the discussions occurring among students and remediate any misconceptions and inaccuracies as they are occurring. Summative peer feedback, by well trained students, has been reported as an effective means to evaluate students while reducing instructor workload [28]. 2.4 Wiki Usage Data A final method of assessment, which is most frequently reported in the literature, is to use a student’s quantity of work as part of his final grade. In most cases, studies report that students are not only graded on the quality of the content they are producing but on the quantity of their contribution to the end product. Wiki usage data has been used in several studies to provide the instructors with a portfolio of information that tracks posts to individual users [5, 11, 12, 28, 32, 33]. The literature on assessment for wikis used as learning environments suggests that while students should work collaboratively, they should be assessed individually [7]. Several studies reported higher student satisfaction and a perception of fairness in the assessment of group work because of the capability of individualized assessment in wiki learning [34, 35, 36, 37, 38, 39, 40, 41]. However, there are concerns that have surfaced in the literature that the usage data may not accurately reflect the group process [4, 42]. First, quantities of posts are not an indication of learning; and second, the quality of the post is a more important consideration as evidence of learning. Infrequent posts do not necessarily indicate that a student is not learning [3, 42]. Student teams may choose to collaborate face-to-face and designate one student to post their work. And qualitative studies have indicated that students read and re-read wiki content without contributing to the conversation [43]. Paradoxically, one study hypothesized that if more in class face-to-face time had been allocated to the wiki activity, perhaps greater evidence of collaboration would have been seen in the wiki learning environment [44]. Therefore, the recommendations found in the literature are that the usage data should only be used as a small part of the assessment process.

3 Recommendations for Practice The authors of this work contend that the barriers to effective assessment stem from the fact that wikis were never designed to be used as a learning environment. Therefore, several changes to the wiki engine need to be considered before effective and efficient assessment can be designed and tested.

64

P. Pusey and G. Meiselwitz

3.1 Facilitate Communication within the Wiki Wikis offer several modes of communication – the threaded discussion associated with each content page, email, and direct edits to the content page. However, the only way to provide private feedback to students is to use email. The authors of this article have found that the email tool within some wikis often require several clicks to communicate directly with students. Furthermore, email communications within some wiki engines leaves no record of the private conversations. Therefore, creating a communication system within the wiki that allows for private conversations to occur would be helpful in building a virtual community [17]. An effective private way of communicating with students would also resolve some of the frustrations suggested by another study which reported student frustration and confusion because multiple learning platforms are used [5]. The students in this study used course management software for most tasks and used the wiki to communicate with their tutors. In order to encourage students to use the wiki platform, the authors of this study suggested that the instructor should refuse to answer questions by email [46]. However, improvements to the wiki email to link directly into the email systems that the instructors and students currently use would be a better solution. 3.2 Comment Tool Learners benefit when they are provided frequent and detailed feedback [10, 40]. In one study English as a Second Language students required a large amount of immediate feedback [5]. Zorko [5] reported that the ability to react quickly when there was a problem and to provide guidance was motivating for his students. Furthermore, Zorko reports that students can respond to the comments more quickly and produce a better end product. Currently, all comments must be made to students using the threaded discussion, direct edits to the page, or in email. A comment tool, similar to the one used in popular word processing revision tools, would facilitate the feedback process. First, it would enable the instructor to provide targeted feedback directly linked to sentences, paragraphs, or words. Therefore, when the instructor is providing feedback they no longer need to identify where, in the text, the feedback is targeted. Second, it would enable peers to provide feedback without directly editing the page. Students don’t like other students directly editing their pages [3, 4]. Formatting issues have occurred, and changes have been undone [43]. Furthermore, students have reported that they are reluctant to edit each other’s work [9]. Therefore, a targeted comment --using a callout box or stickynote type tool – would allow students to suggest changes to the page without actually making them. 3.3 Colored Instructor Comments The threaded discussion area is still a valuable source of evaluation data. By participating in the conversations occurring in the threaded feedback the instructor has an opportunity to monitor and assess the learning as well as the collaborative experience [12]. Several studies reported on the use of peer assessment in threaded discussions to help learners improve the content of their wiki page [17, 19]. Using the threaded discussion section is very helpful for students to provide feedback and to enable them to defend their suggestions [12]. However, one study reported the

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments

65

concern that in the large body of information and threaded discussions it was feared that instructor comments would get lost and students would not see the feedback [32]. If the instructor comments appeared in a unique color they would be easily identifiable by the students and would not get lost. 3.4 Connections to Plagiarism Checking Software It is easy for students to cut and paste content directly from their web browser into their wiki content pages. Several studies reminded students not to directly cut and paste into the wiki [3, 3]. In fact, in one study, students expressed concern about material copied directly from books and Internet sites [43]. With so much of wiki content being text-based, a tool that checked pages for plagiarism within the wiki would be ideal. Sites such as Turnitin are used by educational institutions to check their students work for originality (http://turnitin.com/static/index.php). As a first step, giving instructors the ability to test content pages with a single click of the button would help make the assessment process more efficient. However, providing students access to this tool would also serve as a metacognitive tool to improve their awareness of plagiarism and the need for proper citation. 3.5 Integrated Rating Tool It has been suggested that when students have been properly trained to use the tools, peer assessment can be equivalent to instructor assessments and can improve end products [19, 22, 45, 46, 47]. Furthermore, peer feedback, especially formative feedback, can reduce the time commitment on the part of the instructor [31]. However, wikis only currently facilitate qualitative feedback. The qualitative feedback is an integral part in helping students improve their contributions to the wiki, however most grading systems are quantitative (A-F, 0-4, 25 points, etc.). Therefore, the addition of a flexible rating tool would aid in the assessment process. The proposed tool would work similarly to the feedback systems in many online shopping and auction sites which use a star system to accumulate buyers’ feedback into information about the reliability of the seller or product. The system would link directly to a rubric and accommodate multiple ratings based on the number of subtasks being assessed [29]. The rating scale should be flexible to accommodate local assessment scales – i.e. ranges of letters and numbers. Finally, to facilitate best practice for peer assessment the system should allow for anonymous ratings [27, 47, 48]. Of course, a reporting tool for the instructor, which would export the rating into a spreadsheet, would also be necessary.

4 Conclusion This work is part of a larger study that evaluates the use of wikis to change attitudes in preservice teachers. It began in 2009 with a review of best practice for implementation of wiki as a learning environment [1]. After five semester of implementing wiki learning environments it was clear that the best practice found in the literature could help an instructor successfully facilitate collaborative work. However, the struggle to effectively and efficiently assess collaborative groups up to 180 students led these authors to the literature for answers.

66

P. Pusey and G. Meiselwitz

Attempts have been made to automate the assessment process in a wiki but these attempts were not successful, therefore, the assessment process is still a laborintensive process on the part of the instructor [4, 41]. The examples from the literature quizzes, rubrics and peer assessments are useful for effective formative and summative assessment but the processes are not efficient ones. Therefore, we suggest improvements to the wiki engine to better facilitate instructor and peer assessment. We suggest improved communication tools -- private conversations between instructors and students, colored instructor comments, and a commenting tool -- a flexible rating tool that directly connects to the assessment rubric and exports reports to spreadsheets; and plagiarism checking within the wiki. Whether these improvements to the wiki engine are ever implemented, future work should focus on empirical work that evaluates assessment practice so that instructors have complete body of work on best practice for implementing wikis as learning environments.

References 1. Pusey, P., Meiselwitz, G.: Development of Heuristics for Implementation of Wiki technology as Learning Environments in Higher Education. In: Stephanidis, C., et al. (eds.) 2009 Human Computer Interaction Conference Proceedings, pp. 507–517. ACM, New York (2009) 2. Angelo, T.A., Cross, K.P.: Classroom Assessment Techniques: A Handbook for College Teachers, 2nd edn. Jossey-Bass, San Fransisco (1993) 3. Wheeler, S., Yoemans, P., Wheeler, D.: The Good, the Bad, and the Wiki: Evaluating Student-Generated Content for Collaborative Learning. Brit. J. of Ed. Tech. 39, 987–9956 (2008) 4. Lin, H., Kelsey, K.D.: Building a Networked Environment in Wikis: The Evolving Phases of Collaborative Learning in a Wikibook Project. J. of Ed. Computing Research 40, 145– 169 (2009) 5. Zorko, V.: Factors Affecting the Way Students Collaborate in a wiki or English Language Learning. Australasian J. of Ed. Tech. 25, 645–665 (2009) 6. Choy, S.O., Ng, K.C.: Implementing Wiki Software for Supplementing Online Learning. Australasian J. of Ed. Tech. 23, 209–226 (2007) 7. Cubric, M.: Wiki-Based Framework for Blended Learning. In: Proceedings International Symposium on Wikis, pp. 11–24. ACM, Montréal (2007) 8. Ebner, M., Kickmeier-Rust, M., Holzinger, A.: Utilizing Wiki-Systems in Higher Education Classes: A Chance for Universal Access? Universal Access in the Information Society, vol. 7, pp. 199–207 (2008) 9. Tal-Elhasid, E., Meishar-Tal, H.: Naples Models for activities, collaboration and assessment in wiki in academic courses. In: Electronic Proceedings of the Eden 2007 Annual Conference (2007), http://www.biu.ac.il/bar-e-learn/eden2007/tal_tal.doc 10. Beebe, R., Vonderwell, S., Boboc, M.: Emerging Patterns in Transferring Assessment Practices from f2f to Online Environments. Electronic J. of e-Learning 8, 1–12 (2009) 11. Trentin, G.: Using a Wiki to Evaluate Individual Contribution to a Collaborative Learning Project. J. of Comp. Asst. Learning 25, 43–55 (2009) 12. Campbell, K., Ellingson, D.E.: Cooperative Learning at a Distance: An Experiment with Wikis. American J. of Bus. Ed. 3, 83–89 (2010)

Assessments in Large- and Small-Scale Wiki Collaborative Learning Environments

67

13. AL-Smaki, M., Guetl, C., Kappe, F.: Peer Assessment System for Modern Learning Settings: Towards a Flexible E-Assessment System. Int. J. of Emerg. Tech. in Learning 5, 5–11 (2010) 14. Reinhold, S.: Wiki trails: Augmenting Wiki Structure for collaborative interdisciplinary learning. Paper presented at WikiSym 2006 (2006) 15. Schaffert, S.: IkeWiki: A Semantic Wiki for Collaborative Knowledge Management. In: Enabling Technologies: Infrastructure for Collaborative Enterprises. IEEE, Los Alamitos (2006) 16. Wang, C., Turner, D.: Extending the Wiki Paradigm for Use in the Classroom, pp. 255– 259. IEEE Press, New York (2004) 17. Hatzipanagos, S., Warburton, S.: Feedback as Dialog: Exploring the Links between Formative Assessment and Social Software in Distance Learning. Learning, Media & Tech. 34, 45–59 (2009) 18. Tharp, T.L.: Wiki, Wiki, Wiki – What? Assessing Online Collaborative Writing. English J. 99, 40–46 (2010) 19. Ren, Z., Baker, P., Zhang, S.: Effects of Student-Written Wiki-Based Textbooks on PreService Teachers’ Epistemological Beliefs. Ed. Comp. Res. 40, 429–449 (2009) 20. Williams, P.: Assessing context-based learning: Not Only Rigorous but also Relevant. Assmt. & Eval. In Higher Ed. 33, 395–408 (2008) 21. Lending, D.: Teaching Tip Using a Wiki to Collaborate on a Study Guide. J. of Information Systems Ed. 21, 5–13 (2010) 22. Bloxham, S., West, A.: Understanding the Rules of the Game: Marking Peer Assessment as a Medium for Developing Students’ Conceptions of Assessment. Assmt. & Eval. in Higher Ed. 29, 7333–7721 (2004) 23. Prins, F.J., et al.: Formative Peer Assessment in a CSCL Environment: A Case Study. Assmt. & Eval. in Higher Ed. 30, 417–444 (2008) 24. Sluijsman, D., Prins, P.: A Conceptual Framework for Integrating Peer Assessment in Teacher Education. Studies in Ed. Eval. 32, 6–22 (2006) 25. Sluijsmans, D.M.A., et al.: Training Teachers in Peer-Assessment Skills: Effects on Performance and Perceptions. Studies in Ed. Eval 41, 59–78 (2004) 26. Sluijsmans, D.M.A., et al.: The Training of Peer Assessment Skills to Promote the Development of Reflection Skills in Teacher Education. Studies in Ed. Eval 29, 23–42 (2003) 27. Wen, M.L., Tsai, C.C., Chang, C.Y.: Attitudes Towards Peer Assessment: A Comparison of the Perspectives of Pre-Service and In-Service Teachers. Innov. in Ed. & Teaching International 43, 83–92 (2006) 28. Minocha, S., Thomas, P.G.: Collaborative Learning in a Wiki Environment: Experiences from a Software Engineering Course. New Review of Hypermedia & Multimedi 13, 187– 209 (2007) 29. Castle, S., Arends, R.: Developing credible performance assessments. In: Castle, S., Shaklee, B.D. (eds.) Assessing Teacher Performance. Rowman & Littlefield Education, Lanham (2006) 30. Andrade, H., Du, Y.: Student perspectives on rubric-referenced assessment.. 3. Practical Assessment Research & Evaluation 10, 159–181 (2005) 31. Irons, A.: Enhancing learning through formative assessment and feedback. Routledge, New York (2008) 32. Elgort, I., Smith, A.G., Toland, J.: Is Wiki an Effective Platform for Group Course Work? Australasian J. of Ed. Tech. 24, 195–210 (2008)

68

P. Pusey and G. Meiselwitz

33. Slovie, P.A.: Use of the Wiki: Encouraging Preservice Teachers’ Construction of Knowledge in Reading Methods Courses. J. of Literacy & Tech. 9, 0975–1535 (2008) 34. Bruns, A., Humphreys, S.: Building Collaborative Capacities in Learners: The M/Cyclopedia Project, revisited. In: Proceedings International Symposium on Wikis, pp. 1–10 (2007), http://www.wikisym.org/ 35. DePedro, P.: New Method Using Wikis and Forums to Evaluate Individual Contributions in Cooperative Work While Promoting Experiential Learning: Results from Preliminary Experience. In: Proceedings International Symposium on Wikis, pp. 87–92. ACM, Montréal (2007) 36. Clark, D.J., Redmond, M.V.: Small group instruction diagnosis: Final report. Seattle, WA: University of Washington Seattle. ERIC Document Reproduction Service No. ED217954 (1982) 37. Hassaien, A.: A Qualitative Student Evaluation of Group Learning in Higher Education. Higher Ed. in Europe 31, 138–150 (2007) 38. McGraw, P., Tidwell, A.: Teaching Group Process Skills to MBA Students. In: A Short Workshop. Ed. & Training, vol. 43, pp. 162–170 (2001) 39. Elliott, N., Higgins, A.: Self and Peer Assessment – Does it Make a Difference to Student Group Work? Nurse Ed. in Practice 5, 45–48 (2004) 40. Gupta, L.M.: Enhancing Student Performance Through Cooperative Learning in Physical Sciences. Assmnt. in Higher Ed. 29, 63–73 (2004) 41. Armellini, A., Aiyebayo, O.: Learning Design and Assessment with E-tivities. British J. of Ed. Tech. 41, 922–935 (2010) 42. Wang, H.C., et al.: An empirical exploration of using Wiki in an English as a second language Course. In: Proceedings of the 5th IEEE International Conference on Advanced Learning Technologies, pp. 155–157. IEEE, Los Alamitos (2005) 43. Matthew, K.I., Felvegi, E., Callaway, R.A.: Wiki as a Collaborative Learning Tool in a Language Arts Method Class. J. of Res. On Tech. in Ed. 42, 51–72 (2009) 44. Judd, T., Kennedy, G., Cropper, S.: Using Wikis for Collaborative Learning: Assessing Collaboration Though Contribution. Australasian J. of Ed. Tech. 26, 341–354 (2010) 45. Hubert-Williams, N.: Facilitating Collaborative Learning using Online Wikis: Evaluation of their Application within Postgraduate Psychology Teaching. Psych. Learn. & Teach. 9, 45–51 (2010) 46. Schneider, D.K., et al.: Conception and Implementation of Rich Pedagogical Scenarios Through Collaborative Portal Sites: Clear Focus and Fuzzy Edges. In: International Conference on Open and Online Learning (2003), http://tecfa.unige.ch/proj/seed/catalog/docs/icool03schneider.pdf 47. Tsai, C.C., Liu, E.Z.F., Lin, S.S.J., Yuan, S.M.: A Networked Peer Assessment System Based on a Vee Heuristic. Innov. in Ed. & Teaching Int. 38, 220–230 (2001) 48. Freeman, M.: Peer Assessment by Groups of Group Work. British J. of Ed. Tech. 20, 289–299 (1995)

Teacher Agents: The Current State, Future Trends, and Many Roles of Intelligent Agents in Education Kevin Reed1 and Gabriele Meiselwitz2 1 Department of Information Technology, Northern Virginia Community College, Alexandria Campus, Bisdorf 352, 3001 N. Beauregard St., Alexandria, VA 22311 USA 2 Computer & Information Sciences Department, Towson University, 8000 York Rd. Towson, MD 21252 USA [email protected], [email protected]

Abstract. Since their development in the 1980's, Intelligent Tutoring Systems (ITS) have experienced a widespread success in such varying areas of education as military training, personal tutoring, and vocational instruction[9], [12], [19], [36]. ITSs are not without limitations, however, and have often proven to be costly and inflexible. Combining these systems with Intelligent Agents (IA), first proposed in the 1990s, is intended to address some of the shortcomings of ITSs; notably the cost of building new learning objects. While IA provide a mechanism for generating dynamic content tailored to a specific learner, a lack of standardization in IA ontologies and a narrow focus on pedagogy provides rich veins for reseach. In this paper we broadly survey the development of IAs in education with an eye towards further exploration of their possibilities. Keywords: Intelligent agents, education, chatbot, tutoring systems.

1 Introduction Speaking at the 2010 commencement of Hampton University, U.S. President Barack Obama remarked First and foremost, your education can fortify you against the uncertainties of a 21st century economy. In the 19th century, folks could get by with a few basic skills, whether they learned them in a school like Hampton or picked them up along the way. For much of the 20th century, a high school diploma was a ticket to a solid middle class life. That is no longer the case. Jobs today often require at least a bachelor's degree...[2] Mr. Obama's observations iluminate a movement that has been emerging since the century mark. U.S. Census data shows that college enrollments nationwide are up 17 percent since 2000, while college costs continue to climb rapidly [42] and government support for higher education continues to decline. Of particular concern in these diverging trends is the fact that many more of the students enrolling in college, and particularly community colleges, since 2000 require intervention services and remediation in the core subjects of mathematics, reading, and writing [31]. Rising A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 69–78, 2011. © Springer-Verlag Berlin Heidelberg 2011

70

K. Reed and G. Meiselwitz

enrollments, too few instructors, stagnant budgets, and students poorly prepared for the rigors of a college education produce a toxicity that must be diluted if the U.S. is to retain its competitiveness in the global economy of this century. The traditional method of addressing student deficiences and individual needs has been the personal tutor, generally an individual who is both a subject domain expert, and takes the time to get to know his or her students personally, so that instruction can be matched to the student's particular learning style and goals. With increasing success, this role has been shifting (or at least supplementing) from human to human interaction to human to machine sessions; starting in the early 1960's [9]. Initially, these systems, typically known as Computer Aided Instruction (CAI), automated routine testing and drilling, with little differentiation between their operation and earlier wholly mechanical systems -- except they ran on computers. But advances in the fields of artificial intelligence and computer hardware, along with evolving perspectives in education and cognitive psychology, enabled ambitious researchers in the 1970s to model and develop 'intelligent' computer based instruction, commonly referred to as 'Intelligent Tutoring Systems' (ITS) [9].

2 Intelligent Tutoring Systems: Design, Characteristics, and Types The term 'Intelligent Tutoring System' is expansive in scope, entailing any computer application that possesses intelligence and is used in instruction. Generally, an ITS should be comprised of one or more autonomous software agents (discussed in greater detail below) that interact with students, present content, and provide assessment, without the immediate oversight of human instructors [9] [11]. ITSs have been constructed from a variety of languages and tools, and run on different platforms; and vary in the degree of intelligence they exhibit [11]. All of them, however, share specific architectural characteristics: the domain or expert model, the student model, the tutor model, and the interface or environment model. [9], [17]. Each of these is a component of a larger system. 2.1 Design Domain Model. The domain or expert model contains the knowledge, behaviors, and formalisms that make up the content being taught, both declarative (facts) and in some measure, procedural (processes for manipulating facts). It is a repository of the rules and inferences a subject-matter expert would follow would to solve a particular problem, such as doing addition in mathematics or diagnosing a breakdown of an automobile air conditioning system. Called an 'expert system' in artificial intelligence (AI), many researchers also refer to this component as the 'cognitive model', as there is an implicit relationship between the content and the 'knowledge state' of the expert working through the problem. The knowledge state, in turn, are those points transitioned to along the path in which the expert possesses increasing understanding of the task at hand. Student Model. The student model tracks and saves the active knowledge states of the student user in constructs such as Bayesian networks, so that progress can be

The Current State, Future Trends, and Many Roles of Intelligent Agents in Education

71

mapped. This component provides the student with a map of his or her learning, and dovetails into the tutor model. Student models have three tasks ([45] as cited in [41]): 1.

2.

3.

By explicit (asking questions) or implicit (tracking the student's navigation) means, they must collelate data about the student and compare it to other student responses. This data must be used to map the student's knowledge and learning. Taking the form of 'buggy' models (representation's of the student's knowledge as deriviations from established 'expertise'), the model will attempt to predict future student responses. These are compared to the actual responses, and a more precise map of the student's knowledge states can then be made. The model must further diagnose the current state of the student's knowledge, so that optimal learning strategies can be chosen to present subsequent domain information for mastery.

Tutor Model. Wenger stated that when 'learning is viewed as successive transitions between knowledge states, the purpose of teaching is accordingly to facilitate the student's traversal of the space of knowledge states' ([45] as cited in [41]). The tutor model, which Wenger called 'pedagogical expertise', monitors the progress captured in the student model, and intervenes in situations where there are discrepancies between the rules established in the domain model. Aptly named, it is the role of the tutor model to manage adaptive teaching strategies, such as learning path maps and recommendations; sending the student remedial instruction and positive reinforcement through the interface model as warranted. These goals are addressed through two modes of support states ([45] as cited in [41]): 1. 2.

Diagnosis, in which the ITS extrapolates state information from the student's observable behavior, knowledge level, and learning style Didactics, in which an appropriate curriculum is dynamically created and adjusted to match the student's current level.

Interface Model. Not surprisingly, the interface model is the means through which an ITS user interacts with the system – both human student and human instructor. The complexity or 'richness' of the interface is dictated by the detail needed for the student to view and complete a problem, and for human monitor to administer the system, making refinements to the domain model as needed. Wenger suggested that the Interface must consist of two component ([45] as cited in [41]): 1. 2.

The discourse model, which detects abiguities in student answers and provides a corrective responses. Knowledge presentation, which provides guidance to keep the student from missing key elements of the knowledge being conveyed.

2.2 Characteristics Woolf et al., 2001 [46] describe six characteristics or abilities of ITSs. They must: • generate 'appropriate instructional material' − This quality distinguishes intelligent tutors from computer aided instruction, which uses a finite set of problems and responses.

72

K. Reed and G. Meiselwitz

• model student performance − Assessments need to be formative and immediate, so that a catalog of student strengths and weaknesses can be created and used to update instructional material and provide feedback. ƒ Formative assessments are used to determine if students are learning the material currently being presented. They are intended as a 'feedback loop' for the instructor. This type of assessment can be contrasted with summative assessements, which are generally used to measure a student's overall performance with the objective of assigning a grade or mark. [16] • model expert performance − This is the location of the tutor's knowledge base, which gives the tutor domain intelligence. • model pedagogical strategies − Provides the tutor with the adaptibility to accommodate students at various levels of learning. ƒ To accomplish, the tutor must be 'aware' of a student's current knowledge state and learning style; and possess the ability to define learning strategies dynamically. • model natural dialog − A tutor must at least give the appearance of natural communication. (This is being realized now through developments in natural language processing). • self-monitor performance − A tutor should be capable of reflective activity that will improve its 'teaching' over time. Interestingly enough, Woolf et.al., stated that every ITS should possess at least one of these qualities; however it seems more likely that every truly intelligent tutor should possess all of them. Domain knowledge and student tracking are features of CAIs, necessary but insufficient for genuine teaching.1 Reflecting on the ongoing processes, developing strategies for aiding students; in effect 'getting to know' the learner are essential virtue for any teacher. There is a clear and self-validating relationship between Woolf et.al.'s list of ITS features and the architecture of these tutors that has evolved over the last twenty years, as shown in figure 1. 1

Meeriam and Caffarella identify four distinctive theories of education ([22] as cited in [35]): 1. Behaviorist: Focused on skill development and training, views the learning process as tantamount to behaviorial change 2. Cognitive: emphasizes 'learning how to learn', sees learning as an internal mental structuring 3. Humanistic: defines the learning process as an act of self-fulfillment and stresses selfdirected learning 4. Social/situational: proposes that learning is the relationship that evolves between student and environment, and that the role of teaching is to create or identify communities of practice. In addition, an often referenced theory of learning in ITS literature is constructivism. Based on the work of Vygotsky and others, constructivist-based learning articulates a model similar to that of the social/situational theory: learning is a process of 'enculturation' into a knowledge community [35].

The Current State, Future Trends, and Many Roles of Intelligent Agents in Education

73

Fig. 1. Relationship between Woolf et.al.'s list of ITS features

Intelligent Agents. In a typical ITS, these abilities are manifested by a collection of autonomous but tightly coupled intelligent agents. A working definition of 'intelligent agent' can be elusive, because they are the subject of extensive research and reporting. Generally, though, they can be described as autonomous enities that observe and impinge upon their environment, which is a combination of application, data storage, knowledge source, and network. They are rational, goal directed agents [32], and usually software constructs [3]. 2.3 ITS Types ITSs can further be classified by the cognitive models they employ: Cognitive or Constraint Based [37]. Constraint Based, often referred to as the Behavioral approach, has its roots in the work of B.F. Skinner and other behavioral psychologists and identifies learning in terms of 'cause and effect' chains [8] or constraints [25]. An example of such a constraint would be: if is true, then must be true, otherwise something is wrong. This learning model is focused on an immediate, declarative problem and has no contextual awareness beyond the chain. Consequently, it is easier to implement as a rules based tutor, but is limited in its ability to provide feedback and reinforcement to the learner [37]. The Cognitive model is based on the theories of John Dewey, a philosopher of education, and the research of educational pyschologists Lev Vygotskey, Jean Piaget, and others [8]. It is holistic in approach and views learning as both declarative (facts) and procedural knowledge construction in a social context. Within the framework of an ITS, rules built on this model could take the following path [37]: if the goal is to classify a shape and the shape has three sides then classify the shape as a triangle because triangles have three sides Facts are easy to code, procedure is more challenging. However, the cognitive model has the clear advantage in its inherent feedback loop.

3 An Additional Classification We observe that there is a further classification of ITSs, and specifically those that constructed on the cognitive model. This additional subdivide is predicated on the

74

K. Reed and G. Meiselwitz

source and substrate of their knowledge domain. While hardly trivial, there are domains in which declarative knowledge is well established and the procedures for deriving and manipulating facts are well known. Disciplines such as mathematics and the sciences are in this area. Their rules and knowledge construction can be described by human experts and crafted into ontologies2 that can be used to built a knowledge base for an ITS. It's no coincidence that the most successful and widely known tutors in existence today provide training in mathematics and the sciences. Projects discussed in the literature include the PUMP Algbra Tutor Project, a partnership between Carnegie Mellon University and the Pittsburgh School System; SHERLOCK, a U.S. Air Force initiative to train jet plane mechanics; AutoTutor, a Physics instruction system at the University of Memphis; and CIRCSIM-Tutor, a natural language ITS project at the Illinois Institute of Technology, which teaches 1st year Medical students about circulatory system pressure. These systems are 'successful' because they encompass clearly defined rules and facts [9] [11]. In other domain areas in which content and knowledge is more unstructured, constructing an ontology is more difficult [28]. This is both the opportunity and challenge of the World Wide Web. An almost unlimited amount of data can be found there, but its minimal organization makes it hard to locate and use information efficiently. This is the issue that the 'Semantic Web'3 is intended to redress [13], though the degree of uncertainty associated with web content makes this an arduous undertaking [18]. Still, the web is an unequalled source of material, and can be mined as a knowledge base for an ITS. We propose, then, to prototype an ITS that helps teach adult learners to read by using content from the web as source material. Reading is a complex activity that typically takes years to master, but is a foundational skill in education and in the workplace. Without this skill, young adults can never hope to obtain the educational levels that President Obama stated are 2

Ontologies are formal representations of knowledge within a domain. They define a vocabulary and semantics to describe the concepts within a domain and the interrelationships between those concepts [13]. More concretely for this discussion, an ontology is a text-based construct of reference/knowledge that is built with an ontology representation language by a domain expert; and can be consulted by intelligent agents as part of their knowledge base [10]. They are conceptually similar in purpose to XML. A simple example, decribed in the OIL (Ontology Inference Layer) ontology representation language, would be [10]: class-def defined herbivore subclass-of animal, NOT carnivore slot-contraint eats value-type plant OR (slot-constraint is-part-of has-value plant) A number of ontology representation languages have been created, with perhaps the most 'successful' being OWL (Web Ontology Language), which became a W3C recommendation in February 2004 [20]. Typically, domain experts will use ontology authoring tools such as Protege-2000 (http://protege.stanford.edu/) to create and modify ontologies, as their syntatic complexity makes hand coding error prone and tedious. Developing ontologies is time consuming and very expensive [23], and so a significant amount of research is ongoing to find ways to automate the process [37]. 3 The term 'Semantic Web' was coined by Tim Berners-Lee and describes the technologies and mechanics that will allow machines, and specifically intelligent agents, to understand the meaning, the 'semantics', of information on the web [4]. This is an area of much active research.

The Current State, Future Trends, and Many Roles of Intelligent Agents in Education

75

essential in his commencement speech. Nevertheless, this remains an ongoing problem in post secondary education. A 2004 U.S. Department of Education study reported that 42% of 1st year students in public two-year institutions needed remediation [38]. At Northern Virginia Community College, the second largest school of its kind in the nation, and situation in one of the country's best educationed regions; over 60% of traditionally aged college students require remediation in reading [24]. There are numerous reasons why so many students come out of high school reading poorly. Social promotions and lack of adequate diagnosis in K12 impact student performance [27]; as do school district retention efforts (school systems 'dumb down' their programs to reduce student dropout). Additionally, a reluctance to inflict a sense of 'low self-esteem' on students [5]; and the simple fact that more students are enrolling in colleges today than in previous generations [43] all contribute to a situation that now needs to be addressed on an ongoing basis. The use of a machine based tutor can help in aleviating this educational and social problem, by providing reading level appropriate material and assessments for students to practice. There are currently a number of machine based tutoring systems with ITS characteristics, but none appear to use the web as source of reading material. The use of 'canned' material can be problematic, as the content may be of no interest to the student and subsequently discourage additional practice. [29], [30].

4 Towards an Intelligent Reading Tutor We envision a Cognitive model ITS in which remedial students select topics of personal interest, such as Sports news, or that have some utility, like want ads. A repository in the ITS would then be checked for matching material and, if found, presented to the student. Otherwise, the system would dispatch an agent to retrieve candidate documents from the web and place them into a database. This process would be ongoing. In the database the documents would be checked for reading level difficulty and semantically annotated with the ITS's 'priming' ontology – a basic set of patterns and responses that will provide a substrate for subsequent assessments. Because we are not attempting to map complex domain knowledge about the document contents, we believe that machine annotation will be possible. The material will then be annotated with Artificial Intelligence Markup Language (AIML), an XML compliant language that is used to create life-like dialog for an A.L.I.C.E chatbot [1]. We expect that this will enable the ITS to use a chatbot4 for assessment, which will be informal and conversational. A possible use scenario would be: 1. Student selects reading material, which is fetched from database 2. The reading material is preannotated with agent generated AIML statements that ask the most fundamental of questions, such as 'Is this article about X' and 'When do the events talked about happen?' 4

Chatbots, chatterbots, and conversational agents are 'talking' computers with antecendents in the research of Alan Turing. They are software constructs that, through various mechanisms used to acquire the appearance of natural language understanding, provide an interactive conversation with human users. The first actual implementation was 'ELIZA', developed at MIT in 1966 by Joseph Weizenbaum; with subsequent agents created to interact in commerce, education, and entertainment, including Second Life. [44]

76

K. Reed and G. Meiselwitz

3. Using a simple web interface, the student will read the selected material and answer the chatbot's questions. We will explore the feasibility of using a Baysian network to analyze the probability that the student's responses are correct. 4. The student's 'correct' responses will be used to formulate additional AIML statements. Answers marked as 'incorrect' will be flagged for review by a human expert. The student will be aware that he or she is 'training' the chatbot to understand the article being read, and is thereby contributing to a community of readers, in support of a constructivist or social/situational learning process. Obviously, the chatbot cannot truly ascertain whether or not the student actually comprehends the material read, but we will look for emerging patterns in the student responses indicative of student learning – one of which will be the 'success' of subsequent readers in responding to the chatbot's questions. Scoring will then be dispatched to the agent that will maintain student profiling (and a human instructor for any arbitration), so that future reading selections are more finely tuned to the student's reading level and interest. We will also study the chatbot dialog to determine whether this type of unstructured communication might be useful in creating more complex ontologies. The ITS we are proposing is not intended to replace human to human interaction and instruction. Learning is a highly social activity, especially in one as fundamental as reading comprehension. We wish to examine the interesting ramifications such an ITS might have on human-computer teaching and learning support, but ultimately, we hope to provide a tool that adults can use to improve their reading.

References [1] Foundation, A.I. (n.d.). AIML – The Artificial Intelligence Markup Language, http://Alicebot.org , http://www.alicebot.org/aiml.html (retrieved December 1, 2010) [2] B., O.: Educated Citizens in a Changing World. Vital Speeches of the Day 78(8), 364– 366 (2010) [3] Bellifemine, F., Caire, G., Greenwood, D.: Developing multi-agent systems with JADE. John Wiley & Sons Ltd., West Sussex (2007) [4] Berners-Lee, T., Hendler, J., Lassila, O.: The Semantic Web. Scientific American Magazine (2001) [5] Brandon, C.: Nearly 30% of college freshmen can’t read, write, or calculate (2008), http://www.thefiveyearparty.com:http://www.thefiveyearparty. com (retrieved December 23, 2010) [6] Capuano, N., Carrolaggi, P., Crestani, J., Gaeta, M., Herber, E., Sangineto, E., et al.: A Virtual Organization for e-Learning. In: 1st International Kaleidoscope Learning GRID on Distributed e-Learning Workshop [7] Cartelli, A.: Semantics, Ontologies and Information Systems in Education: Concerns and Proposals. Issues in Informing Science and Information Technology 3 (2006) [8] Conway, J.: Educational Technology’s Effect on Models of Instruction. Educational Technology (1997), http://udel.edu/~jconway/EDST666.htm (retrieved November 26, 2010) [9] Corbett, A.T., Koedinger, K.R., Anderson, J.R.: Intelligent Tutoring Systems. In: Helander, M., Landauer, K., Prabhu, P. (eds.) Handbook of Human-Computer Interaction, 2nd edn., pp. 849–874. Elsevier Science B.V, Amsterdam (1997) [10] Devedizic, V.: Education and the Semantic Web. International Journal of Artificial Intelligence in Education 14, 39–65 (2004)

The Current State, Future Trends, and Many Roles of Intelligent Agents in Education

77

[11] Freedman, R.: What is an Intelligent Tutoring System? In: Ali, S.S., McRoy, S. (eds.) Intelligence, vol. 11 (3), pp. 15–16 (2000) [12] Gasevic, D., Hatala, M.: Ontology mappings to improve learning resource search. British Journal of Educational Technology 37(3), 375–389 (2006) [13] Hendler, J.: Agents and the Semantic Web. IEEE Intelligent Systems, 30-37 (2001) [14] Hermann, J.F.: The Silicon Tutor: A critical examination of the uses of intelligent tutoring system in writing assessment and pedagogy. Doctoral Dissertation, Indiana University of Pennsylvania (2005) [15] Huang, W., Webster, D., Wood, D., Ishaya, T.: An intelligent semantic e-learning framework using context-aware Semantic Web technologies. British Journal of Educational Technology 37(3), 351–373 (2006) [16] Interactive Educational Systems Design, Inc. (n.d.). How Jamestown Reading Navigator Supports Research-Based Instruction for Struggling Adolescent Readers Formative and Summative Assessment, http://www.readingnavigator.com: http://www. readingnavigator.com/mkt/assets/formative_and_summative_asse ssment.pdf (retrieved December 21, 2010) [17] Jia, J.: An AI Framework to Teach English as a Foreign Language: CSIEC. Association for the Advancement of Artificial Intelligence, pp. 59–71 (2009) [18] Laskey, K. J., Laskey, K. B., Costa, P. C., Kokar, M. M., Martin, T., Lukasiewicz, T. (eds.). Uncertainty Reasoning for the World Wide Web (2008), http://www.w3.org:http://www.w3.org/2005/Incubator/urw3/ XGR-urw3-20080331/ (retrieved November 19, 2010) [19] Maddux, C.D., Johnson, D.L.: Information Technology in Education: Some Reasons for Optimism. Computers in the Schools 26, 83–88 (2009) [20] McGuinness, D. L., & van Harmelen, F. (n.d.). OWL Web Ontology Language Overview. OWL Web Ontology Language Overview, http://www.w3.org/TR/owl-features/ (retrieved November 19, 2010) [21] Memon, Q., Khoja, S.: Academic Program Administration via Semantic Web. In: Proceedings of World Academy of Science, Engineering, and Technology, vol. 37, pp. 695–698 (2009) [22] Merriam, S., Caffarella, R.: Learning in Adulthood: A Comprehensive Guide. JosseyBass Incl., San Francisco (1998) [23] Murray, T.: Authoring intelligent tutoring systems: An analysis of the state of the art. International Journal of Artificial Intelligence in Education, 98–129 (1999) [24] Office of Institutional Research, Student placement rates across different developmental areas: 2004-2007. Northern Virginia Community College, Office of Institutional Research (2010) [25] Ohlsson, S.: Constraint-Based Student Modeling. In: Greer, J., McCalla, G. (eds.) Student Modeling: The Key to Individualized Knowledge Based Instruction, pp. 167–189 (1994) [26] Pahl, C., Holohan, E.: Applications of Semantic Web Technology to Support Learning Content Development. In: Koohang, A. (ed.) Interdisciplinary Journal of E-Learning and Learning Objects, vol. 5 (2009) [27] Pennington, M.: Why Johnny still can’t read (2008), http://www.edarticle.com:http://www.edarticle.com/k-12subject-areas/reading/why-johnny-still-cant-read.html (retrieved December 23, 2010) [28] Rajput, Q.N., Haider, S., Touheed, N.: Information Extraction from Unstructured and Ungrammatical Data Sources for Semantic Annotation. In: Proceedings of World Academy of Science, Engineering and Technology, vol. 40, pp. 295–303 (2009) [29] Ramachandran, S., Atkinson, R.: An Evaluation of Intelligent Reading Tutors. In: Woolf, B.P., Aïmeur, E., Nkambou, R., Lajoie, S. (eds.) ITS 2008. LNCS, vol. 5091, pp. 731– 733. Springer, Heidelberg (2008)

78

K. Reed and G. Meiselwitz

[30] Ramachandran, S., Stottler, R.H.: An Intelligent Tutoring System for Adult Literacy Enhancement. In: Proceedings of the fifth Internation Conference on Intelligent Tutoring Systems. Montreal (2000) [31] Reed, C.B., Conklin, K.D.: Enrolling in College, Ready or Not. The Chronicle of Higher Education 52(8) (October 2005) [32] Russell, S., Norvig, P.: Artificial Intelligence: A Modern Approach, 2nd edn. Prentice Hall, Upper Saddle River (2003) [33] Shabajee, P., McBride, B., Steer, D., Reynolds, D.: A prototype Semantic Seb-based digital content exchange for schools in Singapore. British Journal of Educational Technology 37(3), 461–477 (2006) [34] Shafrir, U., Etkind, M.: e-Learning for depth in the Semantic Web. British Journal of Educational Technology 37(3), 425–444 (2006) [35] Smith, M.K.: Learning theory, the encyclopedia of informal education (2009), http://www.infed.org:http://www.infed.org/biblio/b-learn.htm (retrieved December 22, 2010) [36] Snae, C., Brueckner, M.: Ontology-Driven E-Learning System Based on Roles and Activities for Thai Learning Environment. Interdisciplinary Journal of Knowledge and Learning Objects 3 (2007) [37] Stamper, J.C.: Automatic Generation of Intelligent Tutoring Capabilities via Educational Data Mining. Charlotte, NC (2010) [38] Stotsky, S., Wurman, Z.: College Students Who Can’t Do Math Or Read Well. Minding The Campus Reforming Our Universities (2009), http://www.mindingthecampus.com/originals/2009/12/by_sandra_ stotsky_and_zeev.html (retrieved December 22, 2010) [39] Tankeleviciene, L., Damasevicius, R.: Characteristics of Domain Ontologies for Web Based Learning and their Application for Quality Evaluation. Informatics in Education 8(1), 131–152 (2009) [40] Tomsett, C.: Reconfigurability: creating new courses from existing learning objects will always be difficult. Journal of Computer Assisted Learning 21, 440–448 (2005) [41] Urban-Lurain, M.: Intelligent Tutoring Systems: An Historic Review in the Context of the Development of Artificial Intelligence and Educational Psychology (1996), http://www.cse.msu.edu/rgroups/cse101/ITS/its.htm (retrieved December 21, 2010) [42] Census Bureau, U.S. (n.d.). College Enrollment Up 17 Percent Since (September 17, 2000). M2 Presswire [43] USA Today. One-third of students need remedial college math, reading (2010), http://USAToday.com:http://www.usatoday.com/news/education/ 2010-05-11-remedial-college_N.htm (retrieved December 23, 2010) [44] Wallace, R.: Alicebot, Alice Blog (2009), http://alicebot.blogspot.com/ (retrieved December 23, 2010) [45] Wenger, E.: Artificial Intelligence and Tutoring Systems: Computational and Cognitive Aproaches to the Communication of Knowledge. Morgan Kaufmann, Los Altos (1987) [46] Woolf, B.P., Beck, J., Eliot, C., Stern, M.: Growth and maturity of intelligent tutoring systems: A status report. In: Forbus, K., Feltovich, P.J. (eds.) Smart machines in education: The coming revolution in educational technology, pp. 99–143. AAAI Press, Menlo Park (2001) [47] Woolf, B., Eliot, C.: Customizing the Instructional Grid. Applied Artificial Intelligence 19, 825–844 (2005)

Interpreting User-Generated Content: What Makes a Blog Believeable? Rahayu Ahmad and Wayne G. Lutters Department of Information Systems, University of Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, MD 21250, USA {arahayu1,lutters}@umbc.edu

Abstract. Blogs have democratized participation for everyone who wants to publish on the Internet. This trend, however, is problematic for readers trying to determine trustworthy and believeable sources. This blog diary study, allowed us to understand what aspects influenced believability and ultimately utililty through capturing routine blog-surfing behaviors. We found that literary appeal constituted by authentic and personalized stories positively influence the believability of blogs. Blogs that stimulated readers’ desire for novel and rare information also enhance their believability. Other factors that significantly correlated with believability included aesthetics, matching information needs and user expectations.

1 Introduction Web logs (blogs) have become an essential cord in the social fabric of the Net. A blog is usually defined as a frequently updated webpage and generally have customized navigation forms with dated entries organized as an archive in reverse chronological order [18]. Blogs can come in many different forms: filters, personal journals, and notebooks [13]. Filter blogs are mainly comprised of selective links by the authors, which often become pointers to other information resources. Personal journal blogs are more diary-like, containing personal thoughts and records of daily events. Lastly, notebook style blogs are usually in the form of long and focused essays. While blogs’ form and function have shifted over time, these solo-authored, widely read, personal information sources remain the bedrock of our user-generated infoculture [16]. Blogs have democratized participation for everyone who wants to publish on the Internet. Their popularity is largely dpendent on commercial blog software, such as Blogger [4]. Through such user-friendly software, any individual can feasibly create and publish contents on a potentially international scale with little technical knowledge. Given the personal voice, opinionated writing style, and opportunity to author anonymously, how do readers evaluate the believability of individual blogs?

2 Background Most research on the concepts of credibility or believability in social media has been conducted on web pages, a more general genre of online authorship than blogs. It has A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 81–89, 2011. © Springer-Verlag Berlin Heidelberg 2011

82

R. Ahmad and W.G. Lutters

focused primarily on trustworthiness and expertise. Trust is described as the willingness of the author to make valid assertion while expertise is the ability (competency) of the author in making those assertions [12]. In perceptions of website credibility, researchers have emphasized three factors: navigation, usability, and author background/affiliation. Their studies have demonstrated that websites with a professional look and feel and ease of use positively influence the perception of credibility [12]. The site author’s background and identifiable personal information also has positive impact [10]. Scholars have argued that mechanisms for establihsing a perception of credibility among readers may be meaningfully different from general websites and traditional media [15][19]. It has been speculated that trust with readers are facilitated by developing a personal relationship through disclosing the intimate personal and professional lives of bloggers, allowing readers to see the world through their eyes [22]. These authentic experiences, shared in the form of narratives, are a quintessential trait of blogs [2][20]. Compared with credibility in web sites, author identifiable information has no significant influence towards blog believability or credibility perception. One study demonstrated that there were no differences in perceived credibility when the blogger was identifiable versus when they were anonymous [7]. This implied that the identity of the writer was not being used to evaluate the credibility of the blog content. Another study suggested that bloggers build their credibility by showcasing their intellectual abilities through their analytical arguments and insightful commentaries of issues or events [2]. In addition, among experienced Internet users, blogs are perceived as more credible compared to traditional media such as newspapers or TV. [14]. Futhermore, unlike other media, bias in blogs was considered acceptable to readers [14]. These current studies suggest that blog readers may use distinctive factors in evaluating the believability of blogs, which have not yet been explored in greater detail.

3 Study Design Informed by the limitations of prior studies, we designed a contextual diary study that best preserved natural user behavior. An email invitation was sent out to undergraduate and graduate students and their friends who regularly maintain or read blogs. Twenty people participated in this study. Of the participants, 20% were male and 80% were female. The range of participant ages was 24 to 45 years, with a mean age of 28.5 years and a mean of 16 years of education. Our participants’ demographics were representative of the population of bloggers as indicated by Pew Internet and American Life Project [16]; in which half of all bloggers are under the age of 30. Based on answers to the self report question about their blog reading habits 20% reported doing so on daily basis, 50% were weekly, and 30% monthly. Participants were randomly presented with a search scenario to identify blogs that they would want to recommend to a friend within three broad interest areas: health, parenting, and travel. The rationale of having diverse types of blogs was to see whether there are any significant differences of factors influencing believability. The three areas were selected to represent various degrees of consequences and opinionated contents.

Interpreting User-Generated Content: What Makes a Blog Believeable?

83

For example health blogs would likely have higher consequences of believability and a lower degree of opinionated contents compared to parental and travelling. Based on the scenarios, participants were asked to search for blog at home on their own time, bookmarking each one that they found to be useful, with a minimum request of five. For each discovered blog, they were asked to complete an online diary consisting of a brief online survey with both open and 5-point Likert-scale questions. The open ended questions asked participant to briefly summarize the blog to ensure that they were actually reading it. The particants were also asked to describe their reasons for selecting the blog. As for the 5-point Likert scale questions (1=strongly disagree and 5=strongly agree) participants rated the blog on dimensions developed from Flanagin and Metzger [11] like believability, bias, persuasiveness and interest. The participants also rated each blog using 10-point scale (1=very low and 10= very high ) in property terms proposed by Rubin and Liddy [19]: design, accuracy, timeliness of information, transparency of author information, literary appeal, profit purpose, appeal, and believability. Participants were given a maximum duration of two weeks for completing these online diaries. After analyzing the participants’ compilation of selected blogs and their evaluation for each one, we invited 5 people for a follow-up interview session to better understand their rationale for selecting those blogs and to provide further explanatory detail for our data analysis. These participants were selected based on the extremes of their property ratings, focused on those with large variance in their factors that might influence believability. The interview session was contextualized in the sense that both participants and the interviewer revisited the blogs of interest during the interview session. This interview was conducted through the Skype application which permits screen sharing between the two parties.

4 Analysis In total 77 blogs were analyzed. We performed correlational analyses to identify factors that correlated with believability. We also coded the participants’ responses to the open ended question "why do you think the blog is useful?” to help interpret the statistical results. The relationship between all variables and the blog’s believability was investigated using the Spearman correlation coefficient since the measures were ordinal in nature. There was a moderate correlation between the two variables literary appeal and believability (rs = .37, p < .05; n=77). The variable curiosity trigger also had moderate correlation with believability (rs = .32, p < .05; n=77). The other three variables that were weakly correlated with believability were aesthetic, (rs = .26, p < .05), matching needs (rs = .22, p < .05) and matching expectation (rs = .25, p < .05). The following section provides a discussion of our results.

5 Result and Discussion 5.1 Literary Appeal and Believability Similar to the web credibility literature [10], we observed that participants valued reliable citations and references in judging the believability of the blogs:

84

R. Ahmad and W.G. Lutters

I think this blog is very useful for parents because it provides a scientifically proven result. (Tracy)1 There are a lot of links and information about diabetes. The information is a concrete, supported by links, videos, academic research and other reliable resources. (Melissa) However, the importance of these factors was most dominant in the medical scenario given the serious consequenses of applying fraudulent knowledge. For other scenarios where citations are absence, participants relied on self disclosure of the bloggers and authenticity of their personalized stories in making a believability judgement. Self-disclosure has been shown to increase trust in workgroups [5][24]. For bloggers, the self disclosure tended to be holistic covering not only aspects of her/his work, but her/his self and personal life [22]. Through stories, photos, and personal links a blog’s personality can reveal much about the author’s values, orientation, and biases. In our study, participants valued the personalized diaries which enforced the believability of the blog. One participant explained: It gives a daily account of a person living with diabetes, which gives a balanced and realistic story to believe in. (Kate) The participants also emphasized the authenticity of the blogs’ content as one of the criteria they used in recommending it to others. Participants perceived real people’s experiences as authentic and trustworthy, as one noted: It [the blog] provides a perspective on Bali done by a foreigner who is actually living there instead of simply visiting as a tourist. (Sam) The authentic and genuine experience allowed readers to relate to the author and immerse themselves in the situation, thus enabling greater believability of the content: “When I want to go blogs, I like them to be like talking to themselves, don’t act like they are at work, or a teacher or university professor. I just want to have their pure feeling because it helps me to relate and be more attune to the situation.” (Cheryl) This finding potentially explains why the author background has not been found to be critical for determining credibility. The personalized voice and first-hand experience of the bloggers are cues for determining their competency and hence the believability of their blogs. This observation is also supported by a recent finding that authors’ authenticity and passions are a reliable measure for determining blog’s content credibility [15].

1

Our stylistic convention is to identify quotations from the interviews in italics and direct excerpts from the diaries in Courier font.

Interpreting User-Generated Content: What Makes a Blog Believeable?

85

5.2 Curiosity Trigger and Believability In the studies of online trust, curiosity has been perceived as critical component in a two staged trust model [17]. Based on this model, trust is formed gradually, initially started with exploratory stage and continue to commitment stage [17]. In exploratory stage, the user did not engaged yet in a specific web site and trying to decide whether or not to explore the web site. Egger [9] argued that users initial reaction to a web site is influence by the its appeal and its ability to arouse the users’ curiosity for further exploration. In this study, blogs that have meaningful headings/title and offers variety of information aroused their desire to explore the blog in details as described by our participants: “I could see that thediabetesblog.com would be more informational. Thus I would explore this blog more.” (Melissa) “It combines food and travel, so i'm curious about the topic, on how food varies around the world.” (Adam) Most of our participants were motivated by the desire to seek out novel information, also known as diversive curiosity [3]. Thus, authors who satisfied the readers’ desire for novel and intriguing information, enhanced the believability of their own blogs.One participant noted: “I was curious to know the different between the types of diabetes stages,and from reading her blogs and links provided from this blog, I get to know them.” (Tania) As described by other researchers [1][14], blogs are perceived to be a credible medium due to the perception of their ability to provide insightful, independent information. Our study found that participants were seeking blogs that provided exquisite and interesting information that was hard to find elsewhere, which eventually lead to higher believability ratings. The following diary notes explain: This blog shows the more "hip" side of Bali life. As opposed to other blogs about Bali/Indonesia which tend to be on the denser, more intellectual side of the blogosphere spectrum, this blog provides a lighthearted view of the nightlife, style and beach culture of Bali. (Sam) If someone wants to travel around the world and eager to know the hidden treasures of the destinations, this [the blog] is the right place to find. (Adam) In a prior study, the novelty of a blog’s content significantly influenced the playfulness perception of the travel’s blog which eventually led to a positive intention to visit the places[6]. This implied that the novelty of the blog content is used as a reliable indicator to actually make decisons. Our study provides further explanation of this novelty impact for the believability perception of blogs.

86

R. Ahmad and W.G. Lutters

5.3 Aesthetics and Believability Aesthetics, described as the overall look and design of a site, has been established as one of the factors that influence the credibility of websites. In Fogg’s [12] credibility study involving 2500 participants, the general look and feel was the top most factor. Professional looking web sites were often rated as more credible. For this study, generally particpants described blogs with high aesthetic qualities as: “A good blog should have a neat design, not too colorful, and has a appealing color scheme. It also has to have contemporary design” (Adam) Look and design impacts the judgement of the competency of the blogger, and, hence, influence the beleivability of the contents, as one participant highlights: “I think that an author who would spare some time on finding a good layout is actually dedicative about his/her blog.” (Melissa) There was also a strong reliance on images for verifying the blogs’ contents and the blogger’s experience particularly in travel blogs as illustrated by the following excerpt: This blog is based on his real experience which inclusive tips and photo as a proof. (Lisa) Our findings suggest that similar to web credibility perception, the look and design of blogs is being actively considered in evaluating believability. However, it has a somewhat weaker correlation with believability when compared to literary appeal or writing style. This finding might be attributed to the nature of blogs which are diary like, with a higher degree of personal disclosure, resulting in greater attention to its writing style and contents. The look and design, however, supplemented other cues in determining the competencies of the bloggers. 5.4 Matching Information Needs, Prior Expectation and Believability Increasing numbers of people are relying on blogs as primary sources of information. It is reasonable to argue that the ability of blogs to provide sufficient and appropriate information to the readers enhances their believability perceptions. Our diary responses demonstrated that users acquired the information they were looking for which eventually influence their believability perceptions: The links provided on the site are suitable for our needs of information. (Lyn) It [the blog] helps me to plan out when should I go and expect upon vacation to Langkawi by next month. (Lisa) It has been demostrated elsewhere that readers are often attracted to blogs for their depth of information [23]. The comprehensiveness of information may signal the expertise of the blogger to either compose or compile useful information, hence

Interpreting User-Generated Content: What Makes a Blog Believeable?

87

positively influencing the believability of the blog. The excerpts below demonstrated our participants’ views: It consists of everything that a diabetic needs to know. (Abby) The current posts and seem to mostly be relevant to child development and the responses seem to be appropriate for the topic of the original post. (Gilmore) In the public relations literature, an important property of trust is the trustee’s expectation about the motives and behaviors of a trustee [8]. If those expectations are met, trust is seen as having been warranted. If not, either the level of trust will be lowered or distrust may guide future actions. Similarly, in the blog literature, the audience of a blog has expectations for the authenticity of the blog [20]. In Baumer et al.’s [2] study, readers recognized the personal opinion and narratives in a blog as part of the blog’s authenticity. We shared a similar observation in our study where participants expected the blogs to be based on firsthand experience as precisely described by Tania: “My expectations are personal experience from the author and also other useful links related to the topic” (Tania) On top of that, readers also had some expectation for the frequency of updates, visual style, navigability, responsiveness, appropriateness, and other aspects [2]. From the previous discussions we have observed that aesthethics (visual style) and matching information needs (appropriateness) positively correlate with believability. Having most of the readers’ expectation met, led to perception of believability for a blog.

6 Conclusion This study has a few key limitations. Firstly, it employs a correlational analysis method which prevents us from making cause and effect conclusions. Future work can establish a reliable model of believability perceptions by incorporating the aspects identified in this study. Secondly, although we tried to cover a wide range of blogs topics, our three categories are just the beginning of a comprehensive study. Blogs that are dedicated to crisis management (in event of natural disaster), for example, may give greater importance to the “timeliness” factor for asessing the believability. The value of this particular study is twofold. Firstly, previous studies have mainly focused on credibility perceptions of blogs with other traditional media. These studies have highlighted that blogs are perceived as more credible particularly among experience Internet users. Previous studies also demonstrated that traditional measures like author identifiability did not influence the credibility perception in blogs. These suggested that other factors may be more important in determining the believability and usefulness of the blogs. Our multi-method study is a step toward addressing these gaps. In our field analysis, some key factors like literary appeal and aesthetics emerged as central traits, confirming past literature. However, literary appeal has taken a different meaning where the emphasis is on the personalized and

88

R. Ahmad and W.G. Lutters

authentic experience of the authors which positively influence the believability and utility of their blogs. We also found an emerging factor, curiosity trigger, described as the ability to stimulate readers desire for novel information, to influence the believability of blogs. Readers were attracted to blogs that provided exquisite and unique information which could not be found anywhere else. This finding supports other studies who found that people considered blogs as source of rich and deep alternative information source, which eventually led to positive credible perceptions [1]. This study also provided explanatory details that readers have certain expectations of look and design, writing style and appropriateness of information presented. Secondly, as our information sources are increasingly dominated by user generated content, the question of credibility or believability has become more relevant. Although blogs offer myriad opportunities for learning and gaining new information, improper consideration of the contents’ believability and reliability may lead to harmful consequences especially in riskier scenarios such as with medical information. This research is useful in highlighting the measures used by readers in believing the blogs and hence can be used for assessment of such potential harmful consequences and the formation of appropriate Web policies and education both formal and informal.

References 1. Andrews, P.: Is Blogging Journalism? Nieman Reports 57(3), 63–64 (2003) 2. Baumer, E., Sueyoshi, M., Tomlinson, B.: Exploring The Role Of The Reader In The Activity Of Blogging. In: Proceedings of ACM Conference on Human Factors in Computing Systems (CHI), pp. 1111–1120 (2008) 3. Berlyne, D.E.: Conflict, Arousal, And Curiosity. McGraw-Hill, New York (1960) 4. Blood, R.: How Blogging Software Reshapes The Online Community. Communications of the ACM 47(12), 53–55 (2004) 5. Bos, N., Olson, J., Gergle, D., Olson, G., Wright, Z.: Effects Of Four Computer-Mediated Communications Channels On Trust Development. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 135–149. ACM Press, New York (2002) 6. Chen, Y.C., Shang, R., Li, M.J.: The Effects of Traveler’s Blogs on the Behavioral Intention to Visit a Tourism Destination. In: International Conference on Business and Information, Malaysia (2009) 7. Chesney, T., Seong, D.S.K.: The Impact Of Anonymity On Weblog Credibility. Int. Journal of Human-Computer Studies 68, 710–718 (2010) 8. Doney, P.M., Cannon, J.P.: An Examination Of The Nature Of Trust In Buyer-Seller Relationships. Journal of Marketing 61, 35–51 (1997) 9. Egger, F.N.: Affective Design Of E-Commerce User Interfaces: How To Maximize Perceived Trustworthiness. In: Proceedings of the International Conference on Affective Human Factors Design. Academic Press, London (2001) 10. Eysenbach, G., Kohler, C.: How Do Consumers Search For And Appraise Health Information On The World Wide Web? Qualitative Study Using Focus Groups, Usability Tests, And In-Depth Interviews. BMJ 324, 573–577 (2002)

Interpreting User-Generated Content: What Makes a Blog Believeable?

89

11. Flanagin, A.J., Metzger, M.J.: Perceptions Of Internet Information Credibility. Journalism and Mass Communication Quarterly 77, 515–540 (2000) 12. Fogg, B.J., Marshall, J., Laraki, O., Osipovich, A., Varma, C., Fang, N., Paul, J., Rangnekar, A., Shon, J., Swani, P., Teinen, M.: What Makes Web Sites Credible? A Report On A Large Quantitative Study. In: CHI Conference Proceedings (2001) 13. Herring, S., Scheidt, L., Bonus, S., Wright, E.: Bridging The Gap: A Genre Analysis Of Weblogs. In: Proceedings 37th Annual HICSS Conference (2004) 14. Johnson, T.J., Barbara, K.K.: In Blog We Trust?: Deciphering Credibility of Components of the Internet among Politically Interested Internet Users. Computers in Human Behavior 25(1), 175–182 (2009) 15. Kang, M.: Measuring Social Media Credibility: A Study on a Measure of Blog Credibility. Institute for Public Relations (2010) 16. Lenhart, A., Fox, S.: Bloggers: A Portrait Of The Internet’s New Storytellers. Pew Internet & American Life (2006) 17. McKnight, D.H., Choudhury, V., et al.: Trust In E-Commerce Vendors: A Two-Stage Model. In: International Conference on Information Systems (ICIS). Association for Information Systems, Brisbane (2000) 18. Nardi, B.A., Schiano, D.J., Gumbrecht, M., Swartz., L.: Why We Blog. Communication ACM 47(12), 41–46 (2004) 19. Rubin, V., Liddy, E.: Assessing Credibility Of Weblogs. In: Proceedings of the AAAI Spring Symposium: Computational Approaches to Analyzing Weblogs, CAAW (2006) 20. Schiano, D.J., Nardi, B.A., Gumbrecht, M., Swartz, L.: Blogging By The Rest Of Us. In: Proceedings of CHI, Vienna, Austria (2004) 21. Trammell, K.D., Keshelashvili, A.: Examining The New Influencers: A Self-Presentation Study Of A-List Blogs. Journalism & Mass Communication Quarterly 82(4), 968–982 (2005) 22. Van House, N.: Weblogs: Credibility and Collaboration in an Online World (prepared for CSCW Workshop on Trust, unpublished) (October 2004) 23. Wall, M.: Blogging Gulf War II. Journalism Studies 7(1), 111–126 (2006) 24. Zheng, J., Veinott, E., Bos, N., Olson, J., Olson, G.: Trust Without Touch: Jumpstarting Long-Distance Trust With Initial Social Activities. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 141–146. ACM Press, New York (2002)

Extracting Events from Wikipedia as RDF Triples Linked to Widespread Semantic Web Datasets Carlo Aliprandi1, Francesco Ronzano2, Andrea Marchetti2, Maurizio Tesconi2, and Salvatore Minutoli2 1 Synthema Srl Via Malasoma 24 56121 Ospedaletto (Pisa) - Italy [email protected] 2 Institute of Informatics and Telematics (IIT) CNR Via G. Moruzzi, 1 56123 Pisa - Italy {francesco.ronzano,andrea.marchetti,maurizio.tesconi, salvatore.minutoli}@iit.cnr.it

Abstract. Many attempts have been made to extract structured data from Web resources, exposing them as RDF triples and interlinking them with other RDF datasets: in this way it is possible to create clouds of highly integrated Semantic Web data collections. In this paper we describe an approach to enhance the extraction of semantic contents from unstructured textual documents, in particular considering Wikipedia articles and focusing on event mining. Starting from the deep parsing of a set of English Wikipedia articles, we produce a semantic annotation compliant with the Knowledge Annotation Format (KAF). We extract events from the KAF semantic annotation and then we structure each event as a set of RDF triples linked to both DBpedia and WordNet. We point out examples of automatically mined events, providing some general evaluation of how our approach may discover new events and link them to existing contents. Keywords: Knowledge Representation, Knowledge Extraction, Semantic Web, Natural Language Processing, Semantics.

1 Introduction The core aim of the Semantic Web is to provide a set of methodologies, standards, technologies and best practices to make explicit the semantics that lies behind the data exposed over the Web. As a consequence, it is possible to support an easy and serendipitous automatic integration of the great variety of Web contents, thanks also to the exploitation of shared knowledge references like ontologies, lexicons and semantic resources. In this scenario, the Resource Description Framework (RDF) [1] and the Ontology Web Language (OWL) [2] currently constitute the two core W3C standards useful to respectively represent knowledge over the Web and to specify a formalized semantic reference frame. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 90–99, 2011. © Springer-Verlag Berlin Heidelberg 2011

Extracting Events from Wikipedia as RDF Triples Linked

91

Many methodologies to extract knowledge from Web contents have been proposed. They can be divided into two wide groups: - relation or fact extraction systems: they usually apply specific Web mining procedures to extract facts from online contents. Facts are usually represented as attribute-value pairs describing some feature of a given entity of interest, i.e. the population of a specific country. Extracted facts are usually shown to users as search results referring to the part of Web documents. Examples are WebKnox [3], the Grazer System [4] and TextRunner [5]. - interlinking systems: they mine information from poorly structured Web contents and, unlike the previous group of systems, they expose the extracted knowledge over the Web as RDF data. In this context, the Linked Data initiative [6] has defined a set of best practices to represent such knowledge by exploiting RDF and to unambiguously identify entities on a Web scale by means of URIs. A central role is played by DBpedia [7], an extract from Wikipedia contents, representing a hub for many other Linked Data datasets. Many other proposals to produce Semantic Web interlinked datasets have been made in parallel to Linked Data. They usually deal with some particular kind of Web contents and propose specific methodologies, like: - systems to enrich Web content: they apply procedures for keyword extraction or Named Entity recognition over Web pages, automatically producing a set of relevant terms to be annotated through the URI of the referred Wikipedia/DBpedia entity. Open Calais [8], for example, parses documents and points out entities, facts and events. When possible, entities are linked to DBpedia, Freebase or GeoNames URIs. Wikify [9] performs keyword extraction from Web pages, and disambiguates mined terms linking them to the referred Wikipedia entity. - systems to enrich social tagging service: systems like Faviki [10] and LODr [11] allow users to associate a tag to the Wikipedia page describing the referred concept. A different approach to automatically connect user tags to a specific concept of Wikipedia has been adopted by [12], exploiting also Tagpedia [13], a semantic resource for Tag Sense Disambiguation. Considering the attempts to build semantic resources by mining Wikipedia, [14] describes a set of methodologies adopted to extract from Wikipedia an association thesaurus, by exploiting the set of internal links, the taxonomy of Wikipedia categories and by mining the contents of each article. [15] describes how to build Tagpedia, a semantic reference useful to support the disambiguation of tags, by mining the structure of Wikipedia articles. Proposals to extract text snippets from Wikipedia representing facts have been also defined. Specific extraction techniques have been tuned to gather relevant text snippets from other articles [16] or to mine Wikipedia Named Entities over time [17]. Mining of Wikipedia has also been carried out by applying Natural Language Processing: a dump of the English Wikipedia has been shallow parsed and semantically annotated [18]. Applying both shallow and deep parsing to Wikipedia, methodologies to build a common sense knowledge base have been proposed [19].

92

C. Aliprandi et al.

In this paper we propose an approach to mine the unstructured textual data of Wikipedia so as to extract events, representing them through sets of RDF triples and integrating them in the Linked Data dataset. In section 2, after having clarified the notion of event, we present the KYOTO Annotation Format (KAF) and we describe Synthema English Slot Grammar, a deep parser used to automatically produce KAF annotations. In section 3 we describe the procedure to identify events and to represent them through RDF triples exploiting also WordNet and DBpedia. In section 4 we provide some meaningful examples of this procedure as well as some initial evaluation. We conclude in section 5, discussing our future plans to improve the event extraction process from KAF annotated texts.

2 Annotating Textual Documents by Exploiting KAF The goal of this paper is to explore the possibility to mine events from linguistically and semantically annotated textual documents and to try to propose possible ways to represent them as sets of RDF triples tightly connected with core Semantic Web datasets like DBpedia. For this reason we need first of all to define what we mean by an event, but also to specify a possible RDF representation of events as useful Semantic Web knowledge. We assert that an event is something that happens having some relevance in providing information in a particular context: it could be characterized by specific spatial and temporal coordinates. As a consequence, an event is usually built around a specific action or happening. In order to mine events from documents and, in particular, from the contents of Wikipedia articles, we consider the results of their linguistic and semantic analysis encoded in KAF [20], the deep semantic annotation format that we developed in the context of the KYOTO Project. KAF is a language neutral annotation format representing both morpho-syntactic and semantic annotation of documents through a layered structure. Starting from the lower of all its annotation layers, where tokens, sentences and paragraph are identified, in KAF each additional layer is built on top of the lower one, referring to its constituent elements. In this way, several levels of text annotation can be added by different linguistic processors. Moreover, specialized linguistic processors can be developed to generate incremental annotations for each specific layer. In KAF there are three macro-layers of document annotation (see also Figure 1): • morpho-syntactic layer: it groups all the language-specific text annotations. Tokens, sentences and paragraphs are identified in a specific document. Terms made of words or multi-words are pointed out, along with their Part Of Speech. In this layer also chunks and functional dependencies are represented. • level-1 semantic layer: it includes linear annotation of expressions of time, events, quantities and locations. • level-2 semantic layer: it is mainly devoted to represent facts, in a non linear annotation context, thus possibly aggregating evidences from the lower layers of multiple textual sources.

Extracting Events from Wikipedia as RDF Triples Linked

93

Fig. 1. The three macro-layers of KAF document annotation

Our RDF event extraction process mainly exploits the results of the morphosyntactic annotation layer of KAF. In this layer, the following elements are annotated: • word forms: each word form is unambiguously determined and linked to the sentence and paragraph of the text it belongs to. • terms: terms, also composed of two or more word forms, are identified and characterized by lemma, Part Of Speech and, when possible, by the type of referred Named Entity. The link of a term to an external reference can be represented. This is the KAF feature that we have mostly exploited to represent term meaning, either as a WordNet concept, disambiguated by proper WSD, or as a DBpedia entity. • dependency relations: they represent functional relations among terms, such as Agent, Action, Object, Qualifier, When, Where, How. • chunks: they are used to identify structured phrases, spanning one or more terms, like noun phrases, verbal phrases and prepositional phrases. Synthema English Slot Grammar (Syn ESG), an inverse parallel deep parser, is used to automatically produce KAF annotations. It carries out complex Natural Language Processing (NLP) tasks like word tokenization, segmentation, Part Of Speech tagging, dependency parsing, anaphora resolution and functional analysis. Syn ESG is intended to identify relevant knowledge from the raw text, by detecting concepts and semantic relations in texts. Concept extraction and text mining are applied through a pipeline of linguistic and semantic processors that share as a common ground McCord's theory of Slot Grammar [21]. Syn ESG parser - a bottomup chart parser - employs a parse evaluation scheme used for pruning away unlikely analyses during parsing as well as for ranking final analyses. By including semantic information directly in the dependency grammar structures, the system relies on the lexical semantic information combined with functional relations. Besides Named Entities, locations, time-points and dates, Syn ESG detects relevant information like

94

C. Aliprandi et al.

chunks, noun phrases and verbal phrases. The detected terms are then extracted, reduced to their Part Of Speech and functional tagged base form. Finally, syntactically and semantically tagged words are properly encoded in the corresponding KAF annotation, and the specific KAF layer is produced.

Fig. 2. Example output produced by the Syn ESG text parser

In Figure 2 we show the output sample for the sentence: “The Battle of Cannae took place near the town of Cannae in Apulia, during the Second Punic War”, taken from the Wikipedia article describing the Battle of Cannae. The figure shows functional dependencies among terms. Note the reference resolution for the word “Cannae”, that is correctly co-referred to a unique URI (represented by the id #2583). Starting from the described KAF features, in the following sections we detail the extraction of events from KAF annotated documents as well as their representation as RDF triples.

3 Events Extraction and RDF Representation As mentioned before, in a KAF annotated text, mined terms are linked, thanks to a specific Word Sense Disambiguation algorithm, to either WordNet or DBpedia. Summarizing, we assert that the meaning of a term in KAF documents can be denoted by a synsetID of WordNet and/or by a URI of an entity of DBpedia. Unlike DBpedia URIs, WordNet synsetIDs may represent ambiguous Web identifiers. In this paper we are proposing a tentative URI schema for WordNet synsets: the URI that identifies a synset of English WordNet version 3.0 is ‘http://www.kyoto-project.eu/wordnet/ English/30/synsetID’. In the next section we describe how to extract events from KAF documents and how to represent them as sets of RDF triples.

Extracting Events from Wikipedia as RDF Triples Linked

95

3.1 Identifying Events Inside KAF Annotated Documents To extract RDF representations of events from a KAF annotated document, we mainly consider the terms linked to WordNet synsets and to DBpedia entities, and the dependency relations linking terms. We base our event extraction process on the assumption that the nucleus of an event is an action expressed by a verb, i.e. by a term that has been classified as a verb and associated to a WordNet synset. We also assume that the features that characterize an event are included in and represented by the set of dependency relations of the sentence the verb belongs to and connected in some way to the same verb. In particular we consider the dependency relations of subject (the entity that performs the action), object (the entity affected by the action), when (time or interval of happening of the action) and where (place of happening of the action).

Fig. 3. Set of dependency relations characterizing an event

To identify an event we need a set of dependency relations, in particular we need at least a subject or an object dependency relation. The connection of terms with WordNet or DBpedia URIs is essential in order to represent events as Semantic Web RDF triples linked with other datasets. In Figure 3 the set of dependency relations that can be exploited to characterize an event is schematized. Each term is connected to WordNet or DBpedia except the time-interval characterizing the when dependency relation, which is a literal. Given the previous set of dependency relations, in the next section we describe how to represent events as a set of RDF triples. 3.2 Representing Events as RDF Triples Exploiting DBpedia URIs Each event that we identify in a document is characterized by a subset of the dependency relations shown in Figure 3. In addition, the verbal WordNet synset defining the action of the event identified by its synset URI and the terms of the subject, object and where dependency relations are identified by a URI of a WordNet synset or of a DBpedia entity. The when dependency relation is described by a string specifying a time or interval. We can express an event as a set of RDF triples by exploiting ontological properties taken from the DBpedia Ontology or from a properly structured ontology of dependency relations. In what follows we assume use of an ontology, referred

96

C. Aliprandi et al.

to as Dependency Relation Ontology, properly published at the stub ‘http:// dependencyrelat.org/ontology/’ namespace. The when and where dependency relations can be respectively expressed by the ‘http://dbpedia.org/ontology/Date’ and ‘http://dbpedia.org/ontology/Place’ OWL properties of the DBpedia Ontology. The subject and object dependency relations can be respectively represented in the Dependency Relation Ontology by the ‘http://dependencyrelat.org/ontology/Subject’ and ‘http://dependencyrelat.org/ontology/Object’ properties.

Fig. 4. RDF representation of an event

As a consequence, we obtain the RDF representation of an event shown in Figure 4. The two triples describing the subject and the object of an event are reified and thus referenceable by a URI: we assume to use for this purpose URIs published at the stub ‘http://linkedevents.org/events/’ namespace. The two URIs describing these RDF triples are in turn grouped under the same RDF Bag Container and thus both referenceable through a third URI that is also assumed to be published at the same stub namespace. This URI, referred to as the event URI, obtained by aggregating in the same RDF Bag Container the URIs defined reifying the subject and object triples, is represented in Figure 4 by the dashed line. It points out the core features of an event and represents the subject of the RDF triples describing the place and time of happening of that event (properties ‘http://dbpedia.org/ontology/Place’ and ‘http://dbpedia.org/ontology/Date’). To preserve in the RDF representation, also the URL of the Web page the event has been mined from, we also represent the event URI by means of the stub ‘http://dependencyrelat.org/eventProp/fromURL’ ontological property. To conclude, we can notice that the two RDF triples pointing out the subject and the object of an event could be collapsed into a single one if we manage to represent WordNet verbs, constituting the core actions characterizing an event, as ontological properties: we can exploit the hypernym/hyponym taxonomy of WordNet verbs to define a hierarchy of subsumed ontological properties related to them (i.e. in English WordNet 3.0 the verb consume subsumes the verb eat). Each of these properties

Extracting Events from Wikipedia as RDF Triples Linked

97

contributes to the definition of the main RDF triple of an event by linking the WordNet/DBpedia URI of the subject of the action to the WordNet/DBpedia URI that points out the object, in this case through the URI of the ontological properties representing the verb describing the actions.

4 Example of Mined Events To gather initial examples of facts, we annotated a set of Wikipedia articles, mainly related to wars and battles. Here we show a significant example of mined events from a sentence extracted from the Wikipedia page describing the ‘Battle of Gettysburg’:

Fig. 5. Concepts mined from the sample sentence: “The two armies began to collide at Gettysburg on July 1, 1863, as Lee urgently concentrated his forces there”

Figure 5 shows terms and dependency relations mined from the sentence. Two sentences are identified: “The two armies began to collide at Gettysburg on July 1, 1863” and “Lee urgently concentrated his forces there.”. Moreover, five terms are connected to WordNet synsets and two terms are identified through DBpedia entities. The date is a literal and thus managed as a string (July 1, 1863). From this sentence we can mine two events; their RDF representation is shown in Figure 6. Notice that two DBpedia URIs are exploited to refer to the two terms of the sentence: 1. Gettysburg: http:// dbpedia.org/resource /Gettysburg,_Pennsylvania 2. Robert E. Lee: http:// dbpedia.org/resource /Robert_E._Lee Moreover, the URI of four WordNet Synsets are exploited to represent the RDF triples of these events. In particular, the event related to the first sentence is described by a subject- The corresponding RDF triple is reified and further characterized by the place and date of happening of that event (Gettysburg, July 1st 1863). The event related to the second sentence is characterized by a subject and an object but no place and date are specified. Both events are linked to the URL of the Wikipedia page they have been mined from, through the RDF property: ‘http://dependencyrelat.org/eventProp/fromURL’.

98

C. Aliprandi et al.

Fig. 6. RDF representation of mined events from the sample sentence

5 Conclusions and Future Work In this paper we have presented an approach to extract events from documents, annotated using a deep text parser according to the KYOTO Annotation Format. We have defined an event extraction methodology that takes as input KAF annotated documents: it is based on functional dependencies and on the results of disambiguation of the terms through WordNet synsets or DBpedia entities. We have defined a representation of the mined events, as a set of RDF triples exploiting the URI of WordNet and DBpedia to point out the different entities taking part into events. We have provided a significant example of event mining procedure starting from a Wikipedia article. We are carrying out a global evaluation of the effectiveness of our event mining procedure by processing a larger set of Wikipedia articles. In this way we plan to provide also global quantitative evaluations of the quality of events. Even though the event extraction activities are at an early stage, we believe that the methodologies described can be useful to automatically produce Semantic Web data linked with other datasets. In particular, by applying this process to DBpedia, a considerable amount of new RDF triples can be generated. In conclusion, we can state

Extracting Events from Wikipedia as RDF Triples Linked

99

that our approach represents an example of synergy between NLP techniques and the Semantic Web to produce and link over the Web semantically-grounded contents. Acknowledgments. This work is partially funded by the European Commission (KYOTO project, ICT-2007-211423).

References 1. RDF W3C Web Page, http://www.w3.org/RDF/ 2. OWL W3C Recomm., http://www.w3.org/TR/owl-features/ 3. Urbansky, D., Thom, J.A.: WebKnox: Web Knowledge Extraction. In: 13th Australasian Document Computing Symposium, Hobart (2008) 4. Zhao, S., Betx, J.: Corroborate and Learn Facts from the Web. In: 13th International Conference on Knowledge Discovery and Data Mining, San Josè (2007) 5. Banko, M., Etzioni, O.: The Tradeoffs Between Open and Traditional Relation Extraction. In: 46th ACL: Human Language Technologies, Columbus (2008) 6. Linked Data Web Site, http://linkeddata.org/ 7. DBpedia Web Site, http://dbpedia.org/About 8. Open Calais Web Site, http://www.opencalais.com/ 9. Wikify! Web Site, http://www.wikifyer.com/ 10. Faviki Web Site, http://www.faviki.com/ 11. Passant, A.: LODr - A Linking Open Data Tagging System. In: Social Data on the Web Workshop at the 7th Int. Semantic Web Conference, Karlsrhue (2008) 12. Tesconi, M., Ronzano, F., Marchetti, A., Minutoli, S.: Semantify del.icio.us: automatically turn your tags into senses. In: Social Data on the Web Workshop at the 7th International Semantic Web Conference, Karlsrhue (2008) 13. Tagpedia Web Site, http://www.tagpedia.org/ 14. Nakayama, K.: Extracting Structured Knowledge for Semantic Web by Mining Wikipedia. In: Social Data on the Web Workshop at the 7th International Semantic Web Conference, Karlsrhue (2008) 15. Ronzano, F., Marchetti, A., Tesconi, M., Minutoli, S.: Tagpedia: a Semantic Reference to Describe and Search for Web Resources. In: Social Web and Knowledge Management Workshop at the 17th World Wide Web Conference, WWW 2008, Beijing (2008) 16. Adafre, S.F., Jijkoun, V., de Rijke, M.: Fact Discovery in Wikipedia. In: IEEE/WIC/ACM International Conference on Web Intelligence, Silicon Valley (2007) 17. Bhole, A., Fortuna, B., Grobelnik, M., Mladenic, D.: Mining Wikipedia and Relating Named Entities over Time. In: 13th International Conference on Knowledge Discovery and Data Mining, San Josè (2007) 18. Asterias, J., Zaragoza, H., Ciaramita, M., Attardi, G.: Semantically Annotated Snapshot of the English Wikipedia. In: 6th International Language Resources and Evaluation Conference LREC 2008, Marrakech (2008) 19. Suh, S., Halpin, H., Klein, E.: Extracting Common Sense Knowledge from Wikipedia. In: 6th International Semantic Web Conference, Athens, GA, USA (2006) 20. Bosma, W., Vossen, P., Soroa, A., Rigau, G., Tesconi, M., Marchetti, A., Aliprandi, C., Monachini, M.: KAF: a generic semantic annotation format. In: 5th International Conference on Generative Approaches to the Lexicon, Pisa (2009) 21. McCord, M.C.: Slot Grammar: A System for Simpler Construction of Practical Natural Language Grammars. Natural Language and Logic, 118–145 (1989)

Collaborative Sensemaking during Admin Permission Granting in Wikipedia Katie Derthick1, Patrick Tsao2, Travis Kriplean2, Alan Borning2, Mark Zachry1, and David W. McDonald3 1 Human Centered Design & Engineering, University of Washington Department of Computer Science & Engineering, University of Washington 3 The Information School, University of Washington Seattle, Washington, USA {derthick,pattsao,kriplean,borning,zachry,dwmc}@uw.edu 2

Abstract. A self-governed, open contributor system such as Wikipedia depends upon those who are invested in the system to participate as administrators. Processes for selecting which system contributors will be allowed to assume administrative roles in such communities have developed in the last few years as these systems mature. However, little is yet known about such processes, which are becoming increasingly important for the health and maintenance of contributor systems that are becoming increasingly important in the knowledge economy. This paper reports the results of an exploratory study of how members of the Wikipedia community engage in collaborative sensemaking when deciding which members to advance to admin status. Keywords: contributor systems, system administration, collaboration tools, sensemaking.

1 Introduction Decision making in an online community is often a difficult process. In large online communities like Wikipedia, administrators are key to keeping the community functioning, and those administrators come from the ranks of regular users. The decision to grant administrative permissions has important consequences for the entire community. In the process of deciding which candidates should be granted administrative permissions, reviewers must consider a user’s behavior relative to a set of criteria and come to some reasonably shared understanding of the merits of a new potential administrator. This study examines the collaborative activity of an online community deciding who is given administrative privileges. Understanding the processes of admin promotion in Wikipedia by its own members is challenging. Prior work has considered the edit histories of contributors to identify the characteristics of strong admin candidates (Burke & Kraut, 2008). A complementary study (Forte et al., 2009) provides an insider’s perspective on the process, using interviews with administrators and editors to reveal that the deliberation process has become increasingly difficult and unappealing to candidates. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 100–109, 2011. © Springer-Verlag Berlin Heidelberg 2011

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

101

These and other studies (e.g., Leskovec et al., 2010) illustrate that the decision making is complex, needing to account for diverse data and subject to intense scrutiny. However, researchers have yet to consider how the processes and tools available in deliberations about candidates for administrator status are employed. Our work is intended to help close this gap by considering the role of shared analytical tools in collective sensemaking (Heer & Agrawala, 2008).

2 Method We conducted an exploratory study to understand the processes, tools, and information that Wikipedians use to decide the outcome of Requests for Adminship (RfA) cases. In our study, we collected and analyzed three sources of data: (1) semistructured interviews with active Wikipedia editors who participate in the RfA decision process, (2) the archived record of successful and unsuccessful RfAs as maintained by Wikipedia, and (3) a review of the tools that Wikipedians use for viewing and distilling the contributions and activities of an RfA candidate. We designed this multi-perspective view of the RfA process to gain a rich understanding of the interplay among people, processes, and tools in the work of collaborative sensemaking about the promotion of select community members to admin status. 2.1 Interviews We conducted a series of semi-structured interviews with 10 experienced editors to discover how they think about their online interactions with other Wikipedians, and particularly how they develop their understandings of specific individuals. A primary focus area in these interviews was participation in the RfA process, including typical ways of participating in RfA cases and the use of different information tools (e.g., edit history counters). The interviewees worked from locations across the U.S., so all were interviewed remotely, using either telephone, instant message, or a combination of these methods. Interviews were transcribed and each transcript was coded by multiple team members using an open theme coding process. 2.2 Content Analysis We reviewed the online discussion transcripts of 6 RfA cases, examining instances in which candidates were approved, in which candidates were denied administrator status, and in which the candidate withdrew. We discussed all the transcripts individually and then identified prominent themes in the deliberative exchanges. 2.3 Tool Review When an RfA case is presented to the community for consideration, a set of tools are provided on the nomination page, allowing participants to explore different user characteristics and counts of system activities for the admin candidate. See Figure 1 for an example of the basic count- and history-based tools that point to information about all candidates. This group of tools offers a set of links to information on regular and special pages in Wikipedia. The one exception in this group is “count [quick],” which resides on Wikimedia Toolserver.

102

K. Derthick et al.

Fig. 1. A set of link-based tools to regular and special pages in Wikipedia that provide counts in different categories related to a given admin candidate’s actions and related status indicators in the system.

This predefined set of tools appears as a block of links on the nomination/ deliberation page for each admin candidate. Figure 2 below shows an additional “toolbox” of links available at the end of each RfA nomination. These links all point to tools residing outside Wikipedia itself on the Wikimedia Toolserver. Like the information tools in Figure 1, these tools allow reviewers to consider the candidates’ actions based on count-based criteria, some of which extend beyond Wikipedia itself. We examined this tool set, observing the affordances of each available tool for supporting an editor attempting to arrive at an evidence-based view of a candidate administrator.

Fig. 2. A toolbox provided with each RfA nomination, allowing reviewers to explore different dimensions of the candidate’s actions and presence in Wikipedia and other Mediawiki projects.

3 Results To understand how Wikipedia participants decide who should be promoted to admin status, we considered the formal process from the perspective of the participants, from evidence of their deliberative work, and from the system tools used to support the process. Our interviews and reviews of processes and tools yielded detailed insight into the RfA process. These results are presented below with selected quotes to illustrate each of three themes that emerged from our analysis: styles of interaction,

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

103

social networks, and counts of online activity versus the contexts in which those activities occur. These results are then explored in the discussion section. 3.1 Styles of Acting and Interacting A primary means by which an online community develops an understanding of its members is by interpreting the actions and interactions of individual participants. A sense of who a person is develops as that individual’s behavior is assessed—whether that behavior is related to individual acts performed in the shared space of the system, or whether it is his or her behavior when engaged in dialogic exchanges with others. Within Wikipedia, an editor’s actions and interactions are considered important sources of evidence when the community deliberates about the potential promotion of a candidate editor to admin status. Actions and interactions on user talk pages, on article talk pages, in edit summaries, and in private backchannels such as email or IRC chats, become evidence about which the community can deliberate. When deliberating about RfA candidates, editors seek evidence that an editor has been civil when engaging with others during their work. As one interview explained: I tend to value civil interactions. . . . Wikipedia has a few people who are good content contributors, but are also possessed of tempers and tend to blow up at people. In [such] case[s], it’s weighing value: are their contributions worth the people who might leave because they belittle or harass them, or the environment it creates for others? – Dylan Editors who engage in controversial issues with a level-headed approach are also viewed favorably, particularly since working productively in a heated exchange is a necessary part of doing administrative work. As one editor explained, Participation in highly controversial articles will gain people ideological enemies, but at the same time present a very stringent test of the person's ability to deal with complex and emotive issues, interact[ing] with disruptive editors. [These] qualities are often in demand of administrators. – Taylor Yet another kind of interaction style that is considered when admin candidates are being reviewed is how they have engaged “newbies,” whose initial contributions are sometimes more damaging than valuable. A productive, encouraging interaction style when dealing with new editors is important because the community as whole depends on the infusion of new contributors to conduct the overall work of the system. Admin candidates, thus, are expected to exhibit an interaction style that encourages new contributors to work productively within community norms. One interviewee used a personal example to illustrate how this interaction style may be exhibited when interacting with a new contributor who appears to be engaged in disruptive editing: Well, they get the standard warning template(s) that all vandals get, but I’ll also add a personal note underneath the template. Nothing too cuddly or kind -they're disrupting the encyclopedia, after all -- but “I’ll grant you, that was hilarious...but please stop before I have to report you to the admins” is the type of language I'd use. What I don’t do is advise them on contributing in a constructive way -- I figure that even if they have the means they pretty much need to decide that on their own, as I did. – Bill

104

K. Derthick et al.

As this quote suggests, the ways that an admin candidate has interacted with newbies is scrutinized to ensure that the candidate is oriented more toward enculturating newcomers than in executing the rules of Wikipedia. Interpreting such past interactions, however, is not easy. A strong candidate for adminship must know the system rules and norms, but also find productive ways of interacting with people: If they’ve nominated a new editor’s bad page for deletion, [that action] is often the subject of scrutiny at RfA. – Dylan While much of the sensemaking that reviewers engage in while considering admin candidates focuses on arriving at a collective assessment of how productive their interaction styles have been, there is a related consideration about the network of editors within which the candidate is interacting. 3.2 Social Networks in Sensemaking When developing shared understandings in their deliberations about admin candidates, Wikipedians consider social networks. Participants described how they conceptualized the nature of social bonds, how those bonds are built, and how those social bonds influence the way they come to understand what others are saying. What Constitutes the Social Network? The creation of a social network in a system like Wikipedia is about establishing social bonds. While some participants talked about meeting “random” people who just happened to be interested in similar topics, others didn’t always see it that way. Some form of homophily emerges when people show up at a topic and begin to express similar ideas or when people come to a Wikipedia policy page and begin to discuss a policy decision. Of the many interactions participants might have, some begin the formation of social bonds, and some people become “friends.” When asked about how he defined friendship on Wikipedia, one interviewee explained: I have real-world friends I interact with on Wikipedia, but by the standard of interactions on-wiki, I wouldn’t call them “wiki-friends”. “Friendship” online is held to a lower standard than the real world, so I suppose you could classify a wiki-friend as someone you regularly associate with in a common topic or area of interest, and agree with, or have a rapport with. – Dylan In Dylan’s view, friendships in Wikipedia are based in shared interests, having rapport and seeing things generally the same way. Another participant elaborated on this point to include collaboration on projects and shared work as characteristic of building a social bond: Every editor has a group of friends they talk with, collaborate with, create projects with, and so forth. My group consists of editors with whom I enjoy speaking, people I agree with on policies, people whose talk pages I watch, people I trust to take the right course of action on matters dealing with policy. – Manny The comment by Manny reinforces many of the common conceptions of how the social network is built—the idea that repeated interactions and the enjoyment of those interactions are important. Additionally, as his quote indicates, bonds form around having similar ideas about Wikipedia policies. Manny mentions that agreeing on

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

105

policy results in trust with policy decisions. Manny also mentions using a watchlist to watch for changes in other users talk pages. Watching is an important part of building a stronger social bond. Dylan explained the watchlist concept at length: I should point out that a lot of Wikipedia relationships are defined by who watchlists what pages. That is, Person A has Person B’s user talk page on his or her watchlist, and so when Person C comments on Person B’s page, Person A might chime in ... Without watchlists, keeping up to date on communication or changes is reduced to manual refreshing, which inhibits further communication or interaction. So wiki-friends or collaborators often have similar watchlist circles, reinforcing the same types of connections. – Dylan Many of our participants had the everyday conception of the social network as simply individuals who interact frequently with each other. That knowing of one another through interaction can result in two types of judgments. Given that Wikipedia is big, it is unlikely that an observer can determine the complete social network, which may not be well connected. Influence of Social Networks. The potentially disconnected nature of a social network has implications for sensemaking in the RfA process. In that process, first the candidate for administrative privileges is nominated. Other editors can ask questions and comment on the candidate. The social network of the individuals who nominate, support, or oppose a candidate is one aspect of how others involved in the deliberation come to understand what a candidate has done and how to interpret his or her responses to questions. In one interview, Taylor talks about knowing who is nominating and supporting a particular candidate. In the quote Taylor is trying to explain one way to understand a person’s reputation. Taylor says: “Reputation” is hard to qualify or explain, but you start to recognize editors and see how they act and what they say. If you see a candidate that has been nominated by several well-known editors who you have always found to make useful and sensible comments, this will suggest that people whose judgment you trust have a positive opinion about the candidate. This isn’t mentioned in the official criteria, but I’m sure it plays a large part in shaping people’s first impressions. – Taylor Editors find that a candidate’s circle of friends is useful for understanding RfA candidates: In a much more succinct way this same comment was made by Marshall who says “To some degree I look at how other editors whose opinion I usually agree with have voted.” The previous two comments show that both the social network around the candidate and the social network around the reviewer matter. The intersection of these social networks is hopefully the people who are most qualified to judge the many aspects of the candidates’ attributes. While the social network around the candidate and the reviewer are important, Taylor notes that the network is not adequate as the sole basis for a judgment:

106

K. Derthick et al.

This is probably a particularly useful shortcut since Wiki[pedia] is such a large place and the number of people you interact with directly is a tiny fraction of the whole. However, like I say this is a shortcut that may well mislead, so I wouldn’t make a decision based entirely on who supports or opposes, it matters very much what they say and what the candidate says! And of source on what the candidate has done. – Taylor In addition to social network information, activity counts and the contexts for interpreting them are important in the process of deliberating about RfA candidates. 3.3 Considering Counts and Contexts Wikipedia editors often strive to live up to a set of shared values, such as Assume Good Faith, as they sense of the behaviors of other editors, particularly when they are making some decision that has consequences for that editor (e.g., when he or she is an admin candidate). Thinking broadly about an editor, however, can be a difficult information-seeking endeavor considering the vast edit histories that many people have. To address this challenge, participants rely on tools to gain insight into an editor’s history, such as edit counts. Such counts can then be considered in defined distributions across time, articles, or namespaces. Counts thus sometimes perform a thresholding function: I check their edits. I see what they’ve edited, where they’ve edited. Like if you’ve only got a few hundred edits, you’re likely not completely familiar with Wikipedia enough to become admin. Similarly, if you’ve got thousands of edits, but only to a small section, such as only articles about one TV show, or one sports team, or something, you’ve not had enough interaction throughout the entire site to be familiar with it all – Marshall In considerations of RfA candidates, raw edit counts are sometimes applied to specific kinds of work, such as consideration of Articles for Deletion (AfDs): I’ll oppose if they don’t have a track record in AfDs, for example, because I think it’s important for admins to know how to “properly” judge and close AfDs. – Dylan Raw edits counts are used to gauge distributions of work across the system: In general, I like to see a good “mix” in contributions. I generally vote against candidates whose major contributions are to Wikipedia space, instead of articles. – Dylan However, while counts can give a baseline sense of the attention and work of a candidate in the system, our interviewees also noted that attention should be paid to qualities of actions rather than just their counts. This inclination for contextual interpretation of actions is illustrated in Bill’s reflection on his sensemaking habits: An important thing to consider when interacting with any editor -- particularly one who appears to be acting disruptively, like a vandal -- is to look at their contributions. Not all editors do this...frankly, I suspect most don’t. But I always do. Sometimes someone who appears to be a vandal, or a provocateur, or just

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

107

someone who doesn’t have good intentions, is actually acting in perfectly good faith. You can infer this if you are about to warn someone for vandalism but look at their user contributions beforehand and note that, say, up until a day ago this person was doing great work... So perhaps what looks like vandalism, or harsh words, or whatever, to me... perhaps that’s just someone coming across the wrong way. – Bill Unfortunately, existing tools that Wikipedians use do not assist in the iterative exploration between summations and contextualized action. Even when data is provided that goes beyond numbers, it can be difficult for editors to evaluate whether the data is part of a pattern that accurately characterizes the editor, or an isolated event. This evaluation is complicated by the fact that other editors are often involved in the process of evaluating the editor in question, providing selected diffs as evidence in the discussion forum where the candidate is being considered. A “diff” is a page in Wikipedia that shows the differences between edited versions of a page. The diffs an editor chooses to provide may not be motivated by a principle as respected as Assume Good Faith, where an editor tries to create a false impression of a pattern in another editor. Taylor illustrates this point with a hypothetical case where one editor attempts to persuade others that a candidate exhibits bad behavior: Imagine somebody says “X is incredibly rude; look at this comment here (link).” You click on the link and it shows somebody saying something that might be interpreted as rude, but you lack all context. It is like seeing everything people say in sound bites rather than conversation. – Taylor Consequently, he elaborates, the work of principled sensemaking about the candidate requires additional work: You need to do a lot of detective work. For instance, you can see when the comment was made and search back to find the conversation in a talkpage’s archives. You can then read the context and perhaps find out that they were being sarcastic, or making a joke to a friend. – Taylor Actions often need to be considered in their context to be appropriately evaluated in a fashion that aligns with the evaluator’s values. But contextualizing actions is difficult in hindsight, particularly when the evaluator was not present when the action was taken. Of course, this qualitative/quantitative tension in large data analysis is a familiar tension for academic researchers.

4 Discussion Through our study we identify themes that are critical to decision making in the Wikipedia Request for Adminship process. The interview data yielded insight into the different understandings of what community participants should be sensitive to when considering the promotion of someone from their ranks into an administrative role. What participants in the RfA process consider important in their decision varies notably from contributor to contributor. Although within Wikipedia contributors can review community-produced lists and commentary about what makes a good

108

K. Derthick et al.

administrator, there is no requirement that contributors consult such materials when participating. As one participant pointed out, even if such lists or commentaries are consulted, participants in the RfA decision weigh all considerations differently. Further, what a given participant values most in RfA considerations shifts over time. The editors who actively participate in RfA cases (the “regulars”) shifts over time, changing the collective sensitivity to different potential considerations. Through an analysis of RfA cases and the tools that are used in the process, we identified techniques that reviewers use to focus the attention of others involved in the deliberation on specific evidence to support their interpretations, and how that evidence can be a source of shared understanding and differing opinions. The deliberation about whether to promote a candidate to adminship is facilitated within the system itself via an asynchronous wiki-based discussion. The deliberation is public and open to participation from the entire community. The candidate is also expected to participate. The candidate answers a set of questions, some of which are standard, some of which are focused on key events in the community and perspectives that the candidate may have about the practices of the community. Following the questions, discussion ensues where participants use evidence to argue for “support” or “oppose” positions. Sensemaking thus occurs as a collaborative effort of the community as they collectively consider the merits of the candidate. Sensemaking is currently an artifact-based process in which selected pieces of evidence available in the system are introduced and interpreted by the deliberators. These interpretations are subject to debate as participants decide which evidence is relevant to their decision and what the evidence means. The deliberative forum is seeded with a set of tools for retrieving the most commonly considered evidence in deliberations. These tools include such things as a counter of past contributions in different segments of the system and a history of the candidate’s status since joining the system. As the central means of generating shared artifacts for consideration, these tools play a significant role in how sensemaking occurs in online systems. One key part of this sensemaking is the shared interpretation of evidence. The technical tools are very effective at identifying “diffs” contributed by the candidate. Isolated contributions rarely tell the entire story. During deliberation, participants use the tools to explain and provide additional details that contextualize interpretations of the candidate’s actions. Some of the more hotly contested deliberations surround discontinuous behavior by the candidate, such as when a candidate has an early history of being a poor member yet changes to be a strong and valuable member.

5 Conclusion and Future Work Our results suggest a number of important requirements for the design of tools to support collaborative decision-making in large-scale online communities. In particular, systems like Wikipedia suffer from their wealth of activity and behavior data. While each action is recorded by the system, the relationships among interrelated actions are difficult to uncover and interpret. Thus, new tools are needed that can help large-scale communities understand and interpret mass interaction data. Acknowledgments. This research was supported by NSF award IIS-0811210.

Collaborative Sensemaking during Admin Permission Granting in Wikipedia

109

References 1. Burke, M., Kraut, R.: Taking up the mop: Identifying future Wikipedia administrators. In: Burnett, M., Costabile, M., Catarci, T., Ruyter, B., Tan, D., Czerwinski, M., Lund, A. (eds.) Proceedings of the Twenty-Sixth SIGCHI Conference on Human Factors in Computing Systems, pp. 3441–3446. ACM Press, New York (2010) 2. Forte, A., Larco, V., Bruckman, A.: Decentralization in Wikipedia Governance. Journal of Management Information Systems 26(1), 49–72 (2009) 3. Heer, J., Agrawala, M.: Design considerations for collaborative visual analytics. Information Visualization Basingstoke- 7(1), 49–62 (2008) 4. Leskovec, J., Huttenlocher, D., Kleinberg, J.M.: Governance in Social Media: A case study of the Wikipedia promotion process. In: Artificial Intelligence, pp. 98–105. AAAI, Menlo Park (2010), http://arxiv.org/abs/1004.3547

Mining Social Relationships in Micro-blogging Systems Qin Gao, Qu Qu, and Xuhui Zhang Department of Industrial Engineering, Tsinghua University. 100084, Beijing, China [email protected]

Abstract. The widespread popularity and vigorous growth of micro-blogging systems provides a fertile source for analyzing social networks and phenomenon. Currently, few data mining tools can deal with unique characteristics of microblogging systems. In this study, we propose an integrate approach for mining user relationships in micro-blogging systems. The approach starts from macroscopic analysis of social networks by grouping users with the method of maximal strongly connected components (MSCC). Following that, a measure of condensation level of groups are calculated to find out the most influential group , and all groups can be ranked according to this measure; then a new algorithm is presented to evaluate the influence of a specific user within a group. The integrated approach is capable to analyze large amount data sets. It is useful for exploring directions of information diffusion and evaluating the scope and the strength of individual user’s influence in micro-blogging systems. Keywords: Social data mining, micro-blogging systems, information diffusion analysis, and graph mining.

1 Introduction Micro-blogging systems, like twitter.com, let users write a brief text about daily life and update immediately through many different ways, including typing online, text messaging, instant messaging (IM), and sending emails. In recent years, microblogging systems become increasingly popular, and the user scale is huge. A report found that in June 2010, nearly 93 million internet users visited Twitter.com, an increase of 109 % from the previous year1. The large volume of user’s social connection information captured by such systems provides many opportunities for mining social networks and phenomenon. In particular, in a micro-blogging system, users need to explicitly indicate whether they want to hear from another user by “following” or not. Knowing such explicit and directed relationships will allow not only structural analysis of social networks but also enables the study of information flow within the networks. Furthermore, networks in micro-blogging systems overlap heavily with social networks in real life. Many users are familiar with their followers 1

http://www.comscore.com/ger/Press_Events/Press_Releases/ 2010/8/Indonesia_Brazil_and_Venezuela_Lead_Global_Surge_in_ Twitter_Usage

A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 110–119, 2011. © Springer-Verlag Berlin Heidelberg 2011

Mining Social Relationships in Micro-blogging Systems

111

and followees not only in virtual world but also in real life [1], and new users generally join an existing network by accepting friends’ invitations. This makes social relationship analysis in micro-blogging systems more indicative of real social networks than analysis in other online communities. So far, most research of user relationships analysis studied blog communities, social network sites (SNS), and collaborative tagging systems [e.g., 2-4], and few studies micro-blogging systems. Existing methods used in these studies, such as Social Network Analysis (SNA), have limitations in analyzing huge volume of data sets, and most of them cannot incorporate information flow directions into the analysis. In this paper, we propose an integrate approach to mine social relationships in micro-blogging systems. In addition to high efficiency that is required to process huge data size, our approach also provides a tool for exploring the directions of information dissemination between users and evaluating individual users’ influence in information dissemination. The rest of this paper is organized as follows: The following section describes related studies and motivation. Section 3 introduces the new integrated approach. A case of implementation of our method is presented in Section 4. Section 5 shows our conclusion of the work.

2 Related Work 2.1 Analysis of Online Social Networks The popularity of online social network services makes social relationships mining come again into the limelight with the new communication platform. A lot of researches explore web user relationships in blog, SNS or other web communities from different perspectives with topological analysis [6], link analysis [7] and network evolution [8]. SNA is one of the most influential methods in social relationships analysis. Aligning centrality measures are adopted to discover communities in blogs [2], and betweeness measures are used to extract natural community structures of social networks by dividing the network nodes into densely connected subgroups [3]. Tyler et al. [9] proposed an algorithm based on betweenness and centrality to discover user groups in email networks. Previous studies show that most of SNA researches emphasized binary interaction data, with direct and/or weighted edges, and they focused almost exclusively on very small networks [10]. With the development of internet, especially the rapidly spread of micro-blogging systems, the limitations of SNA become more and more obvious. The size of data set is booming. SNA, however, is hard to process a huge data set. Besides, the traditional SNA is difficult to explore the directions of information transmission in micro-blogging systems. There are a few of other methods of online social relationships mining. To mine a directed social network from an online message board, Matsumura et al. [11] simplify the algorithms of Influence Diffusion Model (IDM) [12] in which the influence of a user is evaluated by propagating terms among couples via messages. Kazienko and Musiał [13] present a new method of personal importance analysis to discover

112

Q. Gao, Q. Qu, and X. Zhang

personal social features in the community of email users based on calculation of the strength of relationships between network members, its dynamic as well as personal position of the nearest neighbor. Similarly with SNA, these methods have difficulties in processing large volume of data and analyzing directed information flow. 2.2 Graph Theory Graphs are widely employed as general data structures in modeling complex systems and networks. Mining frequent subgraph patterns is an effective way to research characters of graphs. There are two basic approaches for pattern analysis: the aprioribased approach and the pattern-growth approach. Apriori-based approach begins with a small size frequent subgraph, and proceeds in a button-up manner by generating candidates with big “size” frequent subgraphs having an extra edge, vertex, or path [14]. The apriori-based approach is based on breadth first search, and they require a good-sized system working space. Different from the the apriori-based approach, the pattern-growth approach adopts depth first search and this approach expand ‘small size’ subgraphs by adding new vertexes or edges. Relational graph is a special graphic structure in which each vertex is unique. Relational graphs are frequently used for modeling biological network, social network, traffic analysis and internet analysis. Dense subgraph is a type of relational graphs, which usually denotes close relationship within a group. CluseCut and Splat are two main algorithms of frequent dense subgraphs mining [14]. Some new methods of analyzing huge directed networks are brought from graph theory. Samudrala [15] used graph theory to discover protein structures. The algorithm denotes each possible conformation of a residue in an amino acid sequence with the notion of a node in a graph. Each node gets a weight based on the degree of interaction between its side-chain atom and local main-chain atoms, and draw edges between pairs of residue conformations/nodes that are consistent with each other [15]. Cai et al. [16] proposed a regression-based graph matrix approach to explore hidden communities in heterogeneous social networks, and they validated the approach with the Iris and Digital Bibliography and Library Project (DBLP) datasets.

3 An Integrate Approach of Social Relationships Mining 3.1 A General Information Diffusion Model Garton et al. [17] suggested a regular social network can be described as a finite set of nodes that are linked with one or more edges, and we build an information diffusion model to explore the user network and information flow in micro-blogging systems. In the model, nodes indicate users or user groups, and directed edges represent information diffusion between two users. Based on this model and existing research related to mining of huge and directed graph data, a new integrate approach based on this model is introduced in the following.

M Mining Social Relationships in Micro-blogging Systems

113

F 1. Information diffusion model Fig.

3.2 Step 1: User Groupin ng by Information Dissemination Relationships Micro-blogging users spon ntaneously form a large number of groups by variious interests or different region ns. For social relationships mining, an ideal grouping is tthat information exchange with hin a group is frequently while information exchange w with members out of a group is as less as possible. Based on information diffusion model, we adopt maximal strongly y connected components (MSCC) to group users in miccroblogging systems. A MSCC in micro-blogging systems can be defined in this w way: Given a directed graph G = (V, E), where V is a finite set of nodes, and E (E V× ×V) is a finite set of directed edges. Nodes denote users, and directed edges exprress information flow. For ∀a V ∀b V if there is at least one path between a andd b, then the directed graph G is i a strongly connected component. And if G would noot a strongly connected compo onent when any node or edge were added to G, it iis a maximal strongly connected d component (MSCC). Based on the definition of o MSCC, we defined a user group as a set of nodes witthin which any two nodes can transfer t information bidirectionally. That is, any two ussers can’t transfer information bidirectionally except for that they belong to the saame MSCC.

∈， ∈，

∈

3.3 Step 2: Group Rankin ng by Information Dissemination Paths between Groups Obviously, the importance of groups in information dissemination is different. For this reason, we propose a method m to rank groups according to their contributionss in information dissemination. Each group is denoted as a a node, that is, the internal relationships within a grooup are masked and only the relationships r among groups are visible. The network oof a micro-blogging system is th hen condensed into a directed acyclic graph G´ (Detailss of proof are showed in Appeendix), and each node of G´is a MSCC. We adopteed a topological sorting algorith hm to rank groups in G´with necessary modifications. T The node without any informaation outflow is deleted from G´and put at the endd of the ranking list. This step iss repeated till all nodes are deleted. In the final ranking llist,

114

Q. Gao, Q. Qu, and X. Zhang

the first one is the strongest influential group in the micro-blogging system. The pseudocode of this algorithm is showed as follows: P Empty list that will contain sets of nodes in sequence N

Set of nodes with no outside link

Insert all nodes which have no outside link into N while N is non-empty do insert N into P for each node n in N remove n for each node m with a link e from n to m do remove e 3.4 Step 3: User Influence Estimation by the Probability of Information Dissemination In addition to ranking the influence of groups, knowing the strength of a specific user has on his/her followers within a group, especially in the most influential group, would be interesting from a practical point of view. Dijkstra [5] proposed a wellknown algorithm to explore the single-source shortest path problem for a directed graph, and the algorithm is often used for finding costs of the shortest paths from a node to another node. Inspired by Dijkstra's algorithm, we introduce the concept of width and propose a new index using the name of QIndex to estimate the influence of a certain user in a group. In micro-blogging systems, information transmits via subscriptions (by following others) or retweeting by others. We define the number of nodes from the source node to the target along a path as the distance of that path and define the number of different paths connecting the two nodes as the width. The probability that a piece of information reach the target depends on the length and the width of information transmission paths. The shorter the distance and the more paths existing between two nodes, the more probably information can reach the target node. We assume P as the probability that any user retweets a certain update. Therefore, the probability that the target user can receive this information is p= Ʃ i∈N P di, where N represents the set of all paths from the source to the target, and di is the length of path i. According to observation, it is reasonable to infer that P < 0.5. Therefore, the shortest path from the information source to the target node makes the greatest contribution to p. To simplify the problem, we can set a threshold T and if di is larger than T, pi which reflects the probability that information transmits via path i can be considered approximating to zero. Thus we only consider the shortest path within T. Based on the above inference, we propose an algorithm based on Dijkstra's to calculate the influence of a specific node. For a directed graph G = (V, E), V is a finite set of nodes and E is a finite set of edges and E V×V. The information source node is labeled as vs (vs V). The algorithm is described in following steps:

∈

∈

Mining Social Relationships in Micro-blogging Systems

115

1. Set the distance value to zero for initial node vs and to infinity for all other nodes, and assign a width value to one for vs and to zero for all other nodes. Mark all nodes unvisited, and set vs as the current node, noted as nc. 2. Then we need to calculate the distance and width between the current node nc and all unvisited nodes to which it links. Assume the distance from the source node to the current node is dc, and the width of the current node is wc, n' is an unvisited node which is linked by the current node. Before this round of calculation, the distance and width between the source node and n' is d' and w', and the distance between n' and the source node via nc is dc +1. If dc +1 < d' and dc +1 < T, then the distance between the source node and n' will be replaced with dc +1 and w' will be updated with wc at the same time; if dc +1= d' and dc +1 < T, w' will be changed into w' +1. 3. The current node nc will be marked as a visited node when all of its unvisited nodes are calculated. 4. Set the shortest distance node in all unvisited node as current node nc, and repeat step 2. If there isn’t any unvisited nodes in a distance less than T, QIndex of all visited nodes will be calculated and the algorithm will be finished. If width=0, the QIndex is infinity, otherwise, the formula is QIndex=Distance/Width

(1)

The higher the QIndex, the less probably the target node would receive information from the source node. It is very important to set a proper threshold T based on required accuracy, computing resources and other limitations. In the worst case, the running time of QIndex algorithm is O (ǀVǀ2 + ǀEǀ). Threshold T is proportional to time cost and the level of detail about the final result, that is, a larger threshold T costs more time and gets more detailed results, while the smaller T costs less time and gets more roughly results. If T approximates zero, the time cost of QIndex is close to O (ǀVǀ + ǀEǀ).

4 Validation User and usage data from www.digu.com, a popular micro-blogging system in China were collected. Digu.com was established in February, 2009, and the function and user interface design is very similar with twitter.com. We use snowball sampling method for data collection. Snowball sampling which is a nonprobability sampling method with which future subjects are recruited from acquaintances of exiting subjects, just like a rolling snowball. Twenty users were chosen randomly from the public discussion board of digu.com as ‘seeds’, and the information of user and user relationships were collected for each seed (Table 1). Then all the followers and followees of each seed were chosen as ‘seeds’ again. Our data collection system sends a request to digu.com every 10 seconds, and the whole data collection lasts two weeks (March 23, 2010 - April 7, 2010). Finally we collected 332,122 users and 11,160,822 inter-user connections. Because of computing resource constraints, a smaller sample includes 2,556 users with 35,510

116

Q. Gao, Q. Qu, and X. Zhang

inter-connections were used in further computation. There are 60 users’ relationships changed during data collection. However, the lost data only account for 2.34% of all users so that the effect is limited on the final result. Table 1. A data collection example

Item ID User name Nick name Location Gender Self-introduction Address Homepage Information Privacy The Number of Followees The Number of followers The Number of updates Folloee Follower

11528569 ququjoy Qu Beijing 1 (1-male, 2-female, 3-private) From Chongqing http://pic.minicloud.com.cn/file/default/SIGN_24x24.png http://digu.com/ququjoy false(false-information disclosure, true-information 2 2 7 digu, robot xabcdefg, flyinglin456

Fig. 1. MSCC of the biggest group

Through calculating MSCC, the biggest group which contains 1,426 users is found. Then we randomly choose a user named yoohee1221_to find out the top 5 users most influenced by him. We set T=5, and the result of QIndex calculation shows yoohee1221_ has strong influence on his direct followers, e.g., classyuan, which is not a surprising result. However, although some users don’t follow yoohee1221_ directly, e.g., liuxinwu, QIndex shows that they are as strongly influenced by yoohee1221_ as those directly linked to yoohee1221_ (Table 2.). Result shows using MSCC and QIndex to mine user relationships of micro-blogging systems is effective.

Mining Social Relationships in Micro-blogging Systems

117

Table 2. Result of Qindex

Users classyuan gambol liuxinwu xujun99663 dan123 chervun tuniu harliger zxb888 topidea yuanjuan WDM123 shaun

Distance 1 1 2 3 4 4 4 4 4 4 4 4 4

Width 1 1 2 2 2 2 2 2 2 2 2 2 2

QIndex 1 1 1 1.5 2 2 2 2 2 2 2 2 2

5 Conclusion In this paper, we propose a new approach to mine user relationships in microblogging systems. First, the new method of user grouping by maximal strongly connected components (MSCC) is introduced for social network structure analysis. Second, by condensing the graph and sorting with a topological algorithm, groups are ranked according to their influence on other groups. Third, a new algorithm that inspired by Dijkstra's algorithm is presented to assess the influence of individual users on others. The integrate approach can be applied to explore directions of information diffusion in micro-blogging systems and discover opinion leaders, and the results can serve various purposes, including product advertisement, policy advocacy, viral marketing and other information diffusion applications in micro-blogging systems.

References 1. Java, A., Song, X., Finin, T., Tseng, B.: Why We Twitter: Understanding Microblogging Usage and Communities. In: The Joint 9th WEBKDD and 1st SNA-KDD Workshop 2007, pp. 56–65. ACM, New York (2007) 2. Chin, A., Chignell, M.: A Social Hypertext Model for Finding Community in Blogs. In: 17th Conference on Hypertext and Hypermedia, pp. 11–22. ACM, New York (2006) 3. Newman, M.E.J., Girvan, M.: Finding and Evaluating Community Structure in Networks. J. Phys. Rev. 69(2), 26113 (2004) 4. Girvan, M.: Community Structure in Social and Biological Networks. PNAS 99(12), 7821–7826 (2002) 5. Dijkstra, E.W.: A Note on Two Problems in Connexion with Graphs. J. Num. Math. 1(1), 269–271 (1959)

118

Q. Gao, Q. Qu, and X. Zhang

6. Ahn, Y., Han, S., Kwak, H., Moon, S., Jeong, H.: Analysis of topological Characteristics of Huge Online Social Networking Services. In: 16th International Conference on World Wide Web, pp. 835–844. ACM, New York (2007) 7. Hsu, W.H., Lancaster, J., Paradesi, M.S.R., Weninger, T.: Structural Link Analysis from User Profiles and Friends Networks: A Feature Construction Approach. In: ICWSM 2007, pp. 75–80. ACM, New York (2007) 8. Golder, S., Wilkinson, D., Huberman, B.: Rhythms of Social Interaction: Messaging within a Massive Online Network. J. Com. and Tech. 2007, 41–66 (2007) 9. Tyler, J., Wilkinson, D., Huberman, B.: E-Mail as Spectroscopy: Automated Discovery of Community Structure within Organizations. J. Info. Soc. 21(2), 43–53 (2005) 10. McCallum, A., Wang, X., Corrada-Emmanuel, A.: Topic and Role Discovery in Social Networks with Experiments on Enron and Academic Email. J. Arti. Inte. Res. 30, 249–272 (2007) 11. Matsumura, N., Goldberg, D., Llorà, X.: Mining Directed Social Network from Message Board. In: 14th International Conference on World Wide Web, pp. 1092–1093. ACM, New York (2005) 12. Matsumura, N.: Topic Diffusion in a Community. In: Ohsawa, Y., McBurney, P. (eds.) Chance Discovery, pp. 84–97. Springer, Heidelberg (2003) 13. Kazienko, P., Musiał, K.: Mining Personal Social Features in the Community of Email Users. In: Geffert, V., Karhumäki, J., Bertoni, A., Preneel, B., Návrat, P., Bieliková, M. (eds.) SOFSEM 2008. LNCS, vol. 4910, pp. 708–719. Springer, Heidelberg (2008) 14. Han, J., Kamber, M.: Data Mining: Concepts and Techniques, 2nd edn. Morgan Kaufmann, San Francisco (2006) 15. Samudrala, R., Moult, J.: A Graph-theoretic Algorithm for Comparative Modeling of Protein Structure. J. Mol. Biol. 279(1), 287–302 (1998) 16. Cai, D., Shao, Z., He, X.F., Yan, X.F., Han, J.W.: Mining Hidden Community in Heterogeneous Social Networks. In: The 3rd International Workshop on Link Discovery, pp. 1–26. ACM, New York (2005) 17. Garton, L., Haythorntwaite, C., Wellman, B.: Studying Online Social Networks. JCMC 3(1), http://www.ascusc.org/jcmc/vol3/issue1/garton.html

Appendix Lemma 1. Given , and is a directed acyclic graph.

,

, where

is the condensation of G. G'

Proof. Suppose that there is a cycle c v, e in G and ∀p, q v. Then we get p and q are reachable from each other. Thus c is a strong connected subgraph of G , which is contradictory to the condition that G is the condensation of G. Therefore, G is a directed acyclic graph. Lemma 2. Given a directed acyclic graph any outside link from node n.

,

. If

, there will not be

Proof. Suppose the size of set V is L and nq has an outside link to nq. Ifn has not any outside link, the lemma will be proved; if nq has an outside link to nr, it is easy to prove that r≠p≠q for the reason that G is a directed acyclic graph. Continue this iteration until nr has no outside link, then the lemma will be proved. After L times

Mining Social Relationships in Micro-blogging Systems

119

iterations, nL will be get. We hypothesize there is an outside link from nL to ni. Because G is a directed acyclic graph, the inference that i≠1, …, L and the hypothesis that the size of V is L become contradictory. Therefore, there is at least one node in G which has no outside link.

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations Nirupama Dharmavaram Sreenivasan, Chei Sian Lee, and Dion Hoe-Lian Goh Division of Information Studies, Wee Kim Wee School of Communication and Information, Nanyang Technological University, 31 Nanyang Link, Singapore 637718 [email protected], {leecs,ashlgoh}@ntu.edu.sg

Abstract. Microblogs have received considerable attention in crisis communication due to its real-time nature. Despite this, research has not adequately addressed how users make use of information amongst the vast amount of real-time content available in a crisis. Hence, this study aims to understand information use in crisis situations by employing microblogs. Taylor’s [4] information use environment (IUE) model provides theoretical background for this study. According to this model, there are eight classes of information use. We focus on the IUE surrounding the 2010 Icelandic volcano eruption. Our study uses Twitter to analyze users’ postings related to this volcano. The types of postings were ascertained using content analysis. Our findings suggest enlightenment to be the largest category of information use. Other categories such as status messages, problem understanding and factual data were also reported. Further, humour not previously identified by Taylor [4] emerged as a substantial class of information use. Keywords: Microblogs, Twitter, real-time communication, information use, crisis communication.

1 Introduction Microblogs have become a popular Web 2.0 tool because of their valuable source of user-generated content. Recently, microblogs have received much attention in the field of crisis communication due to their real-time nature of communication [1, 2]. For instance, past research by Hughes and Palen [2] compares the behaviour of microblog users in mass convergence and emergency events from more general use. Study by Starbird, Palen, Hughes, and Vieweg [1] describes the mechanisms of information production and distribution on microblogs during a critical situation. However, owing to the sheer amount of real-time information potentially available in a crisis, it is often not known how users make use of such information. Moreover, the types of information requested or shared by users on microblogs during a crisis situation are also not known. Hence, understanding information use in crisis through the lens of microblog communication is an important research area, but to our knowledge, there are limited studies that have been conducted. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 120–129, 2011. © Springer-Verlag Berlin Heidelberg 2011

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations

121

Thus there is a need for scholarly research to assess information use in microblogs to investigate how users make use of the heterogeneous and distributed information available during crisis situations. The need arises because it is not known how users use the diverse information from microblogs to address their problems in a crisis situation. For example, when a volcano eruption occurs, people create many microblog postings related to the event. This enables users to detect and monitor the volcano occurrence in an up-to-date manner, to address their needs, by simply observing the posts. It is important to understand how users make use of the vast amounts of information available in crisis to make it meaningful for them to understand the scale and extent of the crisis. The link in the chain of such microblog communication in crisis, from real-time detection of events, to dissemination on this platform depends on understanding information use [3]. Therefore the objective of this study is to ascertain the role played by microblogs to support information use in crisis situations. In order to understand information use in a particular setting such as a crisis in microblogs, we use Taylor’s [4] information use environment (IUE) model in our study

2 Literature Review According to Taylor [4], information use is seen as a process where information sought or received by users is put to use by making sense of it so as to address their problems. Taylor [4] looked at the user and the uses of information, and the contexts within which those users deem information that will be useful to them at particular times. Taylor [4] associated information use to be an integral part of users’ everyday life where they construct and process information. Similarly, a context is a setting where users disseminate, seek and make use of information. This classical model proposed by Taylor [4] has been referred as information use environments (IUE). IUE places emphasis on the set of elements that affect the flow and use of information into, within, and out this context. IUE also determines the criteria by which the users judge the value of information. Taylor’s [4] model foreshadowed much of the current contextually-focused research by recognizing four influences on information use namely: individuals involved, constraints and opportunities provided by the context, problems faced and attitudes and approaches to problem resolution. This study makes use of Taylor [4] as a theoretical framework to develop a holistic understanding of the context in which users resort to information use in microblogs. A well-known IUE study by Davenport [5] also highlighted the importance of context in studying information use. However, we chose Taylor’s [4] seminal work as it is one of the pioneering studies done on IUE and has been incorporated in several contextual studies by Kuhlthau [6], Choo [7] and others. Taylor [4] emphasized IUEs to be the bridge between users and their environments whereby the context affects the availability, access to, and use of information. Additionally, Taylor [4] identified eight classes of information use by individuals to resolve their problems: enlightenment, problem understanding, instrumental, factual, confirmational, projective, motivational and personal. This paper focuses on these classes of information use to categorize the types of information used during a crisis situation. Our study describes a crisis as an

122

N.D. Sreenivasan, C.S. Lee, and D.H.-L. Goh

IUE to understand this context within which the affected users make choices about what information will be useful to them. Crisis may be seen as an interruption in the economic and social lives of individuals in the form of various ranges of situations [8]. For example, crises can be of various types such as natural disasters, human-made crises including terrorism attacks, epidemics such as spread of H1N1, nuclear crisis, environmental crises such as oil spill and many more. As crises present highly complex information environments, management of information before, during and after such emergency situations becomes important to influence the way in which such situations can be managed [9]. Crises events in the past have also demonstrated the need to understand how individuals, government agencies, public relation professionals and other nongovernment agencies collect, organize, share, manage and disseminate information during such situations [1]. From an information perspective, crisis situations are demanding owing to the various information challenges that can surface. Some of the problems that individuals face during crises include information overload, lack of relevant information and changing information needs at different stages of a disaster [8]. Other information challenges include integration and coordination of the information from diverse actors involved, information dissemination at the right time via multiple channels and evaluation of the trust in the information sources [8]. Given this background, our study aims to investigate the types of information potentially used by users in crisis on microblogs. As much as its ability to allow fast exchange of information between users, this platform also differs from other social media because of its ability to broadcast concise real-time updates that can be easily accessed using text messaging via mobile communication devices, Web and other third-party applications. With this ability to send messages using mobile devices and also be able to easily broadcast them to a wider audience, they seem to be fit for use during crisis events. Recently, microblogs have received much attention in the field of crisis communication owing to its growing ubiquity, ability to rapidly communicate and cross-platform accessibility [1, 3]. Despite the fact that these studies have suggested that microblogs are effective in handling emergency communications, limited work has been done to understand how users make use of the vast amounts of real-time information on microblogs during crisis to support their needs. Identifying users’ information usage on microblogs in crisis is important for several reasons. For example, microblogs users may turn to this medium in crisis to learn about real-time events, seek information, share information or communicate with affected agencies [9]. Owing to the vast amount of information available in crisis, users often have difficulty in the early phases to search and make use of relevant information to solve their problems. Even when they begin with a direction in mind, many may become confused and uncertain about how to proceed after a short period of time [1]. Hence, this research gap motivates our study. Our goal is then to examine whether microblogs can facilitate information use. We do this by uncovering categories of information use on microblogs during a crisis situation. Our study shares similarities with prior work on microblogs on event-based user behaviour [2] and usefulness of microblogs in emergencies [9], among others, in that we aim to investigate how users adopt microblogs to communicate with each other during crisis. Nevertheless, there are distinct differences between this study and

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations

123

past work that warrant the current research. For example, the study by Hughes and Palen [2] compared features of user behaviour in Twitter during major events .In contrast, our study investigated microblog postings in a crisis situation to understand the potential uses of information. Next, Mills, Chen, Lee, and Rao [9] examined the usefulness of Twitter as a medium for emergency response. On the other hand, our study drew from Taylor’s [4] classes of information use to examine the categories of potential uses in microblogs during a crisis. With this stimulus, our study aims to explore the following research question: RQ: How do users employ microblogs to support information use in a crisis?

3 Data Collection To investigate the research question, Twitter was chosen because it provides a realtime perspective of what people are thinking and doing. According to recent statistics from comScore, Twitter has grown from 37.3 million unique visitors in May 2009 to 90.2 million in May 2010 [10]. Twitter was launched on July 13, 2006 and allows users to post tweets or updates to a network of other users. Tweets are text-based, up to 140 characters in length and can be posted via multiple means such as Twitter.com, text messaging, instant messaging, mobile devices or from other third party clients. Users can indicate if their tweets are public or private. By being public, tweets would appear in reverse chronological order on the public timeline on Twitter’s home page or on the user’s Twitter page where all the updates are aggregated into a single list. Users can follow and read each other’s tweets without giving mutual permission. By being private, tweets are visible only to a user’s followers. Followers are a class of users who receive each other’s updates on their home page. Our paper reports on the use of Twitter during the 2010 Icelandic volcanic ash eruption that forms the IUE for our study. This eruption that began late March 2010 caused a large scale disruption for air travel across several airports in Europe. Due to the volcanic ash, many airports across Europe were shut down and thousands of travellers all over the world experienced flight delays and cancellations. This large scale crisis affected the airline industry, by costing them approximately 130 million pounds a day [11]. Based on this crisis, we characterize the categories of information use during and following the crisis. We chose the IUE surrounding the volcano in Iceland for analysis, as it provided an adequate coverage of tweets that was required for examining information use closely in a critical situation. In our study, tweets are defined as posts made by any Twitter user. The collection dates for user tweets posted during the crisis were from April 13, 2010 to May 5, 2010. The data collection period was determined based on the timeline when the Icelandic volcanic ash eruption was at its peak [12]. Queries were submitted to the Twitter search facility and the search terms used corresponded to name of the Icelandic volcano ‘Eyjafjallajokull’, ‘volcanic ash’ and hashtag ‘#icelandvolcano‘. We used hashtags (i.e. ‘#’) to search for user tweets that relate to specific events during this crisis. For the entire data collection period, we harvested a total of 63435 tweets. From this, we sampled 15% of tweets per day over the entire time period after eliminating duplicates and non-English tweets, resulting in 9641 tweets.

124

N.D. Sreenivasan, C.S. Lee, and D.H.-L. Goh

4 Findings and Analysis 4.1 Coding Scheme This study adopted qualitative content analysis as it uses a naturalistic paradigm to interpret meaning from text data content [13]. Hsieh and Shannon [14] defined this research method where content of text data was subjectively interpreted through a systematic process of identifying themes or patterns. This qualitative content analysis approach is appropriate for our study to inspect the content of the postings to examine the various types of information that users used during crisis situations. To enable this, coding schemes were developed to facilitate content analysis for the user tweet content. We adopted the coding strategy from Taylor’s [4] eight classes of information use. After this, we used the entire sample of tweets (N=9641) to check if Taylor’s [4] categories of information use could be coded to the data. An iterative coding process was used to consolidate and finalize these codes either through combining or omitting existing codes that were adapted from Taylor [4] or developing additional codes if any. Each category then became a label for assigning meaning to the descriptive or inferential information found in the tweet content to categorize them based on Taylor’s [4] eight classes of information use. For example, tweets from users providing contextual information to better understand the situation were categorized as Enlightenment while tweet content providing a detailed explanation of specific problems were categorized as Problem understanding. If the tweet content did not correspond to any of the eight information use categories of Taylor [4], then additional categories were formulated. After finalizing the categories, all tweets in our sample were coded. Each tweet could be coded into more than one content category. The details of all categories used in the coding schemes for tweets are shown in Table 1. The coding consistency was measured through an assessment of inter-coder agreement to ensure that all coding issues related to the definitions of categories were discussed and resolved. Two coders were used in the study. Both were graduate students who were familiar with microblogging. An inter-rater reliability analysis using Krippendorf’s alpha was performed to determine the agreement among the two coders. The software package, SPSS was used to compute this statistic. For the entire tweet sample, the inter-rater reliability for the coders was found to have Krippendorf’s alpha value of 0.82, indicating strong agreement [15]. 4.2 Results In terms of supporting information use in crisis among users, our results in Table 1 show that microblog communication was mainly used for enlightenment, to provide personal updates, for problem understanding and reporting of factual data. In particular, our analysis indicated that microblogs was used primarily to understand the contextual information of the critical situation (23.5 % of the user tweets) so as to make sense of it. Our findings showed higher percentages for this category as users were perhaps making use of the background information to get an idea of the scale of crisis. For example, a user tweet was used to provide background information on the rise in travel costs due to volcanic eruption in Iceland (“News post: Volcano air,

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations Table 1. IUE categories N=9641 Category Enlightenment

Personal

Problem understanding

Factual

Motivational

Projective

Instrumental

Confirmational

Description

Example

Contextual information to Flight disruptions in better understand the situation Europe get even worse: Thick drifts of volcanic ash blanketed…. http://bit.ly/8YuCz2 Status messages of users Woke up to the news that there was an earthquake in China & a volcano in Iceland. Explanation of particular Iceland farmers try problems to save herds from ash: http://bit.ly/93Cr8N #msnbc Precise data (e.g. numbers, 800 flee as Iceland figures) volcano rumbles http://bit.ly/blzK2p Different perspectives on the Iceland Volcano same problem Was God's Response To British Ad Ban http://tinyurl.com/2c 8kj7s Predicting or forecasting Iceland volcano crisis will hit European airlines worse than U.S., experts say. http://bit.ly/bUWum V Providing instructions on how Stranded by a to handle certain situations volcano? Handy survival guide! http://bit.ly/aug00k Information verification @andrewqh Is that the name of the other glacier volcano in Iceland?

Non-IUE categories Humour

Expressing humour

Irrelevant

Spam

% 23.5

14.7

12.2

11.6

5.3

3.7

3.5

3.1

Iceland, good luck 8.1 and please turn off your volcano before I fly to Europe in two weeks! Silly Kissers…. 2.6

125

126

N.D. Sreenivasan, C.S. Lee, and D.H.-L. Goh

tourism costs may top $3.5 billion in Europe - The Age http://cli.gs/qYs6j”). Next, microblogs were used to update their status messages in the form of what they were facing currently in this crisis. One example, was a user tweeting about his disrupted flight journey during the volcano eruption (“Ready for an 11 hrs flight over that nasty volcano, hope it will let me come back home when I`m supposed to”). Hence our observation on the top two frequently occurring categories supports the notion that microblogs are used for disseminating critical updates and first-person accounts of the disasters [9]. The third highest category of information use was related to in-depth understanding of the problems pertaining to the current situation. For example, a user tweet highlights the problems that stranded travellers are facing due to volcano (“Airlines appeal to passengers to give up seats to travellers stranded by the volcano ash crisis http://is.gd/bGH63”). This user tweet is categorized as problem understanding as it provides a better comprehension of specific problems such as flight delays in the event of air travel disruption because of which airlines are rebooking the tickets of stranded travellers. The fourth category of information use was related to reporting of factual data. One example was a user tweet that reported factual volcano data in the form of numbers (“Eyjafjallajökull: activity still the same - plume height 3-6km, wind now blowing ash to the NW, Strombolian explosions continue #volcano”). Additionally, we found that other categories of information use related to different perspectives on the same problem (5.3%), forecasts (3.7%), instructions to handle certain situations (3.5%) and information verification (3.1%). These showed relatively lower percentages. This was perhaps because a majority of the users were broadcasting real-time information related to the crisis situation instead of concentrating on specific topics. It was also found that humour (8.1%) was a substantial category of information use that was not identified by [4] as shown in Table 1.

5 Discussion On examining the information use categories in user tweets, our results indicate that the main themes of information use could be grouped into topics such as sharing contextual information, providing detailed information on problems, reporting of daily routines and facts. These results support the categorization adopted by past studies that have examined the nature of microblog information during crisis [1, 2]. We found that these categories from our findings were consistent with the broadcastbased information sharing categories outlined previously by Hughes and Palen [2]. Categories such as personal updates were similar to the tweets corresponding to firsthand information synthesis identified by [1]. At the same time, our study also highlighted that users not only share useful updates about the crisis situation but also resort to humour to reduce their stress levels. On a different note, fewer tweets were found related to categories such as asking questions for the purposes of information verification (3.1%) and instructions to handle certain situations (3.5%). This may be attributed to the high information demands of crisis situations. Owing to the rise in demand, users may broadcast real-

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations

127

time information to increase their awareness about the situation rather than directing specific questions [2]. Another reason could be that the tweets that contain background information related to the critical situation may also contain instructions and information verification questions in the form of pointers to other sources. For example, user tweets such as “AccuWeather.com - Weather Video - Volcanic Ashes Disrupt Travel: Should travel insurers pay up after volcano disruption?... http://bit.ly/aQC6dD” suggest that the URL contained in the tweet encompasses the contextual information along with the questions pertaining to a specific topic such as travel insurance. Next, our findings on the information use categories reported that users expressed different perspectives on the same problems in their tweets (5.3%). These findings suggest that users looked at the problems that they were facing because of crisis in different ways. For example, user tweets such as as “News Feature: What's God trying to tell us with the volcano in Iceland... http://bit.ly/bko3Ao” indicate that users were looking for spiritual support to manage the volcanic crisis. Another interesting finding from our study showed that humour not identified by [4] was found to be a substantial class of information use (8.1%). The appearance of humour as an emotional manifestation could be attributed to the way in which people vent their stress or frustration during crisis [16]. For example, humour was used to relieve stress while facing problems of being stranded in airports for prolonged periods owing to the volcanic eruption. User tweets such as “Its earth day today...well earth already celebrated its day with fireworks we'd seen in Iceland's volcanic eruption! - stranded at heathrow” indicate the emergence of humour. This finding highlights the importance of microblogs in crisis by suggesting that emotions are used as a coping mechanism to overcome problems.

6 Implications, Limitations, and Future Research Arising from our findings, the following implications for research and practice may be derived. First, this study contributes to our improved understanding of crisis as an IUE. Our findings indicate that crisis influences the way different users make use of real-time information found on microblogs. For example, our findings demonstrate that many users share contextual information about the crisis to understand the situation and be well-informed. At the same time, other users reported on their firsthand experiences of the critical situation by providing insight and varying perspectives to the same problems. These findings imply that users want to take initiative in providing different types of information about the crisis situation to raise awareness and at the same time assist the affected individuals and agencies. If they do not have first-hand knowledge of the critical situation, they rely on other media sources by providing links to pass on information on microblogs. Second, our study contributes to the extant literature on crisis informatics by exploring the role microblogs play to facilitate emergency communication. Our study sought to examine the types of information that users make use of in crisis situations. Moreover, our findings on the emergence of humour as a substantial class of information use during crisis imply that users circulate light-hearted chatter amongst each other to spread cheer and hope. For example, by sharing jokes on Icelandic

128

N.D. Sreenivasan, C.S. Lee, and D.H.-L. Goh

volcano eruption in their tweets, users not only use humour to help each other cope with the crisis but also instill optimism to overcome from the after-effects. This finding implies that in times of crisis, users are hopeful that by being optimistic and cheerful, they can overcome their problems better in crisis. Next, our findings on how users make use of information from the vast amount of real-time information available in crisis have implications for crisis mitigation and risk communication agencies. They can track these postings to take steps to manage the situation and supplement with required information if needed. Doing so may help them improve their overall process of emergency warning, response and recovery procedures. Another implication that can be derived from our findings is that users make use of real-time information on microblogs extensively during crisis. This implies that under such critical situations, it is possible that false and inaccurate information gets circulated that may prove devastating to affected individuals. Our study suggests that crisis mitigation agencies should moderate the microblog postings or set up official microblog accounts during a crisis situation to disseminate accurate and up-to-date information. However, our study has certain limitations that can be addressed in future research. First, we examined only one microblogging site, Twitter, for our analysis. Users of other microblogging services might differ in their information use that may require separate investigations. Second, only one crisis situation was used in our analysis, although a major one. Using more than one crisis situation for investigation, would allow better generalizability of our results. In addition, we only captured Twitter posts from April 13, 2010 to May 5, 2010. It is possible that tweets in a different time period such as the day the crisis was declared over, may yield other varied and useful results not uncovered in this study. However, this limitation does not affect the accuracy of the data used in our study as we chose the timeframe where the volcanic ash eruption was at its peak ensuring an adequate number of tweets for analysis. Finally, future studies could extend our work by examining information use on a different crisis situation or an event to compare and contrast our results. Future research can also focus on exploring the categories of information use that are most important to satisfy user needs during a crisis situation. Our analysis of user-generated content from microblogs can be supplemented with a mixed methods approach using ethnographic studies, focus groups and interviews with Twitter and non-Twitter users to further understand the potential of this new medium for crisis communications. Acknowledgements. This work was supported by Nanyang Technological University Academic Research Fund Program (Tier 1), under research Grant No. 59/09.

References 1. Starbird, K., Palen, L., Hughes, A.L., Vieweg, S.: Chatter on The Red: What hazards threat reveals about the social life of microblogged information. In: Proceedings of the 2010 ACM Conference on Computer Supported Cooperative Work, pp. 241–250. ACM, New York (2010) 2. Hughes, A.L., Palen, L.: Twitter adoption and use in mass convergence and emergency events. International Journal of Emergency Management 6(3/4), 248–260 (2009)

Tweet Me Home: Exploring Information Use on Twitter in Crisis Situations

129

3. Li, J., Rao, H.R.: Twitter as a rapid response news service: An exploration in the context of the 2008 China Earthquake. The Electronic Journal on Information Systems in Developing Countries 42(4), 1–22 (2010) 4. Taylor, R.S.: Information use environments. In: Dervin, B., Voigt, M.J. (eds.) Progress in communication sciences X, pp. 217–255. Ablex Publishing Corp., Norwood (1991) 5. Davenport, T.H.: Information ecology: Mastering the information and knowledge environment. Oxford University Press, New York (1997) 6. Kuhlthau, C.C.: Seeking meaning, 2nd edn. Libraries Unlimited, Westport (2003) 7. Choo, C.W.: The knowing organization: how organizations use information to construct meaning, create knowledge, and make decisions, 2nd edn. Oxford University Press, New York (2006) 8. Coombs, W.T.: Ongoing Crisis Communication: Planning, Managing, and Responding, 2nd edn. Thousand Oaks, CA (2007) 9. Mills, A., Chen, R., Lee, J., Rao, H.R.: WEB 2.0 emergency applications: How useful can Twitter be for emergency. Journal of Information Privacy & Security 5(3), 3–26 (2009) 10. Latest comScore stats show Twitter growth is still strong, http://techcrunch.com/2010/06/18/latest-comscore-stats-showtwitter-growth-is-still-strong/ 11. Ash cloud costing airlines £130m a day, http://www.guardian.co.uk/business/2010/apr/16/icelandvolcano-airline-industry-iata 12. Met Office, http://www.metoffice.gov.uk/corporate/pressoffice/2010/ volcano.html 13. Cavanagh, S.: Content analysis: concepts, methods and applications. Nurse Researcher 4(3), 5–16 (1997) 14. Hsieh, H.F., Shannon, S.E.: Three Approaches to Qualitative Content Analysis. Qualitative Health Research 15(9), 1277–1288 (2005) 15. Krippendorf, K.: Content Analysis. Thousand Oaks, CA (2004) 16. Kendra, J.M., Wachtendorf, T.: Reconsidering Convergence and Converger Legitimacy in Response to the World Trade Center Disaster. Research in Social Problems and Public Policy 11, 97–122 (2003)

Impact of Blog Design Features on Blogging Satisfaction: An Impression Management Perspective Wee-Kek Tan and Hock-Hai Teo Dept. of Information Systems, National University of Singapore, 21 Lower Kent Ridge Road, Singapore 119077, Republic of Singapore {tanwk,teohh}@comp.nus.edu.sg

Abstract. Trends observed among bloggers worldwide suggest that selfpresentation is a main motivation to blog and a greater majority of bloggers do so with the aid of third party hosting services and tools. Thus, providing tools that are designed to meet the usability requirement of bloggers with respect to self-presentation may increase blogging satisfaction. Grounded on the theory of impression management, we propose that a blogger actively engages in impression management with content, functional and aesthetic blog design features. This can lead the blogger to obtain a perceived confirmation from the readers of his or her identity, i.e., perceived identity verification. Perceived identity verification may heighten blogging satisfaction since it helps the blogger to achieve the goal of self-presentation. Results obtained from an online survey, which was preceded by a focus group discussion, provide support for our conjecture. Theoretical and practical implications are also discussed. Keywords: Blog, blog design features, satisfaction, self-presentation, theory of impression management.

1 Introduction Blogging is a highly popular online activity. According to survey statistics from the State of the Blogosphere 2009 [1], the majority of bloggers on the Internet are casual bloggers who seek to express their life, and to share their expertise and experience. Beyond personal motivations, there is also a growing trend in using blog to make money and to exert political influence. These wide-ranging impacts are being wielded by the majority of bloggers using third party hosting services and tools [1]. It is therefore important to develop an in-depth understanding of how blog tools may be designed to meet the usability requirements of bloggers [2]. Since the majority of bloggers use their blog to express their life and personality in the form of online journal [3], i.e. self-presentation, it makes sense to provide design features that support this main motivation. To this extent, functional features to aid in blogging and the ability to customize the blog’s aesthetic elements are two key factors affecting bloggers’ choice of third party tools [1]. Moreover, most bloggers consider personal satisfaction as the key success measure of their blogs [1]. This research thus A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 130–139, 2011. © Springer-Verlag Berlin Heidelberg 2011

Impact of Blog Design Features on Blogging Satisfaction

131

examines how content, functional and aesthetic design features of blog, which aid the blogger in achieving self-presentation, may affect blogging satisfaction. The findings from this research can potentially help researchers and practitioners to design effective blogging tools. Besides blog hosting service providers such as Blogger, social networking sites such as Friendster that incorporate blogging as a pertinent feature may also benefit from this research.

2 Theoretical Background Bloggers typically express their identity on their blogs in the form of personal information such as age and location, emoticons, language styles, occupations, and hobbies. Over a period of time, it is possible for a blogger to develop virtual identity [4]. One fundamental theoretical notion underpinning this process is that individuals proactively present their identity to others using props as explained by the theory of impression management. 2.1 Theory of Impression Management The impressions that individuals leave on other people typically affect how they are perceived, evaluated and treated [5]. Consequently, individuals are inclined to control the impressions that others form of them, i.e., impression management. According to the theory of impression management, this process draws a metaphoric similarity with theatrical performance [6]. In a performance that takes place in a public setting, the actor prepares the required props and scenery in the back-stage before performing with the props and scenery on-stage. In a similar fashion, an individual makes a conspicuous effort to acquire knowledge, skill and props so that s/he can use them to project certain impression upon other people. 2.2 Impression Management on the Web Researchers have found that personal home pages are being used by their authors to publish information on topics deemed important to them together with their social contexts [7]. To this extent, personal home page is an inclusive self-presentation tool that explicates the identity of its author [8]. Indeed, the author of a personal website makes a conspicuous effort to adorn his or her web pages with identity claims that present a coherent online identity [9]. The identity claims could include textual and non-textual contents reflecting the interests, preferences and values of the author, i.e., the props [6]. For instance, aesthetic elements (e.g., image, video and sound) and functional elements (e.g., hyperlink to website related to the author’s interest). 2.3 Blog Design Features for Impression Management Within the context of blog, self-presentation may be applied by bloggers through the creation or usage of blog template and manipulation of the underlying Hypertext Markup Language (HTML) code [10]. Our present study, however, goes beyond the use of basic website elements such as image and hyperlink to examine how

132

W.-K. Tan and H.-H. Teo

sophisticated functional and aesthetic blog design features provided by third party service providers can be used for self-presentation. We shall term this as impression management with blog design features. Functional features such as blogger profile and blogroll are required for bloggers to share their life and experiences with readers. Aesthetic features such as blog header and background enhance the visual appeal of the blog and sustain readers’ interest. In addition, the most important characteristic of blog is the posting of journal entries in a reverse chronological order by the author. Most third party blog tools aim to automate this process in the simplest way possible [2]. For this reason, we consider the content feature as another important dimension of blog design. Content feature aids the blogger in posting both textual and non-textual materials that reflect his or her identity over and above what the functional features can accomplish. In gist, content, functional and aesthetic design features collectively make up the props required by the blogger to present his or her identity to the readers. Fig. 1 depicts sample screenshots of the three categories of blog design features currently available on third party blog service providers.

Fig. 1. Screenshots of blog design features

2.4 Presentational Expectancies Although the theory of impression management explains how bloggers could use the various blog design features as props to present their identities, there is a non-trivial caveat that needs to be taken into consideration. Researchers have noted that social anxiety could develop if an individual is motivated to make a desired impression to a target audience but is hesitant that s/he has the capability to do so [11]. That is, an individual’s self-presentation behavior can be affected by presentational outcome expectancy and presentational self-efficacy expectancy [11, 12]. The former refers to the belief that projecting certain impression will lead to the desired goal and the latter refers to the belief on one’s ability to present that impression.

Impact of Blog Design Features on Blogging Satisfaction

133

3 Research Model and Hypotheses Development Drawing on the theory of impression management as the theoretical lens, we put forth our research model (see Fig. 2) to understand how the provision of various blog design features may fulfill the usability requirement of bloggers in term of impression management.

Fig. 2. Research model

3.1 Impression Management Process and Its Impacts In a technology-mediated setting, tools that facilitate rich social interaction can aid in achieving shared understanding among the communicating parties [13], which is helpful for identity communication. In particular, personal website has often been heralded as a new form of identity communication [14]. Prior research has suggested that it is possible for personal website to communicate nonverbal cues through the use of hyperlinks, emotions, animations and other technological conventions [15]. Collectively, these website elements form the props for individuals to present themselves to the public using the personal websites as the performing stage [6]. In a similar fashion, bloggers are known to present themselves through their blogs [10]. A blogger actively using the various content, functional and aesthetic design features can put across his or her desired impression in a consistent and focused manner to readers of their blog. This can give the blogger a perceived confirmation from the readers of his or her identity, i.e., perceived identity verification [16]. We thus hypothesize that: H1: Impression management with blog design features is positively related to perceived identity verification. In an online community, members whose identities are verified and recognized by others will feel that they are better understood and will be treated in the desired way [16]. Thus, communications among members should have lesser misunderstandings and conflicts. Furthermore, members will be inclined towards increasing their attraction to other community members and commitment to the online community [17]. In a real world setting, identity verification can lead to a satisfying interpersonal relationship because of the resulting shared understanding forged through the verification process [18].

134

W.-K. Tan and H.-H. Teo

In sum, identity verification should lead to contentment with an individual access to community resources, i.e., satisfaction [16]. In our context, we define satisfaction as whether the blogger is contented with the blogging experience to express one’s life and to share one’s expertise and experience with the readers. Accordingly, we hypothesize that: H2: Perceived identity verification is positively related to satisfaction. We also propose a second direct causal pathway between impression management and perceived identity verification. In an online community, identity communication can often help members to connect to others who share similar interests, attitudes and experiences [19]. People with similar personality are more likely to communicate and build relationships [20]. Impression management aids the blogger to present their identity to the readers and thus attracts those who share similar identity. It thus aid bloggers to achieve their goals of expressing themselves and share their experiences and expertise with similar others. Indeed, it is possible for bloggers and their readers to form a community of similar interest [3]. Satisfaction may be viewed as a positive affective arousal on the part of the blogger towards his or her blogging experience [21]. It may be induced in the presence of an advancement factor that brings the blogger closer to the attainment of the self-presentation goal [22]. We posit that impression management with the aid of content, functional and aesthetic design features constitute this advancement factor. We thus hypothesize that: H3: Impression management with blog design features is positively related to satisfaction. 3.2 Antecedents of the Impression Management Process Outcome and self-efficacy expectancies have been widely studied in the extant literature on individual’s use of information technology (e.g., [23]). Both computer outcome and self-efficacy expectancies are known to reduce computer-related anxiety and increase computer usage [23]. Within the online context, Internet self-efficacy expectancy has also been found to enhance individuals’ perception of the ease of use of Internet technology and consequently higher behavioral intention and actual usage behavior [24]. In this study, we adapt the original conceptualization of presentational outcome and self-efficacy expectancies [12, 25] to the context of the blogger using blog design features to manage one’s impression. Presumably, high degree of presentational expectancies should reduce a blogger’s anxiety towards the use of the various blog design features to manage his or her impression and consequently increase their propensity to do so [12]. Accordingly, we put forth our final two hypotheses: H4: Presentational outcome expectancy is positively related to impression management with blog design features. H5: Presentational self-efficacy expectancy is positively related to impression management with blog design features.

Impact of Blog Design Features on Blogging Satisfaction

135

4 Research Methodology 4.1 Research Design This study primarily adopted a quantitative survey design to empirically validate the research model depicted in Fig. 2. Specifically, an online survey technique was used for two reasons: 1) an online survey would have been more convenient for bloggers to participate; and 2) online survey allowed us to administer visual stimuli for the various blog design features to enhance the questionnaire. In addition, we conducted an initial focus group discussion with a group of experienced local bloggers to help us gather insights on two issues. First, we wanted to determine whether the majority of local bloggers do indeed blog to express their life, and share their expertise and experience, as reported by the State of the Blogosphere 2009 [1]. Second, we wanted to determine the blog design features that the bloggers would most likely use for impression management. This was to aid us in preparing a representative set of features for the visual stimuli of the survey questionnaire. 4.2 Focus Group The focus group was conducted with ten bloggers over two sessions of five bloggers each. They were recruited through an online invitation posted on the website of a popular online community for local bloggers. The proceedings of the discussion were audio recorded and transcribed subsequently. In the first part, the participants were asked to share their motivations for blogging and the types of contents posted on their blog. Consistent with the findings of the State of the Blogosphere 2009, six out of ten participants explicitly mentioned selfpresentation as well as experience and expertise sharing as the main motivation for blogging: “I use my blog as a diary that is opened to others, especially my friends and family…; I write just about anything that happened to me on my blog…” Three other participants indirectly hinted at self-presentation as their blogging motivation: “I mostly gossip about what happened in the office…; I give my takes on current affairs, well sometimes I grumbled about bread and butter issues…” In the second part, we explained to the participants the purpose of the focus group and also gave them an overview of content, functional and aesthetic blog design features. We then asked the participants to discuss how these features may be used for self-presentation. Most participants highlighted the content feature as the most useful and obvious one in expressing their identity: “I feel that the content of the blog is usually detailed enough to show my identity…” Aesthetic design features were also given the thumbs up: “I make extensive changes to my blog template to reflect my favorite color, hobby and idols…” Functional design features drew mixed responses. Most participants, however, pointed out that certain features were useful: “I placed many links to other blogs and websites associated with my idols… I guess that counts as some form of social affiliation…; the short blogger profile says everything about me…” Towards the end, the participants were asked to rank the various features in descending order of suitability for self-presentation. Various functional and aesthetic design features were identified from the extant literature (e.g., [26]) and given to the participants. Those features with a ranking score greater than the mean were used in the subsequent survey questionnaire as visual stimuli (see Table 1).

136

W.-K. Tan and H.-H. Teo Table 1. Selected functional and aesthetic design features used in the online survey

Functional Blogger Profile Photo Albums Tagboard Links to Non-blog Websites Blogroll Badge

Mean Std. Dev. Aesthetic 18.0 3.300 Post Content 16.7 3.129 Blog Header 14.9 5.971 Content Area 14.8 2.860 Post Title 14.6 3.627 Blog Background 13.0 5.477 Primary Navigation

Mean Std. Dev. 10.7 1.947 9.9 2.514 9.0 1.700 8.5 1.900 7.4 2.989 6.1 3.381

4.3 Operationalization of Constructs The theoretical constructs shown in the research model (see Fig. 2) were measured with instrument scales either adapted from validated scales or self-developed based on established definitions from the extant literature. Except for satisfaction, which is a seven-point semantic differential scale, the remaining constructs used seven-point Likert scales. The scale measuring content impression was adapted from [27]. The scales for functional impression and aesthetic impression were self-developed based on the content impression scale. Presentational outcome expectancy and presentational selfefficacy expectancy were measured using scales adapted from [12] and [29]. Satisfaction was measured with a scale adapted from [28]. For the scale measuring perceived identity verification, we followed Ma and Agarwal’s prescription [16]. 4.4 Survey Design and Procedure The online survey was conducted over a one month period and participants were recruited via an open invitation posted on the same online community. All participants were checked to ensure that they had an existing personal blog that was actively maintained for at least the past 12 months leading up to the survey. The survey proceeded according to the following sequence. Participants first registered their particular and also undergone a simple validation of their blog. They then read the survey instructions before proceeding to read the description of each

Fig. 3. Screenshots taken from actual survey website

Impact of Blog Design Features on Blogging Satisfaction

137

category of blog design features and view the associated visual stimuli. The visual stimuli were spread over three web pages, one per category of features. For content, the participants were shown a sample screenshot of a blog post editor taken from Blogger together with the preview of the post. For the functional and aesthetic features, the participants were shown a skeleton layout of a blog with the various features marked up (see Fig. 3). Participants must view all features in each category before proceeding to the next category. The survey website ensured that they spent at least five minutes on each category to adequately understand the various features. The participants then proceeded to complete the main survey and a short demographic questionnaire. Most of the participants took less than 30 minutes to complete the entire survey.

5 Results 5.1 Descriptive Statistics There were altogether 261 participants who registered. After removing the invalid responses, the final useable sample size was 244. There were 113 male and 131 female participants. The majority of the participants were aged from 21-30 (67.2%) and were locals (82.4%). On the average, the participants had 3.0 years of blogging experience (δ = 2.232) and 66.4% of them blog on a daily or weekly basis. 5.2 Validity and Reliability of Instrument Scales The data were analyzed with partial least square (PLS) based structural equation modeling. The internal consistency, convergent validity and discriminant validity of the measurement model were assessed to be acceptable. The instrument scales were also deemed to be reliable after assessing their Cronbach Alpha scores. 5.3 Hypotheses Testing We then proceeded to test the structural model with the bootstrapping procedure. The standardized path coefficient, standard error and statistical significance are shown in Table 2. The R2 for perceived identity verification is 0.203, satisfaction is 0.237, content impression is 0.381, functional impression is 0.227 and aesthetic impression is 0.233. Table 2. Summary of hypotheses testing results Hypothesis Path Std. Error Sig. H1 CI → PIV 0.253 0.068 *** FI → PIV 0.067 0.077 n.s. AI → PIV 0.231 0.074 *** H2 PIV → SAT 0.202 0.066 ** H3 CI → SAT 0.243 0.068 *** FI → SAT 0.169 0.067 ** AI → SAT 0.026 0.072 n.s.

Hypothesis Path Std. Error Sig. H4 POE → CI 0.257 0.070 *** POE → FI 0.247 0.071 *** POE → AI 0.230 0.081 ** H5 PSEE → CI 0.436 0.074 *** PSEE → FI 0.293 0.071 *** PSEE → AI 0.315 0.072 *** * p < 0.05 ** p < 0.01 *** p < 0.001

138

W.-K. Tan and H.-H. Teo

6 Conclusion The results from the survey generally support our proposition that a blogger treats one’s blog as a performing stage to actively present one’s identity to a public audience of readers [6, 9]. This is accomplished with the aid of the various blog design features as props. Through this process of impression management, a blogger can indeed achieve the goal of expressing one’s life and sharing one’s expertise and experience. Consequently, the blogger derives personal satisfaction with the blogging process. The main theoretical contribution of this study is providing a plausible explanation as to why and how the majority of bloggers seek to express their life, and to share their expertise and experience. By adopting a process model, instead of a variance model, we provide a rich insight into the antecedent and consequential factors involved in the impression management process. This model may be adapted to explain why and how individuals use social computing technologies that have the capability to reflect one’s identity, e.g., social networking sites and virtual worlds. Practically, our research findings emphasize the importance for third party blog hosting service providers to make available a well-balanced basket of content, functional and aesthetic blog design features to their bloggers. This will help the majority of the bloggers in their impression management endeavor and thus leading to greater satisfaction. Presumably, this will improve the acquisition and retention of new and existing bloggers.

References 1. Technorati State of the Blogosphere (2009), http://technorati.com/blogging/feature/state-of-theblogosphere-2009 2. Blood, R.: How Blogging Software Reshapes the Online Community. Communications of the ACM 47(12), 53–55 (2004) 3. Nardi, B.A., Schiano, D.J., Gumbrecht, M., Swarth, L.: Why We Blog. Communications of the ACM 47(12), 41–46 (2004) 4. Moon, J., Li, J.P., Sanders, G.L.: The Role of Virtual Social Identity through Blog Use in Social Life. In: 12th AMCIS, Mexico, pp. 4085–4095 (2006) 5. Leary, M.R., Kowalski, R.M.: Impression Management: A Literature Review and TwoComponent Model. Psychological Bulletin 107(1), 34–47 (1990) 6. Goffman, E.: The Presentation of Self in Everyday Life. Doubleday, New York (1959) 7. Wynn, E., Katz, J.E.: Hyperbole over Cyberspace: Self-Presentation and Social Boundaries in Internet Home Pages and Discourse. The Information Society 13(4), 297– 327 (1997) 8. Schau, H.J., Gilly, M.C.: We are What We Post? Self Presentation in Personal Web Space. J. Consumer Research 30(3), 385–404 (2003) 9. Vazire, S., Gosling, S.D.: e-Perceptions: Personality Impressions Based on Personal Websites. J. Personality and Social Psychology 87(1), 123–132 (2004) 10. Bortree, D.S.: Presentation of Self on the Web: An Ethnographic Study of Teenage Girls’ Weblogs. Education, Communication & Information 5(1), 25–39 (2005)

Impact of Blog Design Features on Blogging Satisfaction

139

11. Leary, M.R.: Understanding Social Anxiety: Social, Personality, Clinical Perspectives. Sage, Beverly Hills (1983) 12. Maddux, J.E., Norton, L.W., Leary, M.R.: Cognitive Components of Social Anxiety: An Investigation of the Integration of Self-Presentation Theory and Self-Efficacy Theory. J. Social and Clinical Psychology 6(2), 180–190 (1988) 13. Preece, J., Maloney-Krichmar, D.: Online Communities. In: Jacko, J., Sears, A. (eds.) Handbook of Human-Computer Interaction, pp. 596–620. Lawrence Erlbaum Associates Inc., New Jersey (2003) 14. Walker, K.: It’s Difficult to Hide It: The Presentation of Self on Internet Home Pages. Qualitative Sociology 23(1), 99–120 (2000) 15. Walther, J.B.: Computer-Mediated Communication: Impersonal, Interpersonal, and Hyperpersonal Interaction. Communication Research 23, 3–43 (1996) 16. Ma, M., Agarwal, R.: Through a Glass Darkly: Information Technology Design, Identity Verification, and Knowledge Contribution in Online Communities. Information Systems Research 18(1), 42–67 (2007) 17. Swann, W.B., Polzer, J.T., Seyle, D.C., Ko, S.J.: Finding Value in Diversity: Verification of Personal and Social Self-Views in Diverse Groups. Academy of Management Review 29(1), 9–27 (2004) 18. De La Ronde, C., Swann, W.B.: Partner Verification: Restoring Shattered Images of Our Intimates. J. Personality and Social Psychology 75(2), 374–382 (1998) 19. Jensen, C., Davis, J., Farnham, S.: Finding Others Online: Reputation Systems for Social Online Spaces. In: SIGCHI Conference on Human Factors in Computing Systems, Minnesota, pp. 447–454 (2002) 20. Newcomb, T.M.: The Acquaintance Process. Holt, Rinehart and Winston, New York (1961) 21. Briggs, R.O., Reinig, B.A., de Vreede, G.J.: Meeting Satisfaction for TechnologySupported Groups. Small Group Research 37(6), 585–611 (2006) 22. Reinig, B.A.: Towards an Understanding of Satisfaction with the Process and Outcomes of Teamwork. J. Management Information Systems 19(4), 65–83 (2003) 23. Compeau, D.R., Higgins, C.A.: Computer Self-Efficacy: Development of a Measure and Initial Test. MIS Quarterly 19(2), 189–211 (1995) 24. Hsu, M.-H., Chiu, C.-M.: Internet Self-Efficacy and Electronic Service Acceptance. Decision Support Systems 38, 369–381 (2004) 25. Leary, M.R.: Understanding Social Anxiety: Social, Personality, Clinical Perspectives. Sage, Beverly Hills (1983) 26. Stone, B.: Blogging: Genius Strategies for Instant Web Content. New Riders Publishing, Indiana (2003) 27. Callero, P.L.: Role-Identity Salience. Social Psychology Quarterly 48(3), 203–215 (1985) 28. Bhattacherjee, A.: Understanding Information Systems Continuance: An ExpectationConfirmation Model. MIS Quarterly 25(3), 351–370 (2001)

An Exploratory Study of Navigating Wikipedia Semantically: Model and Application I-Chin Wu1, Yi-Sheng Lin1, and Che-Hung Liu2 1 Department of Information Management, Fu-Jen Catholic University No.510, Zhongzheng Rd., Xinzhuang Dist., New Taipei City 24205 Taiwan 2 Department of Business and Management, National University of Tainan 33, Sec. 2, Su-Lin St. Tainan, 700, Taiwan [email protected], [email protected]

Abstract. Due to the popularity of link-based applications like Wikipedia, one of the most important issues in online research is how to alleviate information overload on the World Wide Web (WWW) and facilitate effective information-seeking. To address the problem, we propose a semantically-based navigation application that is based on the theories and techniques of link mining, semantic relatedness analysis and text summarization. Our goal is to develop an application that assists users in efficiently finding the related subtopics for a seed query and then quickly checking the content of articles. We establish a topic network by analyzing the internal links of Wikipedia and applying the Normalized Google Distance algorithm in order to quantify the strength of the semantic relationships between articles via key terms. To help users explore and read topic-related articles, we propose a SNA-based summarization approach to summarize articles. To visualize the topic network more efficiently, we develop a semantically-based WikiMap to help users navigate Wikipedia effectively. Keywords: Navigation, Normalized Google Distance, Semantically-based, SNA-based summary, Wikipedia.

1

Introduction

With the ubiquity of the Internet and Web 2.0 technologies, the WWW has become the main source of information and knowledge in the modern era. The top sites on the Web, as ordered by Alexa traffic rank, are Google, Facebook, YouTube, Yahoo, Live, Baidu, Wikipedia, Blogger, MSN, and Tencent (October, 2010). Wikipedia is the most popular web-based, free-content encyclopedia web site out of the top 10 Web sites. The statistical data from Wikipedia shows that in 2008, the site welcomed 684 million visitors; there were more than 91,000 contributors working on more than 16 million articles. Because of the popularity of applications like Wikipedia, the number of articles in Wikipedia is constantly expanding. As we know, more and more people regard Wikipedia is an efficient means to find needed knowledge, such as searching definitions of terminologies, exploring articles on related topics, and so on. Basically, Wikipedia A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 140–149, 2011. © Springer-Verlag Berlin Heidelberg 2011

An Exploratory Study of Navigating Wikipedia Semantically

141

users browse content in the traditional manner (i.e., by following hyperlinks) when searching for information. However, users may unconsciously change their search goals or get lost when exploring or retrieving information in Wikipedia. In order to make research more effective for the vast number of Wikipedia users, it is highly important to develop effective search or navigation tools to guide users to find and organize needed information or topics. Generally, users invest a great deal of time browsing by following links or searching for specific information. Because of the rapid growth in the volume of information on the WWW, web mining and information retrieval are regarded as key techniques for finding desired information. Web mining tries to extract potentially useful implicit information, link structures and patterns from information units or activities on the WWW. There are three types of web mining techniques: web content mining, web structure mining, and web usage mining. The main difference between web pages and static text documents is that the former contain content as well as link information, and metadata [1][9]. Web content mining exploits IR and artificial intelligence (AI) techniques to extract, mine and analyze information from web pages. Generally, web content mining strategies can be divided into those implicitly that mine information or knowledge from the content of documents, and those designed to improve the search results, i.e., the information retrieved by search engines. IR technology relies primarily on content analysis techniques, but Web pages are usually noisy and contain various types of content, such as text, images, and multimedia. To resolve this problem, some researchers have exploited the hyperlink structure, which provides hyperlink information for a collection of web pages, and proposed ranking algorithms to rank search results. Analyzing the hyperlink structure between WWW pages to support user search activities has attracted a great deal of attention in recent years. Since the link structure encodes a considerable number of latent human judgments, link mining and analysis techniques are employed by commercial search engines, e.g., the PageRank algorithm [2] used by the Google search engine is one of the most well-known link-based algorithms. Almpanidisa and Kotropoulos (2007) [1] proposed a topical information resource discovery algorithm that implements a focused or topic-driven crawler by combining text and link analysis techniques. Their results show that, in the initial stage, the content-based and link-based algorithm does not need a lot of data and that it outperforms comparable methods. In Wikipedia, a topic may contain many articles; thus, it is difficult for users to locate articles relevant to the topic simply by following the hyperlinks in the articles. To address this problem, we propose a semantically-based navigation system that is based on the theories and techniques of link mining, semantic relatedness analysis, social network analysis (SNA) and text summarization. Specifically, we employ a Link Strength (LS) measure to establish a preliminary topic network by analyzing Wikipedia’s internal links [13]. Our goal is to find the specific topic or related subtopics for a seed query (topic) to construct a preliminary internal link-based network. Moreover, we refine the Normalized Google Distance algorithm [3] in order to quantify the strength of the semantic relationships between articles via key terms, and filter out articles that do not have strong semantic relationships. Our preliminary evaluation results demonstrate the effectiveness of applying semantic analysis in an internal link-based network. To help users search for information, we apply centralitybased and cohesive measures in SNA to summarize multiple articles. The measures

142

I.-C. Wu, Y.-S. Lin, and C.-H. Liu

are k-clique and degree centrality, which identify the sub topics of the seed query and key articles of the sub topics respectively. Then, when the user clicks on a topic node that he/she wants to explore, an SNA-based summary is presented on the interface. To more efficiently visualize the semantically-based topic network, an interface as it would appear on PCs and mobile devices is developed to help users navigate Wikipedia effectively.

2 Basic Concepts 2.1 Semantic Relatedness Analysis: Normalized Google Distance Humans acquire the meaning of words and the relationship among words according to their background knowledge. For example, many humans could answer the question of how “polar bears” or “automobiles” are related to “global warming.” However, it is difficult for computers to make judgments regarding the semantic relationships between keywords. How to find the correct tasks to allow computers to automatically extract the meaning of words has recently gained a great deal of attention among the natural language processing and artificial intelligence communities [3] . Basically, there are mainly three measures to estimate the semantic relatedness of different words, which are thesaurus-based, corpus-based and Wikipeida-based measures [5]. Most of the research methods rely on long-term and labor-intensive efforts to construct the semantic relationships among words. Recently, an automatic algorithm has been proposed using the search results counts from the Google search engine, i.e. the normalized Google distance (NGD) algorithm [3]. The NGD algorithm can detect the semantic relationship among terms using the Google search engine. Based on reports from WorldWideWebSize.com, the Google search engine indexes 23.6 billion pages (December, 2010). Thus, the WWW is the largest corpus that can be used to analyze the semantic relationships among words. The main idea of the NGD is to understand the relationships between any two terms according to the number of search results, i.e., the number of return pages. Vitanyi and Cilibrasi (2007) [3] provided a statistical index based on Google page counts, showing the logical distance of a pair of terms called NGD. When the value gets lower, it implies there is a closer relation between two terms. Eq. (1) is the normalized Google distance based on conditional probability. That is, the probability p(x|y) is defined by p(x|y)=p(x,y)/p(y) and . In the NGD equation, y means the probability, as the number of words that Google searched. The result number, by which it is divided, is M (total numbers of document indexed by Google search). p(x, y) is the probability of queries when searching two terms at the same time. Furthermore, because the conditional probabilities are independent of M, we use frequency, i.e., number of search pages, instead of conditional probability to derive the Eq. (2). D( x, y ) =

NGD( x, y ) =

max{log 1

), log 1 )} p( x | y ) p( y | x) 1 1 max{log ), log )} p ( x) p( y)

max{log f ( x), log f ( y )} − log f ( x, y) log M − min{log f ( x), log f ( y)}

(1)

(2)

An Exploratory Study of Navigating Wikipedia Semantically

143

2.2 Automatic Summarization Techniques With the fierce growth of information on the Web, it is important to extract the key abstract content from information sources. Therefore, text summarization techniques and tools are useful to help users find needed information and make decision quickly. Basically, the aim of text summarization is to extract key sentences from one or more documents to represent the meaning of the target document(s). Generally, text summarization can be divided into three phases: analyzing the source text, determining the salient points and synthesizing an appropriate output [7]. For the number of documents, it could be classified as single-document summarization and multi-document summarization. Single-document summarization is generated from the content of one document by different methods that aim to reduce the redundant information. Forsyth and Rada (1996) [6] researched the approach of computing the TF-IDF weight of a document to figure out important terms. Based on their research, they found that sentences are composed from many terms; thus, they attempt to extract important sentences using signature words of the document. Teufel and Moens (1997) [11] employed five heuristic methods: cue, location sentence length, thematic word method, and title methods to extract important sentences from document training sets. They examine the effectiveness of each method and then integrated five methods to analyze the tradeoff between methods. The experiment results show that cue phrases (i.e. in summary, in conclusion, in short, therefore or proper nouns) are the strongest single heuristic and the combination of 5 heuristics will lead to the best performance. Hovy and Lin (1997) [8] mainly considered the important position of sentences (i.e. the first sentence of a document or a phrase), called sentence position, to generate the summarization directly.

3 The System Framework With the emergence of Web 2.0 technologies, social web sites (i.e., social networking websites and micro-blogging services) provide unprecedented opportunities for sharing user-generated content. Wikipedia, one of the most famous collaborative projects on the Web, has become an extremely popular reference database for people seeking information or knowledge. The process for generating a semantically-based topic network is illustrated in Fig. 1. In the following, we describe the modules of the framework. Article pre-processing module. The proposed framework retrieves all link-related articles within three degrees. Besides extracting the hyperlinks from each article, the module stores the content of the article, including the title and its associated links, in XML format. Basically, the Wikipedia is written in wiki markup language which is based on the XML web language composed of wiki-recognized title, content, and most of all containing huge numbers of out-links and in-links. In this phase, we extract and indentify information we want to use from those special symbols of wiki language. For example, the symbols “[” and “]” are used for out-links, while the symbols “[[” and “]]” present the in-links.

144

I.-C. Wu, Y.-S. Lin, and C.-H. Liu

Fig. 1. The process for generating an SNA-based summaries interface

Internal link-based and semantic network analysis module. This module utilizes the proposed link strength (LS) measure to search for topics related to the user’s seed query. It filters out unimportant articles and finds possible subgroups around the seed query. If the value or strength of a link is less than a specified threshold, the system will remove the link from the internal link-based network. Then, the system executes semantic analysis based on the NGD algorithm to filter noise nodes from the initial network. SNA analysis module. This module applies the social network analysis (SNA) indicators defined in social network theory [12] to recognize the role of articles in the topic network based on the results of the previous stage. Basically, we mainly apply centrality and cohesive measures in SNA, i.e., k-cliques and degree centrality, to identify subtopics and key articles of subtopics to help users effectively search for information in Wikipedia. SNA-based summaries generation module. The aim of this step is to generate the summaries of articles based on the SNA analysis results, namely, the interface will present different summarization results based on the topologies of the network. To help users search for information, we apply SNA indicators to summarize single and multiple articles. Then, when the user clicks on a topic node that he/she wants to explore, an SNA-based summary is presented on the interface to help the user quickly read the related articles.

4 Incorporating Semantic Analysis into the Internal Link-Based Network 4.1 Internal Link Analysis by the LS Measure We use the term “article” to denote an entry in Wikipedia rather than a page on the WWW, and the term “node” to denote a word in an article with a hyperlink to another

An Exploratory Study of Navigating Wikipedia Semantically

145

article. The link strength (LS), which indicates the degree of closeness between two articles, is determined by considering the type and frequency of the links between the articles. Our goal is to find the specific topic or related subtopics for a seed query. An article may have three types of links: in-links, out-links, and reciprocal bidirectional or multi-directional links. Two articles linked to each other by the same nodes have a bidirectional link. In addition to the relationships between articles, the frequency of the links (the link frequency) between the nodes is determined by the LS measure, denoted by ζ, which is calculated as follows: ζ(ai , aj) =(f_bi(ai , aj))w1 + w2×f_in(ai , aj)+w3×f_out(ai , aj),

(3)

where f_in(ai , aj) denotes the frequency of in-links from aj to ai; f_out(ai , aj) denotes the frequency of out-links from ai to aj ; and f_bi(ai , aj) denotes the frequency of bidirectional links between ai and aj. Details of setting the relative weights of in-links and out-links and the threshold of the LS value is shown in our recently work [13]. 4.2

Semantic Relatedness Analysis by the NGD Algorithm

For filtering articles with low semantic relatedness in the initial internal link-based network, we conduct further tests for semantic relatedness via the key terms of each article. As we know, it is a difficult task to make judgments of the semantic relationships between keywords via computers. Recently, an automatic algorithm has been proposed using the number of search results returned by the Google search engine, i.e. normalized Google distance (NGD) algorithm, as we introduced in Section 2.1. In our research, we defined the titles of the articles in Wikipedia as the nodes in the NGD algorithm. We can find out the relevancy strength of two web pages by calculating and analyzing the titles of these two pages. A value will be given to the distance of the relationship using the NGD algorithm. Notably, the lower the value is, the higher the semantic relationship is between articles. We filter topics that have a distance value higher than the threshold aim in order to remove noise. The formula of the algorithm is given in Equation (2). Table 1. NGD threshold value of the seed query of “History of personal computers” Title of the Article 74181(ALU) BASIC Bluetooth Burroughs_Corporation Laser_diode Left-handedness Pixar PlayStation_2 The_Walt_Disney_Company Video_game Wii_Remote

Judgment(Irrelevant)

NGD Value

2 2 2 3 2 3 2 3 3 3 5

0.464 0.876 0.971 0.444 1.000 1.000 0.896 0.736 0.421 0.836 0.421

Average: 0.695

146

I.-C. Wu, Y.-S. Lin, and C.-H. Liu

Testing of NGD threshold. Based on Evangelista and Kjos-Hanssen’s research (2006) [4], the expected value of the NGD threshold will be around 0.7. Accordingly, we invited experts from the information management department to determine if the nodes in the topic network are relevant to the seed query of “History of the personal computer”. Table 1 shows the results of the unrelated nodes’ titles with the seed query, derived using the judgment of the experts. The results shows that the average NGD value of all those unrelated nodes is 0.695, which is similar to the execrated results from previous research, i.e., the NGD value is 0.7 [4]. In our the other test, the average NGD value of all those unrelated nodes for the seed query of “knowledge management” was 0.732. Thus, we set the value of the threshold as 0.7 to construct the final topic network toward the seed query. Similarity Calculation. Notably, we also conduct a similarity calculation using the cosine measure to investigate the cohesiveness of each network. Table 2 shows that after applying NGD filtering, the networks will have a higher cosine value than those for which semantic analysis has not been conducted. Table 2. Similarity comparison between IIN and IIN with NGD Queries History of PCs Star Trek Abraham Lincoln Knowledge Management

Internal-link based network (IIN) 0.248 0.149 0.103 0.261

IIN With NGD Analysis 0.240 0.275 0.148 0.291

5 SNA-Based Summaries for the Topic Network 5.1 Process for Generating Summarization Summaries can be divided into three types based on purpose: indicative summaries, informative summaries and critical summaries. In our research context, we generate informative summaries to help users to explore the topics in Wikipedia, with the aid of shorter versions of the original articles. The information summary provides a highly reduced list of the important content of the document, which users could even use to replace the original document. The 30% of a text that is important always represents 80% to 90% of the article’s original focus. Fig. 2 shows the two phases needed to generate the SNA-based summaries, i.e., the analysis phase and the synthesis phase, in our research context. The steps for generating summarization proceeded as follows: (1) Pre-process the source articles; (2) parse articles into terms; (3) select the terms based on the feature set; (4) weight the terms based the role of the articles in the SNA; (5) Calculate sentence scores based on the feature set; and (6) generate summaries from top-N sentences and then resort them based on the sentences’ original order in the articles. We consider the roles of articles in the social network. Accordingly, we then apply different summarization strategies based on the role of articles in the social network. First, the cohesive article (CA) will be identified based on the formula of k-cliques given in the social network analysis (SNA). The analysis results help us label the sub-

An Exploratory Study of Navigating Wikipedia Semantically

147

topics of the nodes in the network. Second, we will find the hub article (HA) of each sub-topic. The HA is analyzed based on the formula of degree centrality given in SNA. This seeks to help the user explore the main topic of each sub-topic. Similarly, we also present a guided summarization of the hub article and its associated articles.

Fig. 2. Process for generating SNA-based summaries

5.2 Identify Sub-topics and Extract the Summaries for the Topics Radev et al. (2004) [10] presented the MEDA algorithm to generate the summaries of multiple multi-documents. The algorithm adopted three features to select the top sentences which are centroid value, positional value, and first-sentence overlap. Similar to previous researchers, we consider the characteristics of the articles in Wikipedia in order to extract key terms and then select the top sentences for extracting summaries of the topic or sub topic. In this work, we proposed three features, which are first sentence, weighted first-sentence overlap and cue phrases to select key terms. Particularly, we incorporated the concept of the position value of a sentence into the concept of the first-sentence overlap to generate the score of each sentence, i.e. weighted first-sentence overlap, wfi, as shown in Eq.(4). Basically, position value, Pi, means the position of the ith sentence in the document, and n is the number of sentences of the target article. Furthermore, the candidate terms can be selected from these sentences, i.e., sentences with high value of wfi., as shown in Eq. (5). The final key terms are selected based on different relative importance of the first sentence Eq.(6).

S1 ,

weighted first-sentence overlap ak , and cue phrases,

CPk

,as shown in

wf i = Sim( f1 , f i ) × Pi where Pi =

n − i +1 n

(4)

n

a k = ∑ wf si × S i

(5)

a'k = λ × top( ak ) + (1 − λ ) × ( S1 + CPk )

(6)

i=2

148

I.-C. Wu, Y.-S. Lin, and C.-H. Liu

6 The Illustrative Example and Application The following table shows the results of the article summarization for the seed query “knowledge management.” We adjusted the relative importance of the weights for the weighted first-sentence overlap, first sentence, and cue phrases based on the Eq. (6). We asked two experts to select the sentences which are highly relevant and useful to assist novices understand the subject of “knowledge management.” There are 20 distinct sentences are selected from the article for we set the value of λ at 0.1, 0.5, and 0.9 respectively, based the Eq. (6). Both of experts agree that the meaningful and useful sentences will be extracted when we set λ at 0.5. Table 3 shows the results of the top-5 sentences selected from the article for we set the value of λ at 0.5. The sentences with bold type denote the overlapping sentences among three different settings of weights. Finally, we present the interface for the semantically-based WikiMap as it would appear on a mobile device and personal computer, as shown in Figure 3(a) and 3(b) respectively. In the future, the effectiveness of the proposed applications will be evaluated in the real setting. Table 3. Article summarization for the seed query “knowledge management” The weight of weighted first-sentence overlap is set to 0.5 1. Knowledge Management efforts typically focus on organizational objectives such as improved performance, competitive advantage, innovation, the sharing of lessons learned, integration and continuous improvement of the organization. 2. Knowledge Management (KM) comprises a range of strategies and practices used in an organization to identify, create, represent, distribute, and enable adoption of insights and experiences. 3. More recently, other fields have started contributing to KM research; these include information and media, computer science, public health, and public policy. 4. Many large companies and non-profit organizations have resources dedicated to internal KM efforts, often as a part of their business strategy, information technology, or human resource management departments (Addicott, McGivern & Ferlie 2006). 5. KM efforts overlap with organizational learning, and may be distinguished from that by a greater focus on the management of knowledge as a strategic asset and a focus on encouraging the sharing of knowledge. Note: Available at: http://en.wikipedia.org/wiki/Knowledge_management (accessed 30 January 2011).

Fig. 3(a). Semantically-based WikiMap Fig. 3(b). Semantically-based Presented in the Mobile Device Presented in the PC

WikiMap

An Exploratory Study of Navigating Wikipedia Semantically

149

7 Conclusion and the Future Work In this present study, we proposed an SNA-based summarization model and the development of search or navigation applications to guide users to find and organize needed information or topics within Wikipedia. We employed the NGD algorithm in the proposed LS measure to quantify the strength of the semantic relationships between articles in the topic network. Our preliminary evaluation results demonstrate the effectiveness of applying semantic analysis in an internal link-based network. To help users quickly read topically related articles, we proposed SNA-based summarization to present single article’s summaries in a newly developed interface. This study also presents the interface for the semantically-based WikiMap as it would appear on PCs and mobile devices. In the future, we will evaluate the precision and accuracy of the summaries for the articles based on the proposed methods. Moreover, the effectiveness of the proposed applications will be evaluated in the real setting. Acknowledgments. This research was supported by the National Science Council and Fu-Jen Catholic University of Taiwan under the Grant No. 99-2410-H-030-047-MY3 & No.409931074078, respectively.

References 1. Almpanidisa, G., Kotropoulos, C., Pitas, I.: Combining Text and Link Analysis for Focused Crawling—An Application for Vertical Search Engines. Information Systems 32(6), 886– 908 (2007) 2. Brin, S., Page, L.: The Anatomy of a Large-scale Hypertextual Web Search Engine. Computer Network 30(1-7), 107–117 (1998) 3. Cilibrasi, R., Vitányi, P.: The Google Similarity Distance. In: IEEE Transactions on Knowledge and Data Engineering, pp. 370–383. IEEE Press, New York (2007) 4. Evangelista, A., Kjos-Hanssen, B.: Google Distance between Words. Frontiers in Undergraduate Research. University of Connecticut (2006) 5. Finkelstein, L., Gabrilovich, Y.M., Rivlin, E., Solan, Z., Wolfman, G., Ruppin, E.: Placing search in context: the concept revisited. ACM Transactions on Information Systems (TOIS) 20(1), 406–414 (2001) 6. Forsyth, R., Rada, R.: Adding an Edge in Machine Learning: Applications in Expert Systems and Information Retrieval. Ellis Horwood Ltd., pp.198-212 (1986) 7. Hahn, U., Mani, I.: The Challenges of Automatic Summarization. Journal of IEEE Computer 33(11), 29–36 (2000) 8. Hovy, E., Lin, C.Y.: Identifying Topic by Position. In: Proceedings of the 5th Conference on Applied Natural Language Processing (ANLP), Washington, DC, pp. 283–290 (1997) 9. Liu (ed.): Web Data Mining: Exploring Hyperlinks, Contents and Usage Data. Springer, New York (2007) 10. Radev, D.R., Jing, H., Stys, M., Tam, D.: Centroid-based Summarization of Multiple Documents. Information Processing and Management 40(6), 919–938 (2004) 11. Teufel, S., Moens, M.: Sentence extraction as a classification task. In: Proceedings of the Workshop on Intelligent Scalable Summarization. ACL/EACL Conference, Madrid, Spain, pp. 58–65 (1999) 12. Wasserman, S., Faust, K. (eds.): Social Network Analysis: Methods and Applications. Cambridge University Press, UK (1994) 13. Wu, I.-C., Wu, C.-Y.: Using Internal Link and Social Network Analysis to Support Searches in Wikipedia: A Model and Its Evaluation. Journal of Information Science 37(2), 189–207 (2011)

Productivity Improvement by Using Social-Annotations about Design Intent in CAD Modelling Process Gerardo Alducin-Quintero1, Manuel Contero1, Jorge Martín-Gutiérrez2 David A. Guerra-Zubiaga3, and Michael D. Johnson4 1 2

13BH, Universitat Politècnica de València, Camino de Vera s/n, 46022 Valencia, Spain Universidad de La Laguna, Avda. Astrofísico Fco. Sánchez s/n, 38206 La Laguna, Spain 3 CIDESI Texas, Caldwell Waston Bldg., 1700 Research Pway, College Station, Texas, USA 4 Dept. of Engineering Technology and Industrial Distribution. Texas A&M University, 3367 TAMU, College Station, Texas, USA [email protected], [email protected], [email protected], [email protected], [email protected]

Abstract. This paper focuses on the New Product Development Process (NPDP) area to contribute to increasing the productivity of CAD users by means of an improved design intent communication using a social-annotation technique. Design teams operate in a similar way to an online social network, and CAD models are not just only a 3D geometry representation, as they reflect the result of an specific modeling strategy, that usually constrains the future capability for modification and reuse of the existing 3D model. Considering this context, this work tries to assess the impact of annotations in the engineering change process in order to determine its influence in the user performance during this process. Preliminary experimental results obtained from several experiments with Spanish CAD students indicate that it is possible to reduce the time needed to perform engineering changes in existing models into a 10-20% range, if those CAD models provide annotations explaining the original design intent. Keywords: New Product Development Process, CAD, Social-Annotations, Design Intent, Engineering Change Process.

1 Introduction The New Product Development (NPD) process is a key activity for the enterprise survival and competitiveness. Empowering the role of design and shortening the development cycle of new products are two well-known strategies to improve this process. In this context, advanced CAD systems are an important tool to support the NPD process, providing a key advantage to improve organizational effectiveness [1]. This paper focuses on this area in order to contribute to increasing the productivity of CAD users by enhancing design intent communication by means of a socialannotation technique, inspired on the idea that design teams operate as an online social network. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 153–161, 2011. © Springer-Verlag Berlin Heidelberg 2011

154

G. Alducin-Quintero et al.

Design intent governs the relationships between modeling features in a part, and between parts in assemblies. It is related to the behavior of the CAD system when a modification is performed on a CAD model. CAD users should analyze the geometry of the part to be generated, and plan what is the most efficient sequence of modeling features, in order to assure that future modifications in the design are managed by the CAD system in a determined expected way. Usually good CAD models are associated to change flexibility and robust variation capabilities [2] [3] [4]. Two kinds of knowledge are related to the design intent concept: declarative knowledge that is knowledge of facts (knowing that or knowing what) and procedural knowledge that is knowledge of how to do things (knowing how) [5]. This means that a CAD model is not just only a 3D geometry representation, but also this model stores the know-how about the strategy used to build it. Capture, share and transfer this complex knowledge is a key element to improve the NPD process. Through this work is attempted to demonstrate the importance of going beyond the capture of design intent in a traditional way, based on non-explicit information associated to the feature types used in the CAD model and their interrelationship. This is aligned to Guerra-Zubiaga [6] work trying to structure different knowledge types to support manufacturing and design decisions. This work proposal includes an innovative solution by using design annotations relative to the CAD modeling strategy, specifically, about the decision making process during the 3D model construction. Design annotations give an important clue about how and why a CAD feature was used to build a CAD model by using just a few words. This research activity is directed to assess the impact of these annotations in the engineering change process and in second place in analyzing the dynamics of the social annotation process. This paper is centered in the first aspect, presenting a pilot study that has been conducted using CAD models related to the automotive industry. These CAD models represent parts with different geometry complexity levels. For each part, two different CAD models were created, with and without annotations. These models were modified by two classes of engineering students (one mechanical engineering group at the undergraduate level, and other belonging to a master degree on CAD/CAM/CIM). Both classes were given the same written instructions where it was explained a series of engineering changes to be performed by modifying the geometry of the CAD models. For each class a control and an experimental group were randomly organized. Experimental groups had to change the models that contained annotations about the design strategy used by the original author of the model, while the control groups had not available this additional information. Preliminary results indicate that it is possible to reduce the time needed to perform engineering changes in existing models into a 10-20% range by using the annotation technique. It was determined the basic knowledge-mapping and examined the problem-solving process employed by the participants in the modification of constraint-based CAD models [7]. The initial hypothesis was confirmed: the experimental group performed the design changes using less time. Additionally it was noticed that certain patterns of behavior of participants coincided with previous research works like: [1] [3] [8]. In the next point, is provided all the details about the experimental design followed to analyze the influence of annotations on the modification of CAD models. Then

Productivity Improvement by Using Social-Annotations

155

results are presented, explaining the limitations and main conclusions of this preliminary study and giving some orientation about future works about this topic.

2 Hypothesis and Methods 2.1 Hypothesis Definition This piece of research work pursues to explore how design annotations influence in the user performance during the engineering change process. This is performed before analyzing the dynamics of the social annotation process, because it is needed to assure that the availability of these annotations provides an added value to the CAD models. This research hypothesis that CAD operators, using annotated models where original design intent is made explicit, are more efficient dealing with CAD model modifications. In this context, efficiency is related to the time used by CAD users when they have to perform a change in a 3D geometric model to accomplish an engineering change order. The author’s perspective is: if the design intent knowledge is made explicit by means of these annotations, the corresponding CAD models will be created with better semantic quality, following the product data quality model by Contero et al. [2]. Author´s vision is that this annotation process can be performed following the behavior of social networks, where knowledge associated to the CAD modeling process is made explicit by the collaborative annotations performed by design engineers. 2.2 Experimental Stage The experimental stage was divided in two phases in order to test the hypothesis: phase I, called as “undergraduate study”, and phase II, named as “postgraduate study”. All the participants were given the same written explanation and modeling tasks with accompanying figures, and they were asked to perform a series of engineering changes that consisted in geometric modifications on the original 3D CAD models. The parts were created thinking that the references used for the creation of each feature impacted the ability for later modification and edition of the geometry. This includes capturing design intent during geometry creation. Specifically, in the compound parts (e.g. automotive components) some references (e.g. datum planes or sketches) were used to separate in groups of CAD features that define each area of the part. The used models are divided in two sets: simple geometry (low amount of CAD features used to build it) and compound geometry (composed by a high amount of CAD features). As simple geometry parts were used a housing of gear box and the part used by Johnson in [9]. As complex geometries were used a component of a PC fan housing base (from Brigham Young University’s NX CAD Manual) and an automotive radiator collector (from a Spanish company). In both cases, the same type of geometry modifications was required: simple changes (e.g. modify the height of one element that is controlled by an expression) and complex changes (e.g. create a copy of a group of features with strong parametric parent-child relationships). Two variants for each CAD model were created: one without any annotation information

156

G. Alducin-Quintero et al.

explaining design intent (named regular CAD model) and other one with explicit textual annotations about the design intent (named annotated CAD model). The content of texts in the annotated CAD models provided information about: • Location. This kind of text comment allows identifying the set of CAD features that define a geometric element of the part (indicating the first and last feature of a specific element). • Design Intent. This commentary allows understanding the intention of the author of the model (e.g. this profile was used to create the main body of the connection). • Modification Procedures. This comment gives information about both recommended steps for making modifications and procedures that must be avoided (e.g. recommendation, don’t use mirror copy feature). 2.3 Undergraduate Experience Context. Firstly, the undergraduate students group (composed by thirty students and denoted as the “undergraduate group”) was split into two groups of fifteen participants. Students were enrolled in a CAD course from a Mechanical Engineering degree in La Laguna University (Spain). The course was based on Autodesk Inventor. The difference between the experimental group and control group was in the CAD model files that they received in order to conduct the study. The experimental group received CAD files with annotations, by using the Inventor Engineer Notebook functionality, that allows to create CAD annotations that support both text and images. The control group received the same CAD files, but stripped of any annotation information. The observed variable was the time expended in performing each modification requested to the participants. Students were given 50 minutes as maximum time to complete all the required modifications. They controlled the time, writing down the initial and final time for each requested geometric modification. Afterwards this time list was rechecked using the time stored into the participant’s CAD files in order to avoid inconsistencies. Availability of numerous groups of CAD users with a similar knowledge and expertise level in modelling is a very limited, outside the academic world. CAD students provide an interesting study population, whose behaviors can be extrapolated to the professional and industrial world. With respect to the sample size, Polkinghorne [10] and Meyer & Booker [11] recommend a number between five and twenty designers for an exploratory phenomenological study, which is verified by the present experience. Participant students showed homogenous basic skills as Inventor users, and completed the modeling exercises as requested. Knowledge Mapping Tasks. In both groups, participants showed a similar amount of declarative knowledge (knowledge about CAD commands) but they had a lack of procedural knowledge (knowledge of how to apply the commands to achieve a goal) due to their short experience. The high number of create-erase contiguous events was an interesting behavior pattern exhibited by the undergraduate group. This is related to the fact that they only knew the most basic commands of the CAD system, so they didn’t waste time trying to use complex commands or trying to find them in the CAD user interface. The majority of the participants that completed the exercises used the most simple and direct solution.

Productivity Improvement by Using Social-Annotations

157

2.4 Postgraduate Experience Context. The postgraduate students group (composed by eighteen students and denoted as the “postgraduate group”) was divided into two groups of nine participants. Students were enrolled in an advanced CAD course from the CAD/CAM/CIM Master degree at Universitat Politècnica de València (Spain). The course was based on Siemens NX CAD system. The experimental group received CAD files with annotations. The control group received the same CAD files, but stripped of any annotation information. The observed variable was the time expended in performing each modification requested to the participants. Students were given 60 minutes as maximum time to complete all the required modifications. They controlled the time, writing down the initial and final time for each requested geometric modification. Afterwards this time list was rechecked using the time stored into the participant’s CAD files in the same way that during the undergraduate experience. At the beginning of this research work, was supposed that the impact of the annotations depends on several factors. One of these factors is the user knowledge and expertise. Taking into account that postgraduate students have a more extensive experience with real design problems, exercises used in this experience, have been taken from real industrial CAD models. That’s the reason that in this second experimental phase was tried to organize a postgraduate students group with real design experience in the industry. This means, that they knew several CAD systems, with a medium level of expertise, being one of these systems Siemens NX. However, this group was observed to have a less homogenous behavior that the undergraduate group, due to the different background of participants. Knowledge Mapping Tasks. It was detected that the participants’ expertise influences their performance during the decision-making process to complete an engineering change request. For example if they must create a copy of a geometric element and they know that in other CAD software there is a command to perform a mirror-copy operation, they invest time to find this CAD function, although perhaps it does not exist in Siemens NX or it is not implemented in the same way than in other CAD system. This behavior is related to the fact that the annotations are easy to visualize and their content is relevant to perform the design change in a specific way. For example, if the annotations had contained direct recommendation to perform the specific design change, this expert reaction could have been avoided.

Fig. 1. Simple Engineering Change Request

158

G. Alducin-Quintero et al.

Fig. 2. Complex Engineering Change Request

3 Results 3.1 General Results and Statistical Analysis In the first experimental stage (undergraduate group) only fourteen students completed the engineering changes requested in exercises one and two, as it is seen in Table 1. The average time for those completing the exercise one was 37 minutes. For this exercise the experimental group (with annotations) there were six students and their average time was 28 minutes, while the control group (without annotations) there were eight students and their average time was 37 minutes. The average time for completing engineering changes in the exercise two was eight minutes. In this case the experimental group had an average of five minutes and the average time of the control group was eight minutes. In both cases the control group had an average completion time that was between 26-37% greater than the experimental group. During the second experimental stage (postgraduate group) the results were: for the first exercise only 16 students completed the engineering change requested with a average time of 49 minutes. Specifically the experimental group (eight students) had an average time of 43 minutes and the control group (eight students) had 52 minutes. In the case of the second exercise only 14 students completed the geometric changes requested with an average time of 50 minutes. The experimental group (seven students) had an average time of 47 minutes and for the control group (seven students) was 54 minutes. In both cases the control group had an average completion time that was between 13-17% greater than the experimental group. Table 1. Experimental Phase I: Undergraduate Group Group Total Students Participating Exercise One Number Completing Exercise Average Time for Exercise Two (min.) Standard Deviation Exercise Two Number Completing Exercise Average Time for Exercise Two (min.) Standard Deviation

Control 15

Experimental 15

8 37 7.305

6 28 8.262

8 8 3.583

10 5 2.846

Productivity Improvement by Using Social-Annotations

159

Table 2. Experimental Phase II: Postgraduate Group Group Total Students Participating Exercise One Number Completing Exercise Average Time for Exercise Two (min.) Standard Deviation Exercise Two Number Completing Exercise Average Time for Exercise Four (min.) Standard Deviation

Control 10

Experimental 9

8 52 21.715

8 43 20.584

7 54 17.321

7 47 17.153

Table 3. Statistical t Analysis for Hypothesis Exercise / Phase E1/Ph I E2/Ph I E1/Ph II E2/Ph II

Time (minutes) Control Experimental 37 28 8 5 54 43 54 47

T -2.243 -2.165 -0.993 -0.729

Significance 0.045 0.046 0.338 0.480

While the results in the first phase (undergraduate group) showed that there is difference between control and experimental groups, the results were further analyzed to detect any statistically significant differences. The results of those statistical tests are shown in the table 3. This table shows the relevant statistic and one-tailed probability. Although in all the cases, experimental group (who’s received additional information through annotations) showed a better performance, this means they used less time than the control group. But only there is a statistically significant difference with the undergraduate group experience. The authors relied on that is because the undergraduate group has more homogenous understanding about design and CAD software than other groups [12].

4 Limitations and Future Work One of the most important limitations of this research work relates to the sample size of the CAD users that participated in the experiences. The sample size was small as this pilot study, being exploratory in nature, was intended to give us suggestions for more extensive studies in the future, where larger number of participants would allow to perform a more robust statistical analysis. Other factor that can be considered a limitation is the users’ expertise. The undergraduate group students showed low level CAD skills (they were enrolled in a basic level CAD course) but a homogenous background. This was a positive factor to obtain a result concordant with the initial hypothesis. The postgraduate group of students had a different profile. The majority of them had a real industry experience in design, showing a heterogeneous professional background with different CAD skills. This heterogeneity could have let to the lack of more conclusive results.

160

G. Alducin-Quintero et al.

Fig. 3. Impact of Design Intent Annotations

Future work will be focused on mitigating some of these limitations mentioned above and improving the analysis of the results. The first step would be to find a larger group of students for participating in the experiments. They should offer a similar experience and expertise. Second step would be the redesign of experiments applying all the experience obtained during the previous study.

5 Conclusions Product design can be considered a social process involving groups of individuals (i.e., design teams) attempting to meet complex product needs through sharing varied expertise and knowledge types [13]. It is a creative process involving networks of people with negotiations and interactions between them; the product design itself is dynamic and evolving [14, 15]. This work explored the effects that design intent annotations had during the engineering change process and assessed the possible relationship between annotations’ content and the designer’s knowledge related to an expertise level. The findings presented in this paper show that the productivity impact through CAD annotations depends of several factors such as geometry among others. It can be shown that if a simple design change is performed on a simple geometry the annotation impact is low or null. However, if a complex change is required on complex geometry the annotation impact is significant. In other words, the annotation impact depends on how it can be structures according to the annotation’s content and the user’s knowledge as shown in figure 3.

Productivity Improvement by Using Social-Annotations

161

It has been shown the need of explore new knowledge infrastructure to support better design decisions, not depending on a particular computer design platform through product design development. Acknowledgement. The authors acknowledge the support received by UPV, CIDESITEXAS, Texas A&M University, and Universidad de la Laguna. The authors also acknowledge the support by Tecnológico de Monterrey through the Research Chair in Nanotechnology and the valuable support of Oscar Martínez and Germán Manacera on the experimental performance of this piece of work.

References 1. Anderl, R., Mendgen, R.: Parametric design and its impact on solid modeling applications. In: Third ACM Symposium on Solid Modelling and Applications, pp. 1–12. ACM, New York (1995) 2. Vila, C., Contero, M., Company, P.: Extended modeling, a tool for cooperative design. In: 6th International Conference on Concurrent Enterprising, Tolouse France (2000) 3. Anderl, R., Mendgen, R.: Analyzing and optimizing constraint-structures in complex parametric CAD models. In: Bruderlin, B., Roller, D. (eds.) Geometric Constraint Solving and Applications, pp. 58–81. Springer, Berlin (1998) 4. Hartman, N.W.: The development of expertise in the use of constraint-based CAD tools. Eng. Design Graph. J. 68, 14–26 (2004) 5. Rynne, A.: AC 2007-2132: Cognitive modelling strategies for optimum design intent in parametric modelling (PM). American Society for Engineering Education (2007) 6. Guerra-Zubiaga, D.A.: A Model to enable Knowledge Maintenance: Supporting Manufacturing Decisions. VDM Verlag, Germany (2009); ISBN 978-3639209983 7. Hartman, N.W.: Defining expertise in the use of constraint-based CAD tools by examining practicing professionals. Eng. Design Graph. J. 68, 6–15 (2005) 8. Bhavnani, S.K., Garrett, J.H.J., Shaw, D.S.: Leading indicators of CAD experience: Paper presented at the Proceedings of Computer-Aided Architectural Design Futures CAAD Futures 1993, pp. 313–334 (1993) 9. Johnson, M.D., Diwakaran, R.P.: Assessing the effect of incentive on computer-aided design intent. In: The ASME 2009 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference, San Diego, California, pp. 523–532 (2009) 10. Polkinghorne, D.E.: Phenomenological research methods. In: Valle, R.S., Halling, S. (eds.) Existential-Phenomenological Perspectives in Psychology, pp. 41–60. Plenum, New York (1989) 11. Meyer, M.A., Booker, J.M.: Eliciting and analyzing expert judgment: A practical guide. Academic Press, San Diego (1991) 12. Reddy, J.M., Finger, S., Konda, S., Subrahmanian, E.: Design as Building and Reusing Artifact Theories: Understanding and Supporting Growth of Design Knowledge. In: The Design Productivity Debate. Springer, Heidelberg (1998) 13. Bucciarelli, L.L.: Designing engineers. MIT Press, Cambridge (1994) 14. May, P., Ehrlich, H.C., Steinke, T.: Mapping Sociotechnical Networks in the Making. In: Bowker, G., Star, S.L., Turner, W., Gasser, L. (eds.) Beyond the Great Divide,Technical Systems and Co-operative Work, Lawrence Erlbaum, Mahwah (1997) 15. Bhavnani, S.K., Garrett, J.H.: Leading Indicators of CAD Experience. In: Flemming, U., Van Wyk, S. (eds.) CAAD Futures 1993, pp. 313–334. Elsevier Science Publishers, Netherlands (1993)

Handshake: A Case Study for Exploring Business Networking for the Enterprise, Inside and Out Laurie E. Damianos, Donna L. Cuomo, and Stan Drozdetski The MITRE Corporation, 202 Burlington Road, Bedford, MA 01730, USA {laurie,dcuomo,drozdetski}@mitre.org

Abstract. MITRE has developed and launched Handshake, a social networking platform for business networking use. We took a unique approach in providing a single MITRE-hosted platform for connecting, collaborating, and networking with internal employees and external partners. The business networking prototype also serves as a research platform for building and deploying other social capabilities and exploring the value of social media for the enterprise. Since its initial release, Handshake has grown to support over 4300 users (~1000 are external participants) and 450 groups. Feedback from our early adopters has been very positive, with compelling stories on how they are able to engage our sponsors, mission users, and other partners in ways not possible with previous tools. We discuss the initial results of our study here and provide examples of how Handshake is able to meet our corporate objective of bringing the larger community to bear on problems of national importance. Keywords: social networking, business networking, social software, social computing, web 2.0, social media, collaboration, collaborative environments, collaborative computing, computer-supported cooperative work, evaluation, business value, web-based interaction, group and organization interfaces, Elgg.

1 Introduction Social networking tools (e.g., Twitter, Linked-In, Facebook, Ning, MySpace) have been proliferating on the internet. We describe a social networking tool as a webbased service which allows users to construct a profile, explicitly declare a connection with other users, view and browse their connections and those of others [1]. Many of these tools can also be used for personal publications, status updates, broadcasting, and direct messaging. Because of the social nature of the tools, they can provide situation awareness of one’s personal network as long as people frequent these sites and make contributions. Statistics indicate that many people visit these sites at least monthly, with frequent users visiting several times a day. According to [2], there are almost 600 million Facebook users worldwide. Most social networking sites primarily support pre-existing social relations [3]; in fact, Facebook users search for people they already know more than they browse for people to meet [1], [3]. While humans can recognize up to 2000 connections [5], they can maintain only 150 [5], [6]. Indeed, the average number of Facebook friends is 130 A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 162–171, 2011. © Springer-Verlag Berlin Heidelberg 2011

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

163

[7], [8]. On average, people have 10-15 strong ties [5] (e.g., close confidantes, people who can be relied upon in an emergency, and people with multiple shared interests [9]), and those networks tend to reinforce beliefs rather than introduce new ideas. Weak tie networks provide access to a variety of ideas and experiences and expand the scale and the scope of a given network [9]. Both strong ties and weak ties can be sustained on social networking tools [3], [10], and latent ties (not socially activated) can be converted to weak ties on these sites [3], [11]. Individuals can use their social capital to gain access to resources from other members of their network [3]. Having connections to individuals outside one’s immediate network can introduce non-redundant information, new perspectives, and other benefits (e.g., employment connections), and therefore weak ties help build social capital [12], [13], [14], [15], [16]. In fact, a MITRE-Babson study on the impacts of technology-mediated networks showed a positive relationship between the number of diverse connections in someone’s network and their innovative behavior [17]. Indeed, social software has become “crucial in supporting and strengthening collaboration and nonroutine work in businesses” [18]. As early as November 2006, approximately 22,000 organizations had Facebook directories [19]. Gartner explains that “The demand from workers is escalating, and they can and are turning to the consumer Internet if their corporate technology provider isn’t offering a solution” [18]. To meet this demand, some organizations have started deploying social networking capabilities, but many are still in the proof-of-concept or early stages [20]. Those sites that have been instantiated are primarily for internal enterprise use only [21], [22], and the Burton Group reports that these organizations are faced with many non-technical issues such as “business case, metrics, policies and controls, roles and responsibilities, employee participation models, and cultural dynamics” [20]. In this paper, we describe our own unique approach of designing and deploying a single social networking platform for the enterprise that supports both internal and corss-organizational business networking.

2 Case Study The MITRE Corporation [23] is a not-for-profit organization with expertise in systems engineering, information technology, operational concepts, and enterprise modernization. In addition to managing several Federally Funded Research and Development Centers (FFRDCs), MITRE supports its own independent technology research and application development for solving sponsors’ near-term and future problems. MITRE has approximately 7000 scientists, engineers, and support specialists distributed across many locations and working on hundreds of different projects for various sponsors in numerous domains. MITRE employees have traditionally relied on email, telephone, and face-to-face meetings to communicate and collaborate with external partners. Artifacts distributed via email get lost or conflicts occur when different copies of materials are edited. Telephone conversations are ephemeral, and face-to-face meetings involve travel and scheduling issues. The company also uses an external Microsoft SharePoint site as a secure document repository, but the strict authentication requirements, the lack of

164

L.E. Damianos, D.L. Cuomo, and S. Drozdetski

lightweight functionality, and the limited social aspects of the tool make it difficult for users to connect socially and collaborate. Forming ad hoc group workspaces around topical areas is also difficult with existing tools. In addition to these problems, individuals want to be able to establish relationships with external parties and manage those contacts as people move from job to job. The recent popularity of tools like LinkedIn and Facebook make connecting and tracking possible, but they are not trusted environments for conducting business and are fraught with overwhelming amounts of personal and extraneous information [24]. As a result, it is currently difficult to capture the business networks and relationships that MITRE staff have with people from other organizations. To address these and other collaboration issues, a research team at MITRE embarked upon building a trusted environment for MITRE and its partners to connect, collaborate, and share new information. Our goals were to improve MITRE’s ability to establish and maintain relationships, form groups and facilitate multiorganizational collaboration around topics, leverage expertise from industry, and bring broader segments of the community to bear on important sponsor problems. In addition, we wanted to enhance the individual – and enterprise – situational awareness around relationships, activities, topics, and communities. A specific focus of this project was how to address and mitigate the risks of this new capability. We needed to ensure that the social networking platform – supporting both internal and external access – and its data-level access model were both secure and usable. From a research perspective, we were also interested in exploring and assessing the value of an enterprise networking tool for business use as well as identifying and evaluating emerging business models enabled by social networking tools. Could the enterprise move away from supporting teams to leveraging networks and make the enterprise, as a whole, more efficient?

3 Approach Before selecting and deploying a social networking tool, we started by gathering requirements and identifying business need use cases. We conducted work practice studies by interviewing and observing 18 MITRE employees who were engaged in cross-organizational projects and teams with sponsors, vendors, contractors, consultants, academia, and other FFRDCs. We documented work practices, assessed needs that were not being met, identified breakdowns and ineffective use of existing technology, and collected additional requirements for our prototype tool. The team also explored existing social networking tools and products – including commercial technology, open source tools, and hosted platforms – to understand what functionality and options existed. We used this data, along with extensive criteria, to select an appropriate tool that would support a broad spectrum of research efforts. Once we had a research platform selected, we built upon it and deployed it outside our firewall. We identified teams and partners to support as we piloted the tool and used their feedback to make improvements. We leveraged these pilots to study risks and costs, deployment mechanics, use and usage issues, integration, and cultural and political dynamics. Throughout this process, we worked with several teams to create a

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

165

framework for evaluating social media and a multi-fold plan for assessing the value of business networking for the enterprise and for the individual.

4 Research Platform 4.1 Tool Selection and Implementation We wanted to create a trusted environment with fine-grained access control and a secure, yet easy, invitation model, all of which had to meet our information and operational security requirements. We required a single platform to allow our employees to engage with each other internally but also to connect and collaborate externally without having to go to another site. The latter proved far more difficult, as no commercial tool supported the single-platform approach to span a firewall. For user adoption, the tool had to be easy to use and require little to no training. Ideally, we wanted it to be similar to popular social tools like Facebook that many of our employees were already familiar with. We were also looking for a low cost solution: an inexpensive tool with no licensing costs to accommodate an unknown and unlimited number of external participants. Finally, in order to support our experiments and allow us to use the tool as a research platform, the tool had to be fully customizable, extensible, measurable, and easy to integrate with other applications. After an extensive market survey, we selected the open source Elgg [25] platform as the basis for our prototype. Elgg’s fine-grained access control model suited our unique requirements for sharing both inside the company and outside, and the ability to host it on our own servers helped meet our information security requirements. Elgg also provided enough out-of-the-box functionality and many of the key features that we wanted. In addition, Elgg was flexible enough to customize to our environment and our needs; its plug-in model allowed for extensibility without modifying the core, and its social graph API made it easy to integrate with other applications and services. An advantage of using open source software for research was that it gave us access to the community of developers; we were able to leverage the contributions of others and share our own plug-ins and enhancements. Through this exchange, we developed a good working relationship with core Elgg developers who were responsive to our needs and concerns. Finally, using an open source platform required no software purchase and no licensing costs, resulting in a solution with no initial investment. Using an agile method of iterative and incremental development, we designed and implemented a working prototype. We focused on creating a trusted environment: we tightened but simplified the permissions model, added new access control levels, encrypted communication via HTTPS, removed public access, and made the invitation mechanism more secure. We also worked on enterprise enhancements: improving the usability and integrating with single sign-on, active directory, email, our intranet portal, and other prototypes. Lastly, we wrote an event listener to log activity for evaluation and research. Our business networking research platform, named Handshake, was first prereleased internally to the corporation and then later made available outside the firewall. Initially, we invited and nurtured several pilot teams to build their communities on Handshake and then used them to drive our early requirements, but

166

L.E. Damianos, D.L. Cuomo, and S. Drozdetski

our user base grew rapidly through word-of-mouth. We encouraged feedback from our users and were able to address concerns quickly. Our agile development methodology allowed us to implement, test, and deploy new features rapidly. 4.2 Handshake Overview and MITRE Customizations Handshake [26] is open to all MITRE employees and invited external participants for connecting, communicating, and collaborating in a trusted environment. Handshake members can form explicit relationships with other Handshake members, create their own profile, and establish groups. MITRE staff must explicitly invite external participants to join Handshake, and these external partners can in turn establish connections with other Handshake members and join groups. All content in Handshake has its own access control settings, so that content owners can regulate access. Awareness of activity on Handshake is provided through activity streams and customizable email or site alerts. The rest of this section describes some of the MITRE-built features and specific customizations for our environment. Access. MITRE employees access Handshake by single sign-on and MITREid (based on OpenID) from inside or outside the firewall. Handshake is closed to the general public, but MITRE employees may invite external partners into Handshake via an authenticated email-based invitation model. External members currently may not invite others into Handshake. Security Features & Permissions. Handshake utilizes a MITRE identity model to distinguish MITRE employees from other Handshake members. MITRE members are labeled as such and visually distinctive. All content in Handshake – groups, profile fields, files, wiki pages, discussion topics, blog posts, bookmarks, and images – has its own access control settings. Content owners may choose to make content accessible to all Handshake members, their connections, members of a specific group, MITRE only, or MITRE members of a specific group. Members may also create their own customized access control lists. To help prevent sharing of information to the wrong audience, all content is clearly marked with an appropriate padlock icon, either open or locked. This icon is colored to indicate the access level. Our model for group permissions restricts content within a specific group to that group’s permission level, ensuring that material is not inadvertently exposed beyond the membership of the group. As an added precaution, MITRE members see a banner as a reminder to share information appropriate to that audience. Profiles. Handshake automatically populates profiles of MITRE employees with their name, email address, and phone number; external participants are required to provide their name and organizational affiliation. All Handshake members may optionally upload a profile picture and add their job title, website, hyperlinked professional and personal interests, education, and a bio. Each profile field is access-controlled so that members can choose what to share with whom. Groups. MITRE employees are able to create their own groups around a project, topic, community of practice or interest, affiliation, hosted conference or recurring meeting, etc. Groups can be open or moderated and can be made private or accessible to larger communities. Elgg’s out-of-the-box group capability includes tools such as

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

167

discussion forums, file repositories, wiki-like pages, blogs, photo albums, and a message board. We added our own generic HTML widget so that users can post announcements, provide links to resources and contacts, or create hyperlinked lists. We have also added a group activity feed, a group tag cloud, and a group wiki tag cloud. A simple customization feature allows owners to disable group tools that are not needed. To allow for more flexibility in group management, we have implemented a multiple co-owner feature, the ability to invite members into Handshake and a specific group simultaneously, sub-group functionality, and a group restore feature (for accidental group deletion). Browsing & Searching. MITRE developers improved the metadata and tag search capability across all content, groups, and users. This enhanced search was incorporated and released to the community in Elgg version 1.7. An autocomplete feature in the search box saves users time by allowing them to select from their established connections or group memberships. Further enhancements allow for browsing of groups by membership level, recent activity, recent visits, alphabet, and creation date. We also dynamically sort users’ groups so that they can quickly access those they visit most often. Finally, we have implemented tag clouds for personal content, group content, and group wiki pages to assist in browsing and navigation. Awareness. Handshake provides widgets on each user’s home page that display recent activity – status updates, file uploads, blog postings, comments on content, new group creation, etc. Members can choose to view their own activity feed, the activity of their connections, activity of all Handshake members, and activity across their groups. Each Handshake group also has its own activity feed to keep members up to date and bring new members up to speed quickly (this MITRE plug-in has been released back to the Elgg open source community). Users can view items in the activity feed only if they are permitted access to the actual content. By default, Handshake users also receive customized email alerts to activity that occurs within their groups. The alerts contain textual content and hyperlinks to both the user who performed the action and to the group where the activity occurred. Users may elect to receive their alerts as a single daily digest, turn off notifications from specific groups, or not receive alerts at all. The email alerts provide users with lightweight access to activity occurring in Handshake. In the near future, users will be able to respond to discussion topics directly from their email. Integration & App Development. In addition to our current work of integrating Handshake with email, we have been coordinating with other teams at MITRE who are building plug-ins, applications, and services for Handshake. Recently, work has been done to integrate status updates and activity feeds with our enterprise intranet. Other teams at MITRE are working on mobile applications to Handshake, a contact recommender system for automatically suggesting connections based on users’ digital footprints [27], and an “introduction” tool allowing users to introduce people to each other. Another research group is investigating automatic profile generation. Future Customizations. We would like to improve the user experience within Handshake by enhancing navigation, browsing, searching and filtering for users, groups, and content. We would also like to implement the ability for group owners to customize groups and create their own widgets. In order to entice people to visit

168

L.E. Damianos, D.L. Cuomo, and S. Drozdetski

Handshake frequently, we are building a “Pulse of Handshake” feature where visitors can view dynamically updated content and learn what is happening, what is new, and what is getting a lot of attention. We are designing an extended permissions model around the concept of affiliated partner organizations and trusted partners; external participants belonging to an organization with a contractual relationship with MITRE will have special rights and permissions in Handshake. We are also working on implementing a lifecycle model for user management as users move from organization to organization and want to stay connected to Handshake. Finally, we will continue to explore integration with other socially enabled services and applications in the enterprise.

5 Evaluation Since the initial release in August 2009, Handshake has grown to support over 4300 users (~1000 are external participants) and 450 groups and communities. We are in the formative stages of a follow-on research project that will allow us to study adoption, motivation, incentives, and facilitation in a longitudinal evaluation of socially-enabled business models. Social media follows a long-tail model where contributions from each individual are small, but there are many individuals. Longtail effects grow rapidly with population size, and impacts are seen in large populations [28]. To conduct a proper assessment, we need time for the technology to be adopted, understood, and work practices adapted. To that extent, we have instrumented Handshake with an event logging functionality and are looking at the relationships and interactions among Handshake users, their contributions, group activity, and “silent” participation (often involving sharing Handshake content outside the tool). We are planning to employ a “use case” evaluation methodology where individual users and groups will start by stating their business goals. We will provide an orientation to the group members, and we will follow up at a later time to assess which factors facilitated or hindered the attainment of their goals. We will also extract lessons learned as a way of identifying techniques to speed adoption, remove barriers to achieving value, and institutionalize the best practices. In the meantime, we have, however, been impressed with the creativity of the MITRE staff in adopting these new ways of working. We have collected anecdotal stories of business value achieved via new socially enabled collaboration techniques from our early adopter groups. For example, a chief scientist was approached by the head of a newly formed association which brings together organizations from across the country. The two needed an easy way to share and comment on policy and related issues affecting their centers nationwide. Today, Handshake is one of their primary means of communication and collaboration: "Handshake is very simple and intuitive, and it's helping people to connect in ways they've found difficult to do using traditional methods." Other staff members are turning to Handshake for some novel problem-solving activities. For example, a MITRE senior principal scientist is helping to envision the company's workforce of the future and how it can best be supported. Using a Handshake group called “MITRE 3.0”, he reached out to over 1,000 employees to

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

169

engage them in envisioning their future needs and work practices. "A social networking platform is uniquely suited to engaging a larger swath of users on various problems. The degree of engagement is high and the interactions cut across users working in various locations and domain areas," noted the group owner. Because of low barrier to entry and wide participation, discussions on Handshake are flattening the traditional notions of company hierarchy. In one discussion thread, a user lamented that MITRE was no longer hosting show-and-tell exhibits for internal research projects. Other members from across the company contributed ideas and later decided to host a science fair. This idea arose spontaneously as a grassroots effort without any corporate encouragement. However, managers tracking the discussion offered funding and communications assistance. The science fair was wellattended, and provided an opportunity for participants to meet face-to-face. In a similar vein, the corporate communications group created the MITRE Alumni group to help identify MITRE employees in historical photograph archives, from as far back as the 1950s and 1960s. Monthly, photographs are shared in Handshake where alumni and current employees identify the subjects and the event. Handshake groups are also proving valuable for supporting technical exchange meetings, multi-organizational working groups, MITRE-sponsor project teams, selfhelp groups, new employees, and other virtual communities such as the teleworker community. For many of these Handshake communities, user profiles provide a unique way of helping members who do not have face-to-face access get to know each other better. Finally, the activity rivers for network and group activities provide a unique situational awareness view not available in our other collaboration tools. One of our users noted that Handshake was an “ecosystem” for collaboration that is leading people to discover information they were unaware of.

6 Lessons Learned Handshake provides a trusted way to connect and collaborate across organizations, geographical locations, and time. By maintaining control over the information exchanges and the membership model, MITRE can support a safe environment for its employees while inviting selected outside partners. The flexible but extensive permissions model allows users to choose what information they want to share with whom, bringing just the right subset of members to bear on the issue being discussed or the artifact being created. In creating Handshake and watching it grow, we have learned – and continue to learn – several valuable lessons. It is important not to “reinvent the wheel.” We did not want to build our own “Facebook for the enterprise” but instead focused on how we could utilize this new technology to extend the ways MITRE works and to leverage the broader community to provide better recommendations and products to our customers. We realized we would have to deal with both the early adopters and those reluctant to change their ways. The early adopters quickly grasped the power of the new platform, eagerly adopted it so they could reach a broader swath of users and partners, and pushed us to provide more capabilities quickly. We also had to market, educate and explain the new paradigm to staff who were reluctant to change their existing work practices. We spent a significant amount of time working with our

170

L.E. Damianos, D.L. Cuomo, and S. Drozdetski

internal corporation communications department to capture and publish the early success stories so these new ways of collaborating could be more easily understood and emulated. We also recognized that our users were not going to abandon their existing tool suites, so we integrated the new capabilities with tools they were already using. We observed, for example, that near-real-time notification via email alerts was one of the major contributions to the adoption of Handshake. Letting users employ channels they were most familiar with to access their business network and communities was a plus for everyone. We made sure that the tools were lightweight and required no training to use, but we did provide education (such as webinars) for new concepts. We understood the implications of culture and tried to design for the digital natives, the digital immigrants, and the digital tourists. We developed a policy for both content and user behavior. Since security was important, we built the security features early and made the permissions model easy to understand. Lastly, we learned that measuring the value of social media in the enterprise cannot be done in isolation with a handful of pilot users. The long-tail effect of establishing connections and maintaining peripheral awareness may only be realized over an extended period of time with many cross-organizational members. Acknowledgments. The authors would like to thank all contributors to the Handshake project: Jesse Ciancetta, Lena Fried, Michael Jett, Austin Kennedy, Ernie Kim, Francine Lalooses, Jon Maul, Ed Overly, Tom Read, and Justin Richer. Handshake was funded, in part, by the MITRE Innovation Program, MITRE’s internal research and development program.

References 1. Boyd, D., Ellison, N.B.: Social Network Sites: Definition, History, and Scholarship. Journal of Computer-Mediated Communication 13(1), 210–230 (2007) 2. Slideshare: Facebook Demographics (January 2011), http://www.slideshare.net/amover/facebook-demographics2011?from=ss_embed 3. Ellison, N.B., Steinfield, C., Lampe, C.: The Benefits of Facebook “Friends:” Social Capital and College Students’ Use of Online Social Network Sites. Journal of ComputerMediated Communication 12(4), article 1 (2007) 4. Wildbit: Social Networks Research Report (2005), http://wildbit.com/wildbit-sn-report.pdf 5. Dunbar, R.: Grooming, Gossip, and the Evolution of Language. Harvard University Press, Cambridge (1996) 6. Dunbar, R.: You’ve Got to Have (150) Friends. In: The New York Times, The Opinion Pages (2010), http://www.nytimes.com/2010/12/26/opinion/26dunbar.html 7. Bodnar, Kip: The Ultimate List: 100+ Facebook Statistics [Infographics], http://blog.hubspot.com/blog/tabid/6307/bid/6128/TheUltimate-List-100-Facebook-Statistics-Infographics.aspx 8. Kissmetrics: Facebook Statistics, http://blog.kissmetrics.com/facebook-statistics/ 9. Donath, J.: Signals in Social Supernets. Journal of Computer-Mediated Communcation, 13(1), article 12 (2007)

A Case Study for Exploring Business Networking for the Enterprise, Inside and Out

171

10. Boyd, D.: Friends, Friendsters, and MySpace Top 8: Writing Community Into Being on Social Network Sites. First Monday 11, 12 (2006), http://firstmonday.org/issues/issue11_12/boyd/index.html 11. Haythornthwaite, C.: Social Networks and Internet Connectivity Effects. Information, Communication & Society 8(2), 125–147 (2005) 12. Putnam, R.: Bowling Alone: The Collapse and Revival of American Community. Simon & Schuster, New York (2000) 13. Granovetter, M.: The Strength of Weak Ties: A Network Theory Revisited. In: Marsden/Lin, Social Structures and Network Analysis, pp. 105–130 (1982) 14. Donath, J., Boyd, D.: Public Displays of Connection. BT Technology Journal 22(4), 71–82 (2004) 15. Wellman, B., Haase, A.Q., Witte, J., Hampton, K.: Does the Internet Increase, Decrease, or Supplement Social Capital? Social Networks, Participation, and Community Commitment. American Behavioral Scientist 45(3), 436–455 (2001) 16. Resnick, P.: Beyond Bowling Together: SocioTechnical Capital. In: Resnick, P., Carroll, J.M. (eds.) HCI in the New Millenium, ch.29, pp. 247–272. Addison-Wesley, Reading (2002) 17. Parise, S., Iyer, B., Cuomo, D., Donaldson, W.: MITRE Corporation: Using Social Technologies to Get Connected (2011), http://www.iveybusinessjournal.com/mitrecorporation-using-social-technologies-to-get-connected 18. Gartner, Inc.: Organizations Face 5 Major Challenges Regarding Social Software (2008), http://software.tekrati.com/research/10150/ 19. Smith, J.: Updated Lists of All Companies and Regions on Facebook (2006), http://www.insidefacebook.com/2006/11/15 20. Gotta, M.: Field Research Study: Social Networking within the Enterprise (2009), http://www.burtongroup.com 21. Millen, D., Feinberg, J., Kerr, B.: Dogear: Social Bookmarking in the Enterprise. In: CHI Proceedings, pp. 111–120 (2006) 22. Gaudin, S.: Cisco’s Quad Takes Social Networking to the Enterprise. Computerworld (2010), http://www.computerworld.com/s/article/9177766/Cisco_s_Quad_ takes_social_networking_to_the_enterprise 23. The MITRE Corporation, http://www.mitre.org 24. Curry, R., Kiddle, C., Simmonds, R.: Social Networking and Scientific Gateways. In: Proceedings of the 5th Grid Computing Environments Workshop (2009) 25. Elgg open source social networking platform, http://www.elgg.org/ 26. Cuomo, D.: Say Hello to Handshake (2010), http://www.mitre.org/work/info_tech/software_collaboration/ 27. Gertner, A., Gaimari, R., Richer, J., Bartee, T.: Contact Recommendations from Aggregated On-Line Activity. Submitted to Conference on User Modeling, Adaptation and Personalization (2011) 28. Brenner, R.: Adopting Social Media: Projects, Pilots, and Politics. In: Cutter Consortium Business-IT Strategies Advisory Service Executive Update, vol. 12(13) (2009)

“Your Team Cohesion is Low”: A Systematic Study of the Effects of Social Network Feedback on Mediated Activity Luciano Gamberini, Francesco Martino, Anna Spagnolli, Roberto Baù, and Michela Ferron Human Technology Lab, Department of General Psychology, University of Padova Via Venezia, 8 – 35131, Padova (Italy) {luciano.gamberini,francesco.martino,anna.spagnolli}@unipd.it, {bau.roberto.htlab,michela.ferron}@gmail.com

Abstract. Collaborative mediated environments compete to provide visitors with social feedback, whose actual effects on visitors’ behavior is poorly known. This study considers feedback based on Social Network Analysis (SNA) and assesses whether this feedback is able to affect user activity in an online collaborative game. The results show that SNA feedback is able to modify group activity beyond a mere novelty effect, especially on the dimensions of the behavior covered by the feedback itself. The results also point to the possible role of task type in accounting for the feedback effect on behavior. Keywords: Social Network Analysis, Social Feedback, Computer Mediated Communication, Online Collaborative Games.

1 Introduction The availability of digital technologies to communicate and collaborate has continuously progressed over the last three decades [1]. A recent result of this development is social networking services implemented on websites and in other mediated environments, which turn users’ network from a latent, hidden infrastructure into an explicit, visible social object with its own value. Models and algorithms, elaborated by Social Network Analysis (henceforth, SNA) [2],1 are usually adopted to extract the structure and properties of a social network; several applications use them to make interactions visible or create social statistics [3], namely as a sort of feedback [4]; however, the effect of this information on user behaviors has seldom been evaluated (with the exception of a few studies; e.g., Morris [5]). The current study aimed to fill this gap by investigating, systematically, the effect of SNA feedback on specific dimensions of mediated activity. 1

For instance, the popular blogging service Technorati provides information on the “popularity” of a blog by calculating the number of links connected to it.

A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 172–181, 2011. © Springer-Verlag Berlin Heidelberg 2011

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

173

Previous studies using SNA-based feedback have found that they are effective in sustaining and increasing user activities [6,7], especially when information is perceived as correct and trustworthy [8]. The current study extends these results by adding two feedback types (density and centralization). These added SNA indices are at the group level (i.e., describing the status of the whole network with respect to a certain dimension) and literature has reported differences when feedback is administered at the group versus individual level [9,10]. Thus they are worth a specific investigation. The first hypothesis served as a preliminarily check to determine whether the newly added conditions could trigger the same effect observed by Martino et al. [6], namely an increase in the number of messages exchanged in the group. The explanation for that effect was that feedback on communication made more relevant the interactional dimension of the activity and then encouraged participants to invest more in this dimension by increasing their behavioral engagement. The number of messages exchanged in the group was taken as an overall, generic measure of this mutual behavioral engagement. H1: The provision of SNA feedback on communication increases the number of messages sent from every player to the other team members with respect to control groups with no feedback provision. The second hypothesis is related to the stability of this effect across subsequent activity sessions, discarding the possibility that the feedback effect is merely due to its novelty. In addition, Martino et al. [6] found that feedback, based on reciprocity, had less stable effects than did feedback based on centrality. As an explanation, in [6] it was hypothesized that some fatigue emerged from the previous session and made subtler aspects of communication (such as those needed in ensuring symmetry between all exchanges) more difficult to control. Thus, in the current study we tested if this explanation would hold true for group feedback regarding communication symmetry: H2: Feedback density increases communication activity in all game sessions, whereas feedback on centralization has a smaller or no effect in the final game sessions. The third aspect considered was the specificity of the feedback effect because, in the literature, it has been assumed but not directly investigated. In fact, there are only a few studies that have systematically evaluated the effect of augmenting groupmediated interaction via the provision of some type of feedback about the communication activity [11,12,13]. These studies have tested the effectiveness of such feedback on the aspects of group processes covered by the feedback itself (which might increase or decrease according to user’s position in the group), somehow presupposing that feedback worked by affecting the specific areas of behavior on which it provided information. In the current study, this assumption was tested directly, since effectiveness and stability were assessed on several dimensions of communication with respect to the specific information provided by the feedback. H3: A specific dimension of communication is always affected by the feedback on that same dimension. More specifically, the value of degree centrality increases after

174

L. Gamberini et al.

the provision of degree-centrality feedback; the value of reciprocity increases after provision of reciprocity feedback; the value of density increases after provision of density based feedback; the value of centralization increases after provision of centralization feedback.

2 Method Participants were grouped into teams that played an on-line treasure hunting game (based on the Crossfire platform) for four consecutive sessions. Participants communicated via dyadic chat (through Skype®2) and their message exchanges were logged and measured. Prior to each new game session, participants in the feedback conditions viewed an SNA index that measured a property of the communication activity of the group from the previous game session. Type of feedback varied between subjects and could cover Centrality, Reciprocity, Density, or Centralization; no feedback was provided in the Control Condition. The two new feedback conditions added in the current study involved 80 voluntary participants with similar characteristics of [6] (age M = 23.05, SD = 2.71; 37 men, 43 women). They were recruited by asking university students met at the university premises; people who agreed to participate where then randomly assigned to the different teams. The overall design was composed by these new conditions in addition to the conditions of the previous study. In total, the design involved 200 participants (86 men and 114 women, age: M = 23.66, SD = 4.14) who were distributed among five between subjects feedback conditions. Each condition included four teams (40 participants). 2.1 Feedback The types of feedback displayed are shown in Figure 13. “Centrality” describes how extensively a social actor is involved in social relationships with other actors in the communication network [2]. In the current investigation, centrality was visualized as a network; nodes represented actors and lines represented the ties between actors. Nodes with a higher centrality occupied a more central position, had a wider diameter, and a darker tone of green (Figure 1a). “Reciprocity” is the tendency of an actor to reciprocate the action of other actor and can be considered as a measure of cooperation [14]. The event in which this symmetry was measured was the initiation of a dyad communication exchange. This event was called “Thread Starting Request” (TSR), where threads were temporally 2

3

We choose to use Skype because it is more intuitive than the communication system embedded into Crossfire. Some informal pilot sessions were conducted to determine whether participants could easily use these two programs together and no difficulty was observed (nor were any difficulties complained about during the actual experimental sessions). Only dyadic communication was allowed in order to make the recipient of the message and then the analysis of the ties in the network more straightforward (of course, participants were allowed to open more chats simultaneously to keep in touch with more than one teammate at a time). While all indices are mentioned in this paper with the name they have in SNA literature, in this study some were re-named with more intuitive labels to better convey their meaning to common users.

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

175

bounded sequences of messages on a single topic [6,15] and a TSR was a message that potentially started a new thread. Participants reciprocated if both individuals in a dyadic chat attempted to start a new thread of messages at some point during the session. The representation of this index (Figure 1b) consisted of a graduated scale; two hands were joined when reciprocity between a certain actor and the rest of the group was at a maximum and the hands were separated when there was no reciprocity.

Fig. 1. Examples of the four types of feedback used in the various feedback conditions: (a) degree centrality, (b) reciprocity, (c) density, (d) and centralization

“Density” measures network cohesion [2]. The representation used in our study (Figure 1c) consisted of a graduated meter with an arrow pointing to the position on the meter that corresponded to the index value. “Centralization” measures the tendency of a network to be more or less centralized around a few actors [2]. The representation used in our study was similar to the one adopted for the density feedback, except for the two icons representing extreme values (Figure 1d). Formulas for calculating indices follow Wasserman & Faust [2], with the only exception of centralization (where the final value was achieved by subtracting the value obtained in the classic formula from 1, which maximized the index value when the network was decentralized). All index values varied from 0 to 1. 2.2 Task and Procedure Each participant was represented by an avatar in a large team-shared virtual world. The world contained hidden goblets (the treasure), different locations participants were free to explore, clues to help to find the goblets, and virtual food to keep the avatars alive. Participants acted in the virtual environment using a mouse and keyboard. To present the activity as a cooperative task, participants were informed that the winner would be the team that found the most goblets for all sessions. Participants who belonged to the same team met in a computer room. After signing an informed consent, participants received instructions on the game and played a 10minute training session to familiarize themselves with the game and the controls. Each experimental session lasted 20 minutes with a 25-minute break in between sessions. Feedback was displayed on the participant’s monitor at the beginning of

176

L. Gamberini et al.

each new session for the feedback conditions; a standard explanation of the feedback visualization and the index value was provided to all participants; no evaluation was implied in the explanation that could encourage the achievement of a specific index value. 2.3 Data Collection and Analysis The data collected were obtained from the chats that occurred during the game sessions. Of these data, the overall number of messages exchanged was calculated as well as Degree Centrality, Reciprocity, Density, and Centralization. The statistic procedure used to test the hypotheses varied according to the nature of the dependent variable; namely whether it measured data at the group or individual level. The former case applied to centralization and density and a mixed design ANOVA was performed, with Condition and Session as main factors. Individual indices included the number of messages, degree centrality, and reciprocity. For these measures, Multi Level Models (MLM) were adopted to account for the non-independency of scores for each individual in the group [16]. These models were implemented using SPSS Mixed Procedure [17], with Session and Condition as fixed effects and group belongingness as a random effect. Effect size was calculated using the d coefficient, as suggested by Williams et al [18]. Post-hoc analyses were conducted by comparing 95% confidence intervals4 for the mean value of groups in each condition and along the four gaming sessions.

3 Results 3.1 Effect on the Number of Messages Exchanged It was assumed that the number of messages exchanged in the group would measure the general feedback effect on communication activity. The results are illustrated in Figure 2. The analysis yielded significant results for the main effects of the factors

Fig. 2. Number of messages exchanged in each experimental condition in the four sessions of game. Error bars correspond to 95% Confidence Intervals. 4

Confidence Intervals were calculated using SPSS EMMEANS procedure for MIXED MODELS (SPSS v. 16).

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

177

“Condition,” F(4, 192.281) = 6.572, p < .001, d = .49 and “Session,” F(3, 222.75) = 10.331, p < .001, d = 0.47 and for the interaction between Session and Condition, F(12, 222.775) = 3.673, p < .001, d = .30. During the first session, the comparisons of estimated marginal means did not reveal any significant differences among conditions. In the second and third sessions, namely after the feedback provision, fewer messages were exchanged in the control condition compared to the feedback conditions; feedback conditions did not significantly differ. In the fourth session, the number of messages remained higher in the centrality and density conditions; however, it did not differ from the control group in the centralization and reciprocity conditions. 3.2 Effects on Degree Centrality Results for degree centrality are displayed in Figure 3. The analysis yielded significant results for the main effects for “Condition,” F(4, 194.431) = 19.764, p < .001, d = .22 and “Session”, F(3, 249.264) = 52.928, p < .001, d = .36 and an interaction effect for “Session” and “Condition,” F(12, 249.264) = 2.013, p = .024, d = .07.

Fig. 3. Mean values of degree centrality in each experimental condition and in the four sessions of the game. Error bars correspond to 95% Confidence Intervals.

During Session 1, all feedback conditions except for reciprocity were higher than in the control condition. From the second session onwards, all feedback conditions differed from the control condition. The initial difference with control, due to coincidence, was maintained steadily during all sessions. However, given that groups in the feedback condition differed from the controls since the first session, it is not possible to attribute the differences in the subsequent sessions to feedback alone. 3.3 Effects on Reciprocity The effects of all types of feedback on reciprocity are shown in Figure 4. The analysis yielded significant results for the main effects of “Condition,” F(4, 186.873) = 15.673, p < .001, d = .28 and “Session,” F(3, 290.570) = 11.140, p < .001, d = .67 and for the interaction between “Session” and “Condition,” F(12, 290.570) = 2.462, p = .004, d = .25. Comparisons of estimated marginal means showed no significant difference

178

L. Gamberini et al.

during the first session. In the second and third sessions, no differences were found between any feedback condition and, excluding Centralization, all conditions yielded a higher reciprocity value than did the control condition. Differences among feedback and control conditions remained in the fourth session, where reciprocity scores were also higher with reciprocity feedback compared to the other types of feedback. Centrality and density feedback were effective but less so than reciprocity feedback in the fourth session. Feedback specificity seems to have played a role here. In addition, since reciprocity was effective on reciprocity feedback in the fourth session, this aspect of communication was not affected by fatigue as hypothesized in [6]. This notion influenced H2 and will be discussed in the conclusion.

Fig. 4. Reciprocity values in each experimental condition and in the four sessions of the game. Error bars correspond to 95% Confidence Intervals.

3.4 Effect on Density Regarding the effect on density, a graphical representation of the mean values of density and how they varied across the different conditions is displayed in Figure 5. This analysis yielded significant results for the main effects of “Condition,” F(4, 15) = 5.985, p = .004, η2 = 0.51 and “Session,” F(3, 45) = 18.280, p < .001, η2 = 0.61. There was no significant interaction between “Session” and “Condition,” F(12, 45) = 2.462, ns, η2 = 0.068. As was the case for degree centrality, it is not possible to attribute any subsequent difference to the feedback and was not possible to test this hypothesis.

Fig. 5. Density values for each condition in the four sessions of the game. Error bars correspond to standard errors.

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

179

3.5 Effect on Centralization The results for centralization are displayed in Figure 6. The analysis yielded significant results for the main effects for “Session,” F(3, 45) = 4.826, p = .005, η2 = 0.20. There were no significant main effects for “Condition,” F(4, 15) =.583, ns, η2 = 0.12 and no significant interaction between “Session” and “Condition,” F(12, 45) = .943, ns, η2 = 0.16. Therefore, it seems that this value decreased from the first to the third session, regardless of the presence of feedback.

Fig. 6. Centralization values for each condition across the four sessions of the game. Error bars correspond to standard errors.

4 Discussion and Conclusion Finding the proper cue to influence how users communicate in a mediated environment has always concerned scholars in human-computer interactions [20]. The study reported here examined the effects of a class of augmenting cues (i.e., SNA feedback) on group communication. The effect on the number of messages exchanged (H1) was confirmed for all types of feedback. Stability (H2) was also confirmed: the effect persisted until the third session and afterwards decreased for reciprocity and centralization feedback. However, the explanation for this decrease is not as we originally hypothesized. Specifically, reciprocity feedback did not lose its effectiveness on reciprocity behavior, only on the general amount of messages exchanged. Thus, reciprocity cannot be considered as generally prone to fatigue as we thought in H2. Rather, it is possible that reciprocity feedback decreased its effect on the exchange of messages because it is not a generic dimension of communication, but a specific one; it is more related to message quantity than to message direction. In fact, feedback based on message quantity (centrality and density) was able to maintain longer effects. This result suggests that specificity affected (H3) reciprocity behavior, as hypothesized, but it also affected message quantity. Specificity effects were not found on centralization, degree centrality, or density. Since there was an initial difference due to sampling error in degree centrality and density among conditions, despite homogeneity in demographics, we can only affirm that differences found in the first session for these indices were maintained consistently during all sessions, without knowing if they were due to the feedback or to other

180

L. Gamberini et al.

differences between the groups. Regarding centralization, there may by other reasons for the lack of effect, which was probably connected to the nature of the task. Centralization concerns the level of hierarchization in the communication structure of the group. Because the goal of communication in this study was to circulate information, participants were unable to predictable who found this information; therefore, centralizing communication was not beneficial to the task. This finding can explain why centralization decreased, regardless of feedback, and suggests that the task can play an important role in defining what feedback information is used and how. In conclusion, each feedback affected at least some dimensions of communication and did so steadily. Therefore, SNA-based feedback can profitably be integrated in electronic communication systems to support user awareness of group interactions. The specificity of the feedback, with respect to the targeted behavior, plays a role in determining its effectiveness. This finding has important consequences in designing a feedback service. In case that service is not customized on a specific activity, it is recommended that designers provide a repertoire of different feedback types that users can chose from based on the needs of their activities. Finally, further studies are needed to examine the role of the task in mediating the feedback effect. Acknowledgments. The study reported here is partially funded by the PASION project (Psychologically Augmented Social Interaction over Networks, reference number 27654 PASION, EU IST program). All authors contributed equally to this work.

References 1. Schmidt, K.: Cooperative work and coordinative practices. Springer, London (2011) 2. Wasserman, S., Faust, K.: Social Network Analysis. Theory and methods. Sage, New York (1994) 3. Isbell, C.L.I., Kearns, M., Singh, S., Shelton, C.R., Stone, P., Kormannn, D.: Cobot in LamdaMOO: An adaptive social statistics agent. Auton. Agents and Multi-Agent Syst. 13, 327–354 (2006) 4. Kluger, A., DeNisi, A.: The effects of feedback interventions on performance: a historical review, a meta-analysis, and a preliminary feedback intervention theory. Psychol. Bulletin 119, 254–284 (1996) 5. Morris, M.E.: Social networks as health feedback displays. IEEE Internet Computing 9, 29–37 (2005) 6. Martino, F., Baù, R., Spagnolli, A., Gamberini, L.: Presence in the age of social networks: Augmenting mediated environments with feedback on group activity. Virtual Real 13, 183–194 (2009) 7. Kirman, B., Lawson, S., Linehan, C., Martino, F., Gamberini, L., Gaggioli, A.: Improving Social Game Engagement on Facebook through Enhanced Socio-Contextual Information. In: Proceedings of ACM CHI 2010, pp. 1753–1756. ACM Press, New York (2010) 8. Gamberini, L., Martino, F., Scarpetta, F., Spoto, A., Spagnolli, A.: Unveiling the structure: Effects of social feedback on communication activity in online multiplayer videogames. In: Schuler, D. (ed.) HCII 2007 and OCSC 2007. LNCS, vol. 4564, pp. 334–341. Springer, Heidelberg (2007)

A Systematic Study of the Effects of Social Network Feedback on Mediated Activity

181

9. DeShon, R.P., Kozlowski, S.W.J., Schmidt, A.M., Milner, K.R., Wiechmann, D.: A multiple-goal, multilevel model of feedback effects on the regulation of individual and team performance. J. Appl. Psychol. 89, 1035–1056 (2004) 10. Ilgen, D.R., Fisher, C., Taylor, M.S.: Consequences of individual feedback on behavior in organizations. J. Appl. Psychol. 64, 349–371 (1979) 11. DiMicco, J.M., Hollenbach, K.J., Pandolfo, A., Bender, W.: The impact of increased awareness while face-to-face. Human-Computer Interaction 22, 47–96 (2007) 12. Losada, M., Sanchez, P., Noble, E.E.: Collaborative technology and group process feedback: their impact on interactive sequences in meetings. In: Proceedings of the 1990 ACM Conference on Computer-Supported Cooperative Work, pp. 53–64. ACM Press, New York (1990) 13. Zumbach, J., Schonemann, J., Reimann, P.: Analyzing and supporting collaboration in cooperative computer-mediated communication. Paper presented at Learning 2005: the Next 10 Years! Taipei, Taiwan (May-June 2005) 14. Mui, L., Mohtashemi, M., Halberstadt, A.: A Computational Model of Trust and Reputation. In: Proceedings of the 35th Hawaii International Conference on System Science, p. 122. IEEE Computer Society, Washington (2002) 15. Yates, J., Orlinowski, W.J., Wörner, S.L.: Virtual organizing: using thread to coordinate distributed work. Working papers, pp. 4320–4323. MIT, Sloan School of Management (2003) 16. Freeman, L.C.: Centrality in Social Networks: I. Conceptual clarification. Soc. Netw. 1, 215–239 (1979) 17. Kenny, D.A., Mannetti, L., Pierro, A., Livi, S., Kashy, D.A.: The statistical analysis of data from small groups. J. Pers. Soc. Psychol. 83, 126–137 (2002) 18. Peugh, J.L., Enders, C.K.: Using the SPSS mixed procedure to fit cross-sectional and longitudinal multilevel models. Educ. Psychol. Meas. 65, 717–741 (1999) 19. Williams, D., Caplan, S., Xiong, L.: Can you hear me now? The impact of voice in an online gaming community. Hum. Commun. Res. 33, 427–499 (2007) 20. Oulasvirta, A., Petit, R., Raento, M., Tiitta, S.: Interpreting and acting on mobile awareness cues. Hum.-Comput. Interact 22, 97–135 (2007)

E-Business Solutions in the Cable TV Industry Viorica Harrison and June Wei College of Business, University of West Florida, 11000 University Parkway, Pensacola, FL 32514 [email protected], [email protected]

Abstract. This paper aims at developing implementable e-business solutions for the cable TV industry to accelerate the information technology and information system adoptions; and thereby, increase value for the operations in the industry. Specifically, an electronic value chain is developed to show how to improve value by using information technology and information system. Then, a set of implementable solution items are developed based on this electronic value chain model. Finally, recommendations are provided on how to implement these solution items. Keywords: Value chain, cable TV industry, electronic business.

1

Introduction

Although many large cable TV providers are fully utilizing the Internet and transforming their business operations to include complex e-Infrastructures and IT applications, other providers are slow to move to e-business solutions because of high costs, risks, business reengineering, and failure to recognize the future value shifts toward online television viewing [2]. Cable TV industry providers that neglect to exploit e-business solutions are not aligning their information systems (IS) with their business and organizational strategies. Since IT is integrated in nearly all business aspects, as technology develops, companies must change IS quickly to meet market demands and compete effectively. Cable TV providers, therefore, should use scalable e-Infrastructures for e-business solutions to increase time-to-market speed, expand business functionality, lower costs, improve customer satisfaction, and increase revenues[3] [4] . As more customers migrate to the Internet and wireless devices for their television viewing, the cable TV industry will encounter considerable reformation. Cable TV providers that do not adopt sufficient information technology applications and solutions in their supply chain will encounter massive revenue losses, similar to what the music industry experienced at the turn of the 21st century, before iPod and iTunes [2]. As it stands now, the abundance of technology and Internet programming are already forcing cable TV providers to offer premium online TV content to their existing customers to keep them from accessing online shows through free Internet sites, such as YouTube and Hulu [1]. Since the US cable TV market is a mature and heavily saturated industry, increased free television electronic content and online A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 182–187, 2011. © Springer-Verlag Berlin Heidelberg 2011

E-Business Solutions in the Cable TV Industry

183

viewing could lead to decreased national cable TV industry growth and increased industry competition [8]. Comparing with other industries, IT adoptions in the cable TV industry are still slow. Increased online television viewing may bring about many operational challenges resulting from increased consumer broadband traffic, limited provider bandwidth, and network system bottlenecks, which will cause significant delays of econtent delivery [9][10]. The purpose of this paper is to develop an electronic value chain model for the cable TV industry, decompose this model into a set of implementable solutions for increasing value for the cable TV industry, and provide recommendations for e-business solution implementations. The remaining of this paper is organized as follows: Section 2 presents a value chain model for the cable TV industry. Section 3 develops a set of e-business solutions based on the value chain model. Section 4 provides discussions, recommendations, and conclusions.

2

Electronic Value Chain Model

Fig. 1 presents a value chain model based on the utilization of IT in the cable TV industry. This model was developed based on Michael Porter’s value chain with five primary activities (inbound logistics, operations, outbound logistics, marketing and sales, and services) [12]. Inbound logistics involves receipt of programming content, equipment, inventory storage and management. Cable TV operators can receive, track, and manage video inventories, such as video-on-demand VoD and pay-per-view PPV, online by processing incoming video libraries in batches [2]. For live content delivery, real-time streaming is used and, therefore, the content simultaneously goes through inbound and outbound logistics to reach subscribers. For equipment inventories, cable TV providers can use just-in-time (JIT) inventory systems in order to achieve efficient inbound logistics. For example, a barcode strip can be assigned to each individual piece of television, Internet, and telephone equipment to facilitate easy e-tracking of inventory issued to technicians and contractors for installation. At the end of the day, scanners can be used to return the bar-coded inventory items to the warehouse to be placed back in stock. Operations in the cable TV industry include in-house content production and packaging, content management and protection for both television and Internet outlets, and facility operations. Cable providers package television programming received from inbound logistics into two main tiers, basic and digital. The digital packages consist of premium content and include access to VoD, PPV, DVRs, and HDTV services [1]. To remain viable and earn profits, content library management and protection becomes increasingly important, especially when incorporating ondemand and pay-per-view options with Internet TV. Also, since many subscribers utilize the Internet to manage their cable TV accounts, offering secure e-payment systems and protecting users’ stored personal data by using encryption is critical for cable TV operators. The incorporation of electronic data interchange (EDI) into facility operations assists in creating more efficient supply chain relationships. EDI technology “enables

184

V. Harrison and J. Wei

the transmission of routine business documents having a standard format from computer to computer over telephone or direct leased lines” [6]. Cable TV providers utilize EDI to improve communication between programming networks and advertising agencies to manage advertising spot schedules and build an e-infrastructure for e-business [7]. Outbound logistics in the cable TV industry involves video, data, and telephone service delivery to subscribed users. Increasingly, cable providers have been using Web-based technologies to deliver television and advertising content over the Internet. Today, e-logistics, including e-service distribution, e-content delivery, and online VoD and PPV order dispatch enable automated distribution of movies, sport and concert events, to the end-user via the Internet. Also, due to e-logistics, viewers are able to actively interact with their service providers and receive products and services at faster speeds. Marketing and sales efforts can also be enhanced by the use of technology. For the cable TV industry, e-marketing includes e-advertising, e-sales, web-based press room portals, online newsletters and e-mail usage. Besides Web portals, e-mail selling and e-advertising is very common in the cable TV industry. An investigation of the top twenty companies’ web pages revealed that each company does an excellent job in eadvertising their own products and services via their own Web pages. Customer service represents the last function within the primary activities of the value chain. Although firms are mostly concerned with cost savings and quality, time savings can ultimately be the deciding factor between keeping a customer and loosing a customer to the competition [11]. Today, cable operators must meet customers’ busy on-the-go lifestyles by providing fast, efficient, and satisfactory customer service. Utilization of technology, such as customer relationship management (CRM) software, online chat, online account management, e-troubleshooting of program and video errors, e-order confirmation of purchased programs, e-feedback, e-testing of equipment, and e-mail will increase customer satisfaction, acquisition and retention.

Fig. 1. IT Adopted E- Value Chain Model for the Cable TV Industry

E-Business Solutions in the Cable TV Industry

3

185

E-Business Solutions

E-business solutions are IT applications that are utilized in an industry’s e-value chain. This analysis paper focuses on five categories of e-business solutions: B2B (business-to-business), B2C (business-to-consumer), C2B (consumer-to-business), B2G (business-to-government), and B2I (business-to-internal). Since the cable TV industry provides video, Internet, and telephone services to business customers, governmental customers, and residential customers, some ebusiness solutions are utilized in more than just one category. For instance, ecommunication and company websites are used in both B2B and B2G categories, while more specific e-communication applications, like e-mail alerts, exist in B2C relationships. Research shows that the majority of e-commerce happens in the B2B and B2C categories since government has been somewhat slower at adopting ebusiness solutions [5]. However, presently, government e-commerce is increasing and companies can greatly benefit from incorporating e-government into their ecommerce business models. The focus of this research paper is to analyze e-business solutions present in the cable TV industry’s e-value chain model. Table 1 lists 54 e-business application items employed by the cable TV industry, summarized by e-business categories. Table 1. E-Business Solutions for the Cable TV Industry 12 Items B2B E-Business Portal

24 Items B2C 3 Items C2B E-News Portal Online Customer Feedback

7 Items B2G Cable in the Classroom

8 Items B2I E-Knowledge management and Sharing System E-Payment System Electronic Online Job E-Payment E-Human Program Guide Search/Application System Resource Management (HRM) E-Procurement Interactive E-Parental Controls E-Public Access E-mail Television Television (ITV) E-Communication E-Mail Alerts E-Military Internet Access Electronic Data Online Intranet Interchange (EDI) Advertising E-911 Service E-search Online Account Electronic EManagement Teams Communication Online Cable Ad E-Customer Employee EInventory Help / Support Company Training and EManagement Website Learning Company Website Online Security Digital Technology Online Online Clearinghouse Shopping Mall

186

V. Harrison and J. Wei Table 1. (continued)

Extranet Interface with Suppliers and Partners Online Partner Ads Wireless E-mail

Online Equipment Testing/Upgrade Online Data Storage Online Multilingual Services E-Interactive FAQ’s Video-OnDemand (VoD) Pay-Per-View (PPV) Online Customization Tools Online Outlet Locator E-Video Error Repair Live Online Chat Online Games/Music Online 3D Content

4

Discussions, Recommendations and Conclusions

The cable TV industry has gone through many technological changes in the last ten years. The introduction of innovative technologies, such as interactive television, VoD, PPV, DVR, online TV, and lately, 3D TV, has revolutionalized the cable business into a complex and dynamic industry. The cable TV industry value is shifting from traditional broadband viewing to online and wireless television viewing access. Consumers’ desires and trends toward on-the-go flexible programming and Internet information access are forcing companies to innovate and “get in front of the [technology] change or the consumers [will] threaten to leave them behind” [2]. As such, many cable TV companies currently employ various e-business solutions to improve efficiency, reduce waste, and increase information flow between activities in the value-chain with the sole end goal—to satisfy the ultimate customer. In this paper, the e- value chain model was developed to illustrate how various applications could increase traditional value creation through the use of IS technology [12]. Further, e-business solutions were analyzed at each step of the e- value chain to better illustrate linkages and relationships between various activities in the e-value

E-Business Solutions in the Cable TV Industry

187

chain, e-business applications, and five e-business groups (B2B, B2C, C2B, B2G and B2I). Clearly, if cable TV operators desire to achieve strategic competitive advantages vis-à-vis their rivals, they need to adopt as many practically useful e-business solutions as possible. Of course, adding random e-business solutions without considering value enhancement, operations efficiency, and cost reductions is not recommended. However, in light of the research and analysis provided in this paper, it is evident that adopting less valuable e-business solutions can lower a company’s status in the market place. A prime example is Armstrong Cable Services, which is ranked twentieth by NCTA in terms of US market share, and which also has the lowest e-business solutions implementation rate. Future research will focus on data collection for the top companies in the cable TV industry to see the implementation patterns of these e-business solutions; and thereby, provide suggestions and guidelines on how to improve value for the cable TV industry in general.

References 1. Amobi, T.N., Kolb, E.B.: Broadcasting, Cable & Satellite. Standard & Poor’s Industry Surveys.from Standard and Poor’s NetAdvantage (retrieved, March 15, 2010) 2. Berman, S., Duffy, N., Shipnuck, L.: The end of television as we know it. A Future Industry Perspective. IBM Institute for Business. [Value Study] (2006) 3. Cable, Internet & Telephone Providers in the US: 51322. IBISWorld Industry Report. from IBISWorld (retrieved February 13, 2010) 4. Dess, G., Lumpkin, J., Eisner, A.: Strategic Management, 4th edn. McGraw-Hill, New York (2008) 5. Gordon, L.: Business-To-Government E-commerce Procurement: Online Centers, Databases, and Tools. AllBusiness. A D&B Company (2000), from http://www.allbusiness.com/technology/software-servicesapplications-internet/10616612-1.html (retrieved April 10, 2010) 6. Krajewski, L., Ritzman, L.: Operations Management: Processes and Value Chain. Prentice-Hall, New Jersey (2006) 7. Larson, M.: Charging hard on EDI: CAB hopes to have nets, agencies on same e-based adsales system next year (cable TV). MEDIAWEEK, 14, 45. p.5(1) (2004) 8. McClellan, S., Morrissey, B.: The future of TV: flexible, interactive and measurable, TV will work a lot like, well, the Web. MEDIAWEEK 18(41), 20(9) (2008) 9. Pearlson, K., Saunders, C.: Managing & Using Information Systems. John Wiley & Sons, Chichester (2006) 10. Turban, E.: Electronic Commerce – A Managerial Perspective. Prentice-Hall, New Jersey (2006) 11. Walters, D., Lancaster, D.: Implementing value strategy through the value chain. Management Decision 38(3), 160–178 (2000); from ABI/INFORM Global. (Document ID: 254418631), (retrieved April 7, 2010) 12. Wei, J., Platt, R., Boyd, W., Jasquith, A.: Development of Standardized E-Business Solutions via E-Chain Analysis in the Digital Utility. International Journal of Services and Standards 2(2), 117–136 (2006)

Measuring the Success of On-line Communities in an Enterprise Networking Tool Lester J. Holtzblatt and Laurie E. Damianos The MITRE Corporation, 202 Burlington Road, Bedford, MA 01730, USA {holtzblatt,laurie}@mitre.org

Abstract. We look at self-forming communities on one company’s externallyfacing business networking tool to better understand factors contributing to the success of a community. We propose that there is no single measure of success of a community; success of different kinds of communities depends on a variety of factors, including the perspective of the stakeholders and the type of community established. Keywords: communities, social software, social computing, web 2.0, social media, collaboration, collaborative environments, group and organization interfaces, asynchronous interaction, collaborative computing, computersupported cooperative work, evaluation/methodology, web-based interaction.

1 Introduction In this paper, we discuss an on-going research project evaluating factors indicative of the success of groups formed within Handshake, a business networking prototype deployed at the MITRE Corporation [1]. There is a considerable amount of published literature on measures of the success of on-line groups [2], [3], [4], [5], [6]. However, most studies show that there is no single quantitative measurement that can be used to assess a community’s success [7]. Although many of the typical quantitative measurements can provide useful information on the activity of a community, none of these is necessarily a true measurement of success [8]. Success also depends highly on perspective – the perspective of the community owners, facilitators (if any), and community members, as well as other stakeholders in the community. Communities form for many reasons, but even communities established for similar purposes may have different ways of evaluating their own success [9]. Based on our initial evaluation of Handshake groups, we have similarly found that a variety of factors contribute to and are indicative of a community’s success. Furthermore, we found that different kinds of communities form within Handshake, and that measurements of success appear to depend on the type of community established. In this paper, we will describe the different types of groups established within our business networking tool and factors associated with success for these different types of groups. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 188–196, 2011. © Springer-Verlag Berlin Heidelberg 2011

Measuring the Success of On-line Communities in an Enterprise Networking Tool

189

2 Background and Motivating Problem The MITRE Corporation is a not-for-profit organization with expertise in systems engineering, information technology, operational concepts, and enterprise modernization. In addition to managing several Federally Funded Research and Development Centers, MITRE supports its own independent technology research and application development for solving sponsors' near-term and future problems. MITRE has approximately seven thousand scientists, engineers and support specialists distributed across many locations and working on hundreds of different projects in numerous domains. For solving problems, staff is expected to seek out and rely on the expertise and knowledge of technical and domain experts distributed across the company. As a result, the corporation places a high value on sharing knowledge across individuals, projects, and business units. Information has historically been shared through the use of ListServs, technical exchange meetings, internal wikis, and Microsoft Sharepoint. For communicating and collaborating with our external partners, MITRE employees have traditionally relied on email, telephone, and face-to-face meetings. Artifacts distributed via email get lost or conflicts occur when different copies of materials are edited. Telephone conversations are ephemeral, and face-to-face meetings involve travel and scheduling issues. The company also uses an external Microsoft SharePoint site as a secure document repository, but the strict authentication requirements and the lack of lightweight functionality make it difficult to collaborate using this tool. In addition to these communication and collaboration problems, individuals want to be able to establish relationships with external parties and manage those contacts as people move from organization to organization. The recent popularity of tools like LinkedIn and Facebook make connecting and tracking possible, but they are not trusted environments for conducting business and are fraught with overwhelming amounts of personal and extraneous information. To address these issues, we explored existing products - commercial technology, open source tools, and hosted platforms - to understand what functionality existed and how people were using such tools. We also performed work practice studies with existing cross-organizational teams in order to document current work practices, assess needs that were not being met, identify breakdowns and ineffective use of technology for communication and collaboration, and gather requirements for our prototype.

3 Case Study Through MITRE’s research program, a team developed and launched Handshake [1], a business networking prototype based on Elgg, an open source platform [10]. The Handshake prototype provides a MITRE-owned space for connecting, collaborating, and networking with both internal and external partners in a trusted environment as well as a research platform for exploring the value of social media in the enterprise. Handshake is open to all MITRE employees. Members can connect with each other, create a profile, and establish groups for any purpose. MITRE staff must explicitly invite external participants to join Handshake, but these external partners can, in turn, establish connections with other Handshake members and join groups.

190

L.J. Holtzblatt and L.E. Damianos

Groups may be open to all members of Handshake, limited to MITRE employees, or restricted to a named subset of members. Groups with moderated access require approval for members to join; in open groups, members can self join. Available group tools include a discussion forum, a file repository, wiki-like “pages,” blogs, photo albums, and a message board. Each piece of content within a group inherits the corresponding group permission but may also be restricted further (e.g., MITRE members of the group). A group owner may appoint one or more co-owners to have group administration rights. Owners and co-owners can add or remove members from the group, change the group access controls, edit group metadata (e.g., tags, related URLs, description, title), or delete any inappropriate content. Awareness of activity on Handshake is provided through activity streams and customizable email alerts. Email alerts provide profile information on the creator of the content, the content itself, and quick links into Handshake. At the time of the study, Handshake had been available for just over one year. There were approximately 3300 registered members, including 800 external participants, and 300 established groups. MITRE use of Handshake was voluntary, as was the creation of each of the groups.

4 Approach For this study, we looked at 243 groups that had been created within a one-year period. Most of the groups were still in the growth stage [11] and had not yet reached maturity. We did not look at newly created groups because they were too new to have established regular usage patterns. We also omitted test groups and hidden groups that had no members. Most of the groups examined were moderated and restricted to group members. On average, MITRE employees represented 87% of group membership although 12% of those groups had less than 50% MITRE employees as members with a few of them almost exclusively made up of external partners. Some groups had up to 6 co-owners, but the median was just one per group. The median size of each group was 11, with 977 members belonging to the largest group (open to All MITRE). We created a short survey that was distributed to 222 owners and co-owners of the 243 groups. We asked 6 questions about the purpose/goal/objective of the group, how they would categorize their group, whether there was a designated facilitator, how they rated the success of their group on a 5-point Likert scale, what factors they considered in evaluating the success, and what other tools they used to support this group. We gave the participants 3 days to respond to the survey. At the end of the second day, we sent a reminder to those who had not yet responded. We had 81 participants respond with their input on 128 groups (108 unique groups). For those groups with multiple respondents, we combined their responses and resolved minor conflicts by averaging scores and making judgment calls on group category. Some survey recipients chose not to participant but indicated to us that their groups had not been successful. The categories used in the survey (see Table 1) had been pre-created by classifying each group based on its description. These categories map closely to the corporate

Measuring the Success of On-line Communities in an Enterprise Networking Tool

191

Table 1. Group Categories

Category Community of Practice (COP) Community of Interest (COI) Project

Definition A group of people who share a common domain, field, or profession A group of people who share a common interest or passion, not directly related to their work A team of individuals collaborating on a funded project with a MITRE project number, or a work team for an assigned group project People comprising an area of work that encompasses different projects A task or advisory-level group of people chartered at the corporate, center, or enterprise-level, usually crossorganizational A MITRE organizational unit; e.g., a department or division A group of people who share a common past or present association with an organization such as a university, company, or professional association A group of people attending a specific event (may be a recurring event) (Survey participant as asked to define)

Program Council

Organization Affiliation

Conference / TEM Other

categories used for the internal Sharepoint site although, for this survey, we added Community of Interest and Affiliation, types of communities not found on the Sharepoint site. In addition to survey results, we collected data from the Handshake database and performed some basic calculations (see Table 2). Examining the group activity by month over time, we then classified the activity as consistent (moderated to high), consistent (low), and little to no activity. Table 2. Handshake Data Used in Study

Metrics & Calculations Group membership model Group permissions Group creation date / # days in existence # owners/co-owners # members

Notes e.g., open/moderated e.g., All Handshake, All MITRE, Group only, My Connections

This final number was pulled only at the time of the study. We were unable to ascertain member join date or whether people had joined and then left.

192

L.J. Holtzblatt and L.E. Damianos Table 2. (continued)

% MITRE members # & type of group content # comments on each type of content Median/max thread depth % contributors

% commenters % lurkers Group activity by month

e.g., discussion topics, blog posts, files, pages, albums

Those who contributed original content and may also have made comments Those who commented but did not contribute original content Those who did not participate in any group activity Based on amount of content & comments

5 Initial Results and Discussion Over half the groups established on Handshake were either Projects (both internal and cross-organizational) or Communities of Practice (COP). See Figure 1. Some types of groups were more likely to be rated as successful than other types of groups (Figure 2). One possible explanation for these results is that success is dependent on the clarity of the community’s mission or goal [2]. Projects and Conferences have clear goals and objectives whereas Programs, Organizations, and Affiliations typically do not have stated tasks or objectives. The goals of COPs or COIs may be fuzzier than the goals of Projects but more clear than that of Organizations. Fewer groups had designated facilitators, but having a facilitator did not appear to be a factor of community success (see Figure 3). The overall level of group activity appears to predict the community’s success (Figure 4). Groups with a consistent level of moderate to high activity were more likely to be rated as successful. This finding also holds true across types of content in groups; successful communities were engaged in more discussions, contributed more files, and created more pages than unsuccessful ones (Figure 5). Indeed, success criteria enumerated by survey participants typically included generation of artifacts, amount of shared content, quality of conversations, and ongoing activity. However, the type of activity a community engaged in differed across group category (Figure 6). Comparing the four most frequently occurring categories: Projects, COPs, COIs, and Organizations, we see that: • Projects were more likely to create/edit pages. This is consistent with the observation that project groups work collaboratively around an artifact. • COPs and Projects were equally likely to upload files and much more likely to do so than COIs or Organizations.

Measuring the Success of On-line Communities in an Enterprise Networking Tool

193

• COIs and COPs were equally likely to post discussion topics and much more likely to do so than Projects or Organizations. However, the depth of discussions in COIs is greater than in COPs. From a separate interview study, we learned that lurking (not actively contributing to group content) may actually be an active behavior, and that lurkers may highly value a community because of what they learn. These findings are consistent with the conclusions of other researchers who view lurkers as valued participants in online communities [9]. Because of the negative connotations often associated with the term “lurker”, Preece and Schneider [12] suggest that the term “reader” be adopted to describe on-line participants in a community who read, but seldom contribute content. Farzan et al. [13] note that there are several reasons for valuing the presence of lurkers in an on-line community. Through their participation they may acquire and use new knowledge they read for beneficial purposes. In addition, lurkers may spread their new knowledge beyond the community and, through word-of-mouth- draw other participants into the community.

Fig. 1. Percentage of Handshake Groups by Category

Fig. 2. Success Ratings Based on Group Category. Bars show the percentage of groups within each category that were rated as successful, neutral, or not successful.

194

L.J. Holtzblatt and L.E. Damianos

Unfortunately, we could not adequately measure the activity of lurkers. While we have access to logins and page views, we have no way of knowing if group members were reading and forwarding the Handshake-generated email alerts. Lacking visibility into these activities, we are unable to determine their level of participation in the community. Were they engaged? Were they reading but not contributing? Were they creating conversations around group content outside the group space?

Fig. 3. Success Ratings Based on Having Designated Facilitator. Bars show the percentage of facilitated or non-facilitated groups that were rated as successful, neutral, or non-successful.

Fig. 4. Success Ratings Based on Level of Activity. For each of 3 levels of activity, bars show the percentage of groups that were rated as successful, neutral, or not successful.

Fig. 5. Success Ratings Based on Content Type. Bars show the percentage of contributed content by type for groups rated successful, neutral, or not successful.

Measuring the Success of On-line Communities in an Enterprise Networking Tool

195

Fig. 6. Content Type by Group Category. This graph shows average distributions of content type across group categories.

6 Conclusions The initial results of our study have proved interesting. We have seen that not all measurements of success apply to each kind of community; group category and type of content are distinguishing factors. The size of a community is not necessarily a measure of success. Our data also indicates that having a group facilitator does not always ensure success, and not all groups require a facilitator to achieve success. New Handshake groups are continuing to form, some existing groups are becoming more established, and other groups have faded away. (In the 4 months following the conclusion of this study, 120 new groups were created). As groups continue to change over time, we will observe the activity of Handshake groups over a longer period as part of a follow-on research study. In addition, we would like to explore lurker behavior. We believe that the percentage of contributing members of a community is not in itself a measure of success; lurkers may be contributing to the success of a community although this is more difficult to measure. Acknowledgments. We thank Betty Fisher and Ernie Kim for their help in creating the survey.

References 1. Cuomo, D., Damianos, L.: Extending MITRE’s Reach: Business Networking For & Beyond the Enterprise (2010), http://www.mitre.org/work/tech_papers/2010/10_2435 2. Majchrzak, A., Malhotra, A., Stamps, J., Lipnack, J.: Can Absence Make a Team Grow Stronger? Harvard Business Review (2004) 3. Maybury, M.: Collaborative Analysis for Information Driven Safeguards. In: International Atomic Energy Agency Symposium on International Safeguards: Preparing for Future Verification Challenges, Vienna International Center, Vienna, Austria, vol. 145 (2010). IAEA-CN-184/145 4. Probst, G., Borzillo, S.: Why communities of practice succeed and why they fail. European Management Journal 26, 335–347 (2008)

196

L.J. Holtzblatt and L.E. Damianos

5. Pugh, K.: Sustainable Communities: Top 40 CSFs for Keeping the Faith (2010) 6. Suarez, L.: Community Builders – Building and Sustaining On-Line Communities by Steve Dale (2010), http://www.elsua.net/2010/08/27/communitybuilders-buildingand-sustaining-on-line-communities-by-steve-dale/ 7. Preece, J.: Sociability and Usability: Twenty Years of Chatting Online. Behavior and Information Technology Journal 20(5), 347–356 (2010) 8. Lazar, J., Preece, J.: Online Communities: Usability, Sociability and Users’ Requirements. In: van Oostendorp, H. (ed.) Cognition in the Digital World, pp. 127–151. Lawrence Erlbaum Associates Inc. Publishers, Mahwah (2002) 9. Preece, J., Maloney-Krichmar, D.: Online Communities: Focusing on Sociability and Usability. In: Jacko, J., Sear, A. (eds.) Handbook of Human-Computer Interaction, pp. 596–620 (2003) 10. Elgg, open source social networking platform, http://www.elgg.org/ 11. Iriberri, A., Leroy, G.: A Life-Cycle Perspective on Online Community Success. ACM Computing Surveys (SCUR) 4(2), 1–29 (2009) 12. Preece, J., Schneiderman, B.: The Reader-to-Leader Framework. AIS Transaction on Human- Computer Interaction 1, 13–32 (2009) 13. Farzan, R., DiMicco, J.M., Brownholtz, B.: Mobilizing Lurkers with a Targeted Task. In: th Proceedings of the 4 Int’l AAAI Conference on Weblogs and Social Media, ICWSM 2010 (2010)

Business-to-Business Solutions for the Cosmetic Industry Manami Koblentz and June Wei College of Business, University of West Florida, 11000 University Parkway, Pensacola, FL 32514 [email protected], [email protected]

Abstract. This paper develops e-cosmetic business-to-business solutions based on strategic value chain management analysis. Specifically, it first investigates the e-business adoption in the major activities involved in the cosmetic business based on strategic value chain analysis. Then, it decomposes these activities into a set of implementable business-to-business solutions. Finally, the implementation patterns of these solutions for the top ten companies are analyzed. The results will be helpful in accelerating e-business adoption in the cosmetic business. Keywords: strategic value chain, electronic business.

1

Introduction

Cosmetics companies are getting more and more high tech in their research for effective anti-aging skin care products. Matching the leaps of technology in pharmaceuticals and biotechnology, cosmetics manufactures are getting more and more sophisticated in its ingredients, delivery, and operational system. In order to increase its efficiency and effectiveness, cosmetics companies often implement supporting information technology (IT) infrastructures. Some IT solutions have been commonly used in the cosmetics industry such as Electric Data Interchange (EDI), Enterprise Resource Planning (ERP), Material Resource Planning (MRP), Supply Chain Management (SCM), and Customer Relationship Management (CRM) [1]. Most cosmetics companies are experiencing moderate growth and are looking for the ways to support upcoming increasing volumes; thus, improving its IT infrastructure would enhance customer service and support upon the business expansion [2][5][8]. In order to develop the companies’ IT infrastructure, they need to ensure the employees’ understanding the need to change and its benefits for the companies. Normally, managers and employees are resistance to change for new systems; thus, it is important to ensure that managers and employees have proper training programs prior the implementation. Policy and procedure changes may cause reluctance among employees and lower productivity rates. Managers and leaders must ensure that the corporate culture adheres and supports to the new IT changes [6]. New IT system implementation issues include the lack of e-business skills, lack of understanding and support from employees, deficiencies in back and front-end integration, together with maintaining compatible technologies with current alliances or partners [7]. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 197–202, 2011. © Springer-Verlag Berlin Heidelberg 2011

198

M. Koblentz and J. Wei

The objective of this paper is to develop a set of implementable business-tobusiness solutions for the cosmetics industry has incurred as well as analyze the independence that the industry with IT related business systems. The remaining of this paper is organized as follows: Section 2 develops a strategic value chain based on e-business analysis. Section 3 decomposes these activities involved in the value chain into a set of implementable B2B items. Section 3 presents a benchmark for the top ten companies to analyze the pattern of the B2B implementation. Lastly, the paper concludes with discussions on how to use these implementable items with recommendations.

2

Strategic Value Chain Model

Based on Michael Porter’s value chain [7], this paper presents an e-value chain model for the cosmetic industry based on five primary and four support activities. These activities involve information technological advancements and information systems. They are designed to increase efficiency and effectiveness of the company’s core ebusiness activities together with improvement of the products and services value. 2.1

Primary Activities

Inbound Logistics – Technological advancement for inbound logistics includes an inventory control management systems such as a reorder point (ROP) system. Just In Time (JIT) inventory systems and inventory database systems are also the useful tools to reduce the level of the inventory. When the company’s inventory decreases, its warehouse costs also become lower. By cutting these costs, the company can use financial resources to produce more benefit for its shareholders. Operations – Technology can contribute to make operations more efficient. Computer-aided design (CAD) technological system has shortened the new product development phase from years to months. Electronic data interchange (EDI) is useful to transfer electronic documents or business data from one computer system to another computer system without human intervention; thus, it makes transactions and communications more speedy and accurate. Those value added systems will be necessity in order to bring more revenue for the cosmetics industry in the future. Outbound Logistics – For the outbound logistics, most of the companies’ retail stores use the point of sale (POS) system. Early electronic cash register (ECR), that has very limited function and communications capability, is mostly replaced by POS system that provides businesses with the capability to retain and analyze a wide variety of inventory and transaction data on a continuous basis. Another advanced technology is radio-frequency identification (RFID). RFID is innovated for the purpose of identification and tracking, which are more advanced technological system, yet still expensive to implement. Marketing and Sales – The companies’ website, e-mail, newsletter, press release and information portals are among the top e-business technologies used to generate more value to the company. Most cosmetics products are consumable goods; thus, the company’s website availability is particularly important for the reorders among its

Business-to-Business Solutions for the Cosmetic Industry

199

repeated customers. Many companies distribute their newsletters including the special deals and web coupon codes to promote more online sales. Also, the company’s website has a role to inform new items and provide continuous information on products. Technology is necessary to offset the future losses from competition [7]. Thus, technology becomes necessity to stay in the competition. Service – In order to serve better for its customers, many companies utilize their website to maintain customer satisfaction and provide the cosmetics products information and feedbacks. Many online stores have hassle-free return and exchange policy to increase customer satisfaction. Customer Relationship Management (CRM) system is a widely implemented system for managing a company’s interaction with customers, clients, and sales prospects. This technology is useful to organize, automate, and synchronize the business process for sales activities, marketing, customer service, and technical support. CRM system’s main objectives are to find, attract, analyze, and win new customers and reduce the costs of marketing and client services [3]. Firm Infrastructure -- E-mail system, internet and intranet, and EDI systems have contributed to improve the communication efficiency between the members of the company’s value chain. Recently, Enterprise Resource Planning (ERP) system that is an integrated computer-based application is widely implemented in many companies. ERP system manages internal and external resources, such as tangible assets, financial resources, materials, and human resources to facilitate the flow of information between all business functions [6]. ERP system eliminates redundant works within companies and improves its efficiency and effectiveness. Human Resource Management – Human resource management uses the technology mainly as hiring and training its employees. Companies also use the website for posting new jobs and providing continuous education such as eLearning system. Email and internet technology become very useful tools to shorten the communication time and decrease the costs. Technological advancement is an integral part of the HR process and creates value through implementation of supporting activities. Technology – Technology affects every aspect of the e-value chain. Technology created the value for the customers and companies by decreasing costs, saving time, and increasing a company’s effectiveness. Technology is the key components of the successful companies in many industries in the future. Procurement – Technology advancement also added value for procurement activity. Online ordering system reduced the time and created quicker process. Also, EDI system linked companies to suppliers, manufactures, distributors and retailers for providing effective communications and transactions.

3

Business-to-Business Solutions

Table 1 below describes the B2B applications that can be adapted in cosmetics industry. The B2B mainly focuses on commerce transactions between businesses, such as between suppliers, manufactures, wholesalers, and retailers via Internet.

200

M. Koblentz and J. Wei Table 1. Business to Business Solutions for the Cosmetics Industry

4

E-Business Applications in Top 10 Cosmetics Companies

Table 2 below indicates the top 10 cosmetics companies and its progress level of implementation with the proposed E-business applications. The top 10 companies are ranked based on the annual sales in 2009 in the US. All these companies are publically-traded toiletry and cosmetics companies. In Table 2, a checkmark was given if a company implements a particular B2B solution. The percentages were also shown to analyze the B2B implementation patterns. B2B E-Business applications include E-procurement, E-mail, store-to-store inventory look-up capabilities, order tracking system, electric payment system, multilingual service, extranet interface with partners, product information management, RFID technology, and E-commerce website for wholesale. All companies have implemented at least half of applications out of 10. The companies that have higher rate of implementation are P&G, L’Oreal, Unilever, Estee Lauder, and Johnson & Johnson. Applications that have lower rate of implementation are order tracking and RFID technology. RFID technology is commonly used for more expensive items such as computers rather than cosmetics products. Because of its high cost of implementation, RFID technology still has not have cost advantage in the cosmetics industry.

Business-to-Business Solutions for the Cosmetic Industry

201

Table 2. Business to Business Applications in the Top 10 Cosmetics Companies

PNG A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 Total

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

9 90%

LOR

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

9 90%

UNI

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

9 90%

AVO

BEI

EST

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

6 60%

7 70%

9 90%

SSD

KAO

JNJ

✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

6 60%

7 70%

9 90%

HEN

✓ ✓ ✓ ✓ ✓ ✓ 6 60%

Percentage (%) 100% 100% 60% 50% 100% 100% 100% 100% 0% 60% 77%

Note: PNT- Procter & Gamble, LOR - L’Oréal, UNI – Unilever, AVO – Avon, BEI – Beiersdorf, EST - Estee Lauder, SSD – Shiseido, KAO - Kao, JNJ - Johnson & Johnson, HEN – Henkel.

From this analysis, some companies have more advanced technologies implementations compared to the others; however, all companies recognize that the implementation of technology is a crucial aspect in the cosmetics industry. The cosmetics industry’s main products are obviously beauty and cosmetics products; however, in order to perform and serve better for their customers, cosmetics companies also need to focus on improvements of technological advancement and IT infrastructure. Thus, implementation of E-Business applications will become more important in the future.

5

Conclusions

Cosmetics industry, as well as the many other industries, is largely dependent on technology in every aspect of its value chain and customer chain. Since those technologies can add significant values for its daily operations, utilizing technology advancement E-Business application implementation can be a company’s core competency. On the other hand, lack of advanced technology can easily set back the company’s position in the competitive market. Porter’s value chain model categorizes technology as one of the supportive activities; however, all of main activities also use some technology in its operating. Technology is no longer just a supportive activity; it is now one of the company’s bloodline. The wise use of technology can have positive effects on a company’s operations as increasing revenue, cutting cost, providing better training for its management and employees, and increasing customer satisfaction [4]. The poor use of technology can affect those

202

M. Koblentz and J. Wei

areas negatively and hurts the company’s trusts and reputations. Technology is the key of the growth for the companies in cosmetics industry. Future research will be conducted to develop the Business-to-Consumer, Businessto-Government, and Business-to-Internal solutions. Moreover, more companies will be added for checking the e-business implementation patterns, and provide recommendations for the cosmetic industry.

References 1. Au, K.F., Ho, D.C.K.: Electronic commerce and supply chain management: Value-adding service. Integrated Manufacturing Systems 13(4), 247–254 (2002) ; from ABI/INFORM Global (retrived November 28, 2010) 2. Bare Escentuals, I.N.C.: - Form 10-K - February 24 (n.d.). Internet FAQ Archives (2010), http://faqs.org, (retrieved November 28, 2010), http://www.faqs.org/sec-filings/100225/BARE-ESCENTUALSINC_10-K/ 3. CRM strategy checklist: Planning for CRM and customer service success (n.d.). CRM Call Center information, news and tips - SearchCRM.com (2010), http://SearchCRM.techtarget.com/feature/CRM-strategychecklist (retrieved December 1) 4. Sanger, D.E. (n.d.). The New York Times - Breaking News, World News & Multimedia. The New York Times - Breaking News, World News & Multimedia (2010), http://www.nytimes.com/ (retrieved November 10) 5. Sephora Weekly Specials (n.d.), http://Sephora.com/promo/beautybonuses/index (retrieved November 28, 2010) 6. Shields, M.G.: E-business and ERP rapid implementation and project planning. Wiley, New York (2001) 7. The Economist Intelligence Unit, The e-business value chain: winning strategies in seven global industries. New York, NJ (2000) 8. Wei, J., Van Der Ende, L., Lin, B.: Customer-Focused E-Business Model for the Oil Industry. The Journal of Computer Information Systems(2009); from ABI/Inform Global (retrieved on November 29, 2009)

Online Crowdsourcing in the Public Sector: How to Design Open Government Platforms Giordano Koch1, Johann Füller2, and Sabine Brunswicker3 1

Chair of Innovation, Technology & Entrepreneurship at the Friedrichshafen Institute for Family Businesses, Zeppelin University, 88045 Friedrichshafen, Germany [email protected] 2 Department of Strategic Management, Marketing and Tourism, Innsbruck University School of Management, 6020 Innsbruck, Austria, CEO Hyve AG, Munich, Germany [email protected] 3 Competence Center Innovation Management, Fraunhofer Institute for Industrial Engineering, 70569 Stuttgart, Germany [email protected]

Abstract: The trend towards “open innovation” has revitalized firm’s interest in tapping into external innovation sources. Firms purposively open their business models to connect internal and external ideas, and to co-create value with partners and users. Internet-based crowdsourcing and co-creation platforms have changed the way how firms implement open innovation. They allow new participatory problem solving and value-creation processes. However, the current discussion on open innovation has hardly touched upon the public sector. This paper investigates if crowdsourcing platforms can be applied in the governmental context, and under which conditions. Results show that crowdsourcing may generate strong interest among citizens and may serve as source of new high quality input. However, our findings also indicate that design principles derived from open innovation projects in the corporate world may not be directly applied in the governmental context; they need to be adjusted and complemented. Keywords: open innovation, crowdsourcing, virtual co-creation platform, design principles, public management, open government.

1 Introduction So far, research on open innovation, distributed and participatory problem solving has focused primarily on the corporate world. In a corporate context the ultimate objective of open innovation is to create valuable offerings for a firm’s customers, to profit from investment’s into innovation, and to improve an individual firm’s economic performance [1, 2, 3]. While this research indicates that “unknown” outsiders can constitute an important source for innovation and value-creation, there is hardly any research dealing with the public sector and public organizations dealing with production, delivery and allocation of goods for citizens [4]. Following the origins of A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 203–212, 2011. © Springer-Verlag Berlin Heidelberg 2011

204

G. Koch, J. Füller, and S. Brunswicker

the word “public”, “public” refers to matters pertaining to people of a community, a nation, or state. “Private”, in contrast, is set apart from government as a personal matter [5]. In the lights of this difference, we claim that the public sector and governmental problems constitute an important area of research on distributed and open problem solving, what we call “open government”. Recent case examples on problem solving activities in the public sector undermine the need for openness in governmental processes. The “Stuttgart 21” initiative in Germany - a large urban development and construction project - is just one example that highlights the limitations of closed problem solving activities. The controversial discussions on the decisions to replace the existing railway station with a new infrastructure mostly located underground exemplify that state agencies should consider citizens as users of public services, who wish to be actively involved. Indeed, treating citizens as customers should be an important objective of public management reforms [4]. To bridge this gap, the following paper tackles the question whether or not internet-based crowdsourcing platforms can be applied to the public sector and under which conditions.

2 Conceptual Foundations: Online Co-creation and Crowdsourcing in the Corporate Context In the open innovation model firms purposively use inflows and outflows of knowledge to accelerate internal innovation, and to expand markets for external use of innovation, respectively [6, p.1]. When opening up to external influences, firms can involve a range of different external actors ranging from suppliers, research organizations, universities [7], customers and users [8, 9]. In fact, the potential impact to be gained from interaction with customers and consumers (and users who may not be customers) during the innovation and value-creation process has widely been recognized. It highlights that consumers and users should not be perceived purely as “value receivers” but also a “value generators” and “co-creators” [10]. However, only when external sources such as consumers and users are willing to engage in co-creation projects and are enabled to share their creative ideas, honestly state their preferences, and comment on existing concepts, valuable contributions that lead to significantly better results can be expected. A compelling and enjoyable cocreation experience is considered as an important success factor and essential prerequisite for valuable external input to co-creation projects [11, 12, 13, 14, 15]. A carefully designed co-creation platform providing engaging and immersive virtual environments is a critical factor in co-creation projects. Consequently, researchers dealt with the question of how to successfully design co-creation and crowdsourcing platforms that ignite users’ interest in participation, allow for creative collaboration, and help to build lively social networks [11]. For example, Nambisan and colleagues [12, 14, 15] suggested that experience in virtual environments subsumes four components: (1) the pragmatic experience, (2) the sociability experience, (3) the usability experience, and (4) the hedonic experience. Based on these dimensions, Nambisan and Nambisan (2008) suggest a set of practices such as the establishment of rating systems, social interaction tools, participant recognition systems, exclusive forums, process transparency, brand fests, and clean technical flows, companies should consider when designing virtual platforms [15].

Online Crowdsourcing in the Public Sector

205

Füller (2010) introduced a virtual co-creation framework from a participant’s perspective addressing five dimensions: 1) content; 2) process; 3) people; 4) motives; and 5) personal characteristics that should be considered when designing a co-creation platform [16]. Based on this framework he presents practical recommendations to design co-creation and crowdsourcing platforms concerning the offered tasks, intensity and extent of interactions, kind of multimedia-rich environment, interaction among participants, offered incentive as well as preferred interaction partners (Table 1). Table 1. Overview of key design principles of crowdsourcing platforms Design Principle Tasks

Description Provide tasks which differ in kind as well as level of complexity. Enable participants to take on different roles such as designer, evaluator or networker.

Intensity and Extent

Allow consumers to engage more often and on a continuous base. Motivate consumers to leave their fingerprint on the platform.

Tools and Multimedia-rich Environment

Provide supportive and empowering contexts. Provide an immersive but simple-to-explore environment.

Interaction among Participants

Offer platforms which encourage intense interaction among participants. Allow relationships to be established and a community to be built. Social networking functionality enriches the communication between participants. Connection to existing social networks allows the leveraging of existing relationships Give direct and honest feedback Offer additional monetary compensation or prizes related to the performance of the participants. Offer a branded platform which allows direct interaction with the company’s developer’s team.

Incentives

Partner

These co-creation frameworks and design principles for crowdsourcing platforms used in the corporate world – mostly by profit oriented companies - may offer valuable hints also for the design and management of co-creation platforms in the public sector by governmental agencies and public administration. However, they may not be directly applicable. Governments and public administration differ significantly from companies, as they face different challenges and duties. Concepts, which are relevant and get discussed in this context, are the approaches of [17], good public governance [18], civic education, and performance management [19]. Hence, it is not only a discussion on giving citizens more “customer voice” but also a discussion on how to educate the citizens, to enable possibly all of them (reach), and to establish a lasting and bidirectional communication process between the citizens and the political institutions and politicians in order to ensure democratization. As mentioned above, little is known if and how crowdsourcing platforms can be applied in the governmental context and how those should look like. Thus, our study aims to contribute to a better understanding and close this gap of knowledge.

206

G. Koch, J. Füller, and S. Brunswicker

3 Open Government Platforms in Practice: Results from a German Action Research Project 3.1 The ‘Aufbruch Bayern’ Case Study and Its Research Design We have chosen the “Aufbruch Bayern” open government initiative to analyze our research question. The initiative represents a unique open government project due to the scope, the functionalities of the platforms, the number of participants and also the output of the open government initiative (http://www.archiv.aufbruch-bayern.de/ start.php). As this project represents an “outlier” case, we chose a single case study research. As we had access to the platform, we were able to collect rich data and observations. This data provided the basis for an in-depth case analysis. We choose an action and participatory research strategy [20]; it leverages a combined approach – participatory action research [PAR, see 21]. The review of existing research on crowdsourcing shows that there is a lack of understanding on the role of crowdsourcing in the public sector. In addition, a more detailed understanding on the appropriate design and management is required. Participatory research design supports the researcher to develop and test design principles. It creates direct insights and reflections in practice. In line with action research principles, this research went through the following major phases: Planning, acting and reflection of actions. In the planning phase the problem of the open government project was identified. Then, we designed the crowdsourcing initiative based on existing findings on how to design crowdsourcing and co-creation platforms. In the action phase, the research team worked in close collaboration with the representatives of the state agency and directly interacted with the participants of the crowdsourcing initiative. During this phase, observations and experiences of the individual researchers were highly important. After completion of the acting phase, the actions were reflected. The reflections and in-depth analysis of data collected during the action phase support the adaption and refinement of existing design frameworks and principles concerning the specifics of crowdsourcing and co-creation in a governmental context. 3.2 Data Sources and Data Collection In July and August 2010, the Bavarian State Government, announced an online participation platform. With the publication of the “call for participation”, Bavarian citizens were asked to suggest their ideas, concepts and best practice cases related to three main policy areas: Family, education and innovation. The duration of the initiative was eight weeks. We implemented the crowdsourcing with an online community platform, and thus we had access to a large set of data collected via the publicly accessible platform. Data such as logfiles (visits, amount and form of activity per user), user data (usernames, sent and received messages on profiles) and information on contributions (number of received comments, number and average of received evaluations) were exported from the platform system data base (MySQL). Data such as comments, contributions and ideas were thoroughly analyzed and reflected by the research team. Observations and direct interactions with participants provided a further important source of information.

Online Crowdsourcing in the Public Sector

207

3.3 The Design of the Crowdsourcing Platform and Evaluation In our case study we designed and implemented the crowdsourcing platform in a structure process. We relied on existing findings and design principles [15; 16]. We built upon the co-creation framework and design principles introduced by Füller (2010) when designing the crowdsourcing platform [16], and further extended and enriched it. To cater for the specifics of governmental matters and problem solving, and to better analyze our research question, we chose a very comprehensive design approach. The implemented design is presented and discussed in the following: • Purpose and goal of the participation platform: The aim of the open government platform project was clearly specified. The online participation platform “Aufbruch Bayern” was the first project of a triennial campaign aiming for systematically integrating the citizens with the means of stat-of-the art communication tools, such as crowdsourcing and social media. It was set-up as a contest. The goal of the initiative was to bring the Bavarian State Government and the citizens closer together by collaboratively discussing and evaluating ideas, concepts and best practices. The concept of participation platform enabled citizen to (1) deal with political questions and challenges, (2) to develop an understanding of the given restrictions of political decision making, and (3) thus to potentially reduce the disenchantments with politics, as well as (4) to increase the perceived empowerment. Moreover, the platform was designed to address and finally reach a variety of groups, which are typically very hard to reach by the traditional communication channels, such as print media, radio or television. By applying a comprehensive recruiting strategy both, the politically interested citizens as well as adolescents, emigrants, and academics which are typically hard to get access to, were reached. • Functionality and tasks of the participation platform: In order to involve these different groups, which differentiated in their interests as well as their skills and cognitive capacity, the online participation platform offered a variety of different functionalities and tasks. Besides the rather complex task of proposing and contributing a well-developed concept, the participants were asked to upload single ideas or best practices. In order to enable participants to take on different roles, such as the above mentioned idea contributor, evaluator or networker, the platform supported the upload of different supplemental materials like written concepts, photos or even documents. Furthermore, the platform allowed participants differentiating between comments on the idea and messages on the pin wall. The latter feature can be seen as a very useful tool in order to enable the communicators and networker to keep the community vivid. It allows participants interacting independently from one’s own contribution. Members were also invited to vote for already existing contribution with an easy “thumb-up” or “thump-down” mechanism. They were also able to comment to the designs of their fellow participants. Due to the fact that all content could be explored without a registration a large number of silent visitors were able to observe the ongoing discussion (400,000 page visits versus 2,094 registered users). This supported educational objectives and awareness creation of the open government platform.

208

G. Koch, J. Füller, and S. Brunswicker

• Intensity and extent: To address intensity and extent, community members were regularly updated about latest developments on the platform. Community managers implemented the following measures to increase intensity and extent: (1) newsletters to reengage “sleepers” into an active community role; (2) an automatic news feed, when somebody commented or evaluated your idea, (3) features of the “latest participants” and “latest ideas” on the front page, which typically has the highest click-rate. These activities helped to engage visitors immediately. • Tools and multimedia-rich environment: The online platform was equipped with a comprehensive portfolio of different social media tools. A twitter channel, the “share” functionality with different social networks, and a separate Facebook fan page, ensured that user generated content could be spread easily and could trigger a “viral” process. Indeed, we observed that participants were highly engaged and also skilled in the production of “creative materials”. At the same time, they manage to drive the distribution of information due to their high web 2.0 literacy. Multimedia enrichment and interconnectivity supported the social and informal exchange of information among users across different communities and via social media, so called word-of-mouth [22]. • Interaction: From research on platforms used in a corporate context we know that people participate in crowdsourcing platforms simply because they can interact with other platform members who share similar interests [13]. When we analyzed the click-behavior of visitors and members on the platform, it was obvious that platform visitors are more interested in other members than in content. They typically explored the community members first before they looked into the ideas or concepts. To address this interaction aspect, the platform was designed to encourage intense interaction among participants, and to allow relationships and community building. Users could write messages to other members, and an up-todate social media connection allowed, encouraged and enabled social interaction. The fact that overall about 450 ideas were shared in other social networks highlights the interest in interacting with members on the platform as well as with private network friends of registered users. • Incentives: We implemented different incentive mechanisms to motivate citizens to participate. Besides intrinsic motivation of individuals, non-monetary extrinsic motives spurred the participation in the crowdsourcing initiative. The three best contributions per category were awarded exclusive prices and the chance to be part of the next governmental declaration. This selection process was executed by the jury (members of the state government), which considered the community vote in their decision. In order to motivate interested participants who lack the skills for contributing an idea, the activity of participants on the platform such as voting and commenting was also rewarded with non-monetary prizes. It is worth pointing out that profiles of politicians were accessible and also accessed very often. Indeed, numerous questions were addressed to politicians. Apparently, the direct exposure and ability to interact with politicians represents an effective incentive.

Online Crowdsourcing in the Public Sector

209

• Partner: In the private sector research suggests offering a branded platform which allows direct interaction with the company’s developer team or even prominent people, such as the star designer or CEO [16]. We transferred this design principle to the public sector and invited well known politicians to the platform, to discussions on the twitter channel and to the Facebook fan page. Overall, the platform was clearly branded as online platform of the Bavarian state government. As mentioned above, we implemented additional functions, features and management processes to cater for the specifics of governmental problem solving. They complement and expand existing crowdsourcing frameworks and design principles: • Community management: We considered community management as a crucial design factor as open government projects usually aim for a sustainable development of an active community. Thus, we emphasized community management, its visibility and also its transparency when designing the platform. Indeed, the “Aufbruch Bayern” project team implemented a very transparent and open communication and community management strategy with precisely defined rules. The team managed to motivate community members to be involved in the management of the community and decision making. This strategy should help to reduce the risk of conflict and to increase the legitimacy of the community management. Observations during the case study and analysis of the contest data also suggested that successful community management requires self-confidence and eventually the courage to resist complaints from members. • Combining online and offline events: Another fruitful tool to recruit and to drive community growth was the combination of online and offline events. The underlying logic is quite simple. By selecting highly engaged participants and giving them the possibility to meet in “real life” and further develop the ideas and concepts of the platform, especially with the support of politicians, the political institutions gave participants the respect and recognition they deserved. Our observations and analysis of data indicate that such events have a tremendous impact on the activity and motivation of the community. The challenge is to successfully connect the offline event and the platform. The “Aufbruch Bayern” initiative had a special subpage where pictures, comments and ideas of the workshops were gathered. During the offline event the community managers posted live on Facebook and used the twitter channel actively. Afterwards the workshop participants were encouraged to spread their thoughts and experiences into the community. The discussion above highlights that our design of the open government platform embraced design principles known from the corporate world and additional complementary design aspects. The platform design was comprehensive and took into consideration the specifics of “Aufbruch Bayern”, a public initiative.

4 Concluding Discussion: Successful Design and Management of Open Government Platforms Overall, the crowdsourcing project was well received and showed a high participation. Indeed, it generated a strong interest among Bavarian citizens. 2,094

210

G. Koch, J. Füller, and S. Brunswicker

participants registered on the platform in order to contribute 750 ideas, concepts or best practices and generate approximately 1,540 pages of content. The community provided 10,932 community evaluations, and 6,342 contributions to the discussion. This indicates that citizens showed a high interest in contributing to the crowdsourcing initiative. Over the period of eight weeks participants spent 760 working days on the platform which sums up to more than 364,800 minutes of residence time on the platform. From phone calls, emails and postings we know that participants spent much more additional time at home developing and discussing their ideas within the family or own circle of friends. This high commitment of time indicates the high interest of citizens to be actively involved in governmental problem solving activities. It shows that citizens should not be treated “just as tax payers” or receivers of public services but rather as active participants and problem solvers in a governmental context. The fact that more than 400,000 viewers grappled over the period of two month with the interactive dialogue-platform undermines the impact of the crowdsourcing initiative in terms of awareness creation. An analysis of the feedback data collected in the post-project survey supports this argument. On average, participants evaluated the design of the platform positive both in terms of functionality and “innovation”. While descriptive figures on the quantity of user generated content highlight a high activity, awareness towards and interest in the initiative, we also looked into the quality of the “output”. In our case study, our observations suggest that user generated content had a positive and constructive connotation. Although there was some criticism, comments were not destructive or even insulting. In addition, officials of the Bavarian State Government stated that ideas and concepts showed a high quality. They also assessed the innovation potential of these ideas and concepts as high. When implementing the action research project, we built upon Füller’s (2010) design principles for crowdsourcing in the corporate world [16]. Indeed, our reflections indicate that these design principles also matter in public problem solving. However our study also revealed that existing frameworks applied in the private sector do either over- or undervalue some design aspects, or completely lack important ones. Table 2 highlights our adapted design principles suggested for the public sector. Summarizing the identified results, we can state that design frameworks derived from research on open innovation platforms and crowdsourcing in the corporate world can support the development and implementation of crowdsourcing in the context of open government. However, for some design dimensions of Füller’s (2010) framework we identified deviating emphasis and slightly different means to best design the respect dimension [16]. In addition, we added two new dimensions, namely “community management” and “offline-events”, which should additionally be considered in future crowdsourcing projects in the public sector. Overall, our research proposes first ideas for an adapted framework to design and implement open government projects. These design principles need to be examined more rigidly in future empirical research. At the same time, they are of high practical value as they guide the design of future open innovation platform projects.

Online Crowdsourcing in the Public Sector

211

Table 2. Adapted design principles of Open Government crowdsourcing platforms Design Principle Tasks

Intensity and Extent Tools and Multimedia-rich Environment Interaction among Participants

Incentives

Description Since political issues are typically very complex and rather difficult to explore, we observed that tasks should be designed in a visual manner so that they can be addressed with photos or stories. The dimension intensity and extent did not show any specifics in terms of platform design but should be taken into account seriously. The “Aufbruch Bayern” case showed that citizens are very cautious to shelter their private sphere when they participate in open government projects. The virtual environment has to ensure this additional requirement. In a political context, the number of lengthy messages and comments was extremely high; Feedback and comments via “thumb-up” or “thumb-down” were hardly used; they don’t seem to be that attractive in governmental matters. Long an intense discussions were the dominant mode of interaction Citizens are often politically or societal motivated. These motives should be considered in addition to the recommended incentives.

Indeed it was considered to interact with a trustworthy partner. The participation of known politicians was important We added community and event management as further principle to Community the framework as it turned out to be crucial for success. Even it may Management not be seen a platform feature it has to be considered as community managers should have access to to special features in order to guide, steer, and manage the community. Interesting features are mechanisms to connect similar ideas with each other, or to mark ideas and members in order to get back to them. Statistics like incoming and outgoing links provide further interesting information for the management. Combining Offline & We further added the dimension offline & online events as policy making is characterized by in-depth and lengthy discussions on Online Events complex problems and topics. We found that a combination of an online platform with offline events (e.g. workshop with politicians) serves as a (1) motivator for participants; (2) multiplier in the context of participant recruiting; and (3) enhances the quality of the content due to offline workshops Partner

References 1. Chesbrough, H.W.: A better way to innovate. Harvard Business Review 7, 12–13 (2003) 2. Chesbrough, H.W. (ed.): Open innovation. The new imperative for creating and profiting from technology. Harvard Business School Press, Boston (2006) 3. Chesbrough, H.W.: In: Chesbrough, H.W., Vanhaverbeke, W., West, J. (eds.) Open innovation: Researching a new paradigm, pp. 1–12. Oxford Univ. Press, Oxford (2006) 4. Hilgers, D., Ihl, C.: Citizensourcing: Applying the Concept of Open Innovation to the Public Sector. The International Journal of Public Participation 1, 67–88 (2010) 5. Perry, J., Rainey, H.: The Public-Private Distinction in Organization Theory: A Critique and Research Strategy. The Academy of Management Review 2, 182–201 (1988)

212

G. Koch, J. Füller, and S. Brunswicker

6. Chesbrough, H.W.: Open business models. How to thrive in the new innovation landscape. Harvard Business School Press, Boston (2006) 7. Fabrizio, K.R.: Absorptive capacity and the search for innovation. Research Policy 2, 1–13 (2009) 8. Brockhoff, K.: Customers’ perspectives of involvement in new product development. Int. J. Technology Management 5/6 (2003) 9. von Hippel, E.: Sticky information and the locus of problem solving. Implications for innovation. Management Science 4, 429–439 (1994) 10. von Hippel, E.: The sources of innovation. Oxford University Press, New York (1988) 11. Franke, N., Piller, F.: Value creation by toolkits for user innovation and design. The case of the watch market. Product Innovation Management, 401–415 (2004) 12. Nambisan, S., Sawhney, M.: A buyer’s guide to the innovation bazaar. Harvard Business Review 6, 109–118 (2007) 13. Nambisan, S., Baron, R.A.: Interactions in virtual customer environments: Implications for product support and customer relationship management. Journal of Interactive Marketing 2, 42–62 (2007) 14. Nambisan, S., Baron, R.A.: Different Roles, Different Strokes: Organizing Virtual Customer Environments to Promote Two Types of Customer Contributions. Organization Science 2, 554–572 (2010), http://orgsci.journal.informs.org/cgi/content/abstract/21/ 2/554 15. Nambisan, S., Nambisan, P.: How to profit from a better virtual customer environment. MIT Sloan Management Review, 53–61 (Spring (2008) 16. Füller, J.: Refining Virtual Co-Creation from a Consumer Perspective. California Management Review 2, 98–122 (2010) 17. Moore, M.: Creating Public Value: Strategic Management in Government. Harvard Business School Press, Boston (1995) 18. Stowers, G.N.L.: Becoming cyberactive: State and local governments on the world wide web. Government Information Quarterly 1, 113–114 (2000) 19. Holzer, M., Kloby, K.: Public performance measurement. An assessment of the state oftheart and models for citizen participation. International Journal of Productivity and Performance Management 54, 517–532 (2005) 20. Checkland, P., Holwell, S.: In: Kock, N. (ed.) Information Systems Action Research, pp. 3–17. Springer, Boston (2007) 21. Whyte, W.F. (ed.): Participatory Action Research. Sage, Newbury Park (1991) 22. Kozinets, R.V., Wilner, S., Wojnicki, A., de Valk, K.: Networks Of Narrativity: Understanding Word-of-mouth Marketing In Online Communities. Journal of Marketing 2 (2010)

An M-Pill Framework in the Electronic Healthcare Nien-Chieh Lee, Hi Tran, Albert Yin, and June Wei College of Business, University of West Florida, 11000 University Parkway, Pensacola, FL 32514 [email protected], [email protected], [email protected], [email protected]

Abstract This paper aims at developing a mobile pills framework in the electronic healthcare by using mobile information technologies. Specifically, an electronic based framework is developed to show how to mobile information technology and information system can be adopted in mobile pills. Then, a set of usability solution items are developed based on this framework. A prototype was created to show the real implementation of an m-pill system with these important features. The findings from this paper will be helpful to managers when making decisions on m-pills development. Keywords: m-pills, electronic healthcare.

1

Introduction

Some electronic (e-) pharmacy prescriptions have been used in the healthcare system. However, mobile prescription on the nationwide scale still does not widely used. In order to offer a new way to patient, getting prescription refill quickly and conveniently, this paper will develop a new framework that will assist pharmacy in saving patient’s time, helping doctor to track doctor shopping, and preventing misread medication names. There are some proposed concepts on electronic prescription. The e-prescribing system benefits over traditional handwritten prescription. However, the system does not send electronic prescription to the patients’ mobile devices. The e-prescribing system suggests that printing out the prescription and hand it to patients, or doctor can send it directly to the pharmacy. This system has limitation as patients may not want to buy prescription immediately or want to find the pharmacy by comparing prices and selecting the lowest price [2]. Numerous articles have been written on electrical records and mobile health [1] [2] [4] [5] [8]. People are embracing and adopting the implementation of mobile technology because mobile technology brings tremendous benefits over traditional paper record. However, there is a lack of research on mobile pills. The purpose of this paper is to develop an m-pills framework with usability features. Specifically, in Section 2, a framework was developed to show the adoptions of mobile technologies used in pills. A set of features were decomposed from this framework. A questionnaire is developed to illustrate important features. In Section 3, a prototype of an m-pills system is developed to implement these important features. Section 4 presents discussions and conclusions. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 213–218, 2011. © Springer-Verlag Berlin Heidelberg 2011

214

2

N.-C. Lee et al.

M-Pill Frameworrk

M-pill framework is a mob bile prescription system framework that not only preveents doctor shoppers but also saves patients’ time in getting medication. Figuree 1 illustrates this framework. It I provides functions such as (a) Doctor uses PC to loginn to m-pills, (2) m-pills generattes authentication code including saving a copy in the mpills Database and sending the code to patient’s mobile device, (3) Patient uuses mobile device to conduct seearch for medication via m-pills, (4) Patient places an orrder on medication, and (5) Pharrmacy receives the order & transmits the estimated pickk-up time to patient’s mobile dev vice.

Fig. 1. M-Pill Framework

In Fig. 1, the doctors creeate their account with m-pills so that the doctors can use their username and passwo ord to log in to m-pills to generate electronic prescriptiions for their patients. When an a electronic prescription is generated through m-pills, mpills saves a copy of thee prescription generated and then it assigns a unique authentication code for each h electronic prescription and sends that authentication code to the patient’s mobile device. Then, the patient can log g in with his own user name which will be made up off the combination of the patien nt’s last name and last four digits of his social secuurity number. For example, Smiith2345; and password will be the authentication code for the first time the patient logs in. Once the patient is logged in to m-pills, he w will change the password to wh hatever password he wants. The patient will use the saame username (combination of last name and last four digits of social security numbber) and the newly created passsword to log in the next time. He then can view the list of prescriptions he has with the most recent prescription on the top of the list. T The patient then selects the deesired prescription. Then, m-pills input the authenticattion

An M-Pill Framework in the Electronic Healthcare

215

code corresponding to the selected prescription into m-pills and do a search on the particular type of medication prescribed. The patient is also asked to enter the search criteria to narrow the search by indicating his preferences for the pharmacies i.e. cheapest price, nearest pharmacy or pharmacy loyalty. M-pills will display the search result by listing the prices for the medication and the corresponding pharmacies according to the preference indicated by the patient. When patient makes his choice, he will be prompted to pay. The patient can either pay then with his credit card through m-pills or pay later when he goes to pick up the medication at the pharmacy. M-pills will also display an option letting the patient indicate whether the medication will be picked up by himself or by another person such as his spouse. If the patient indicates that another person will be picking up the medication for him, a message will display reminding the patient to tell that person who will be picking up the medication for him to bring ID with him as the pharmacy will need to verify the person picking up the medication. When it’s time for the patient to get refills on the medication, the patient can simply enters the authentication code again onto m-pills. Then, m-pills will display the number of refills the patient has left. When there is no more refills allowed for a patient, m-pills will display a message saying “maximum number of refills reached, see doctor.” M-pills is going to keep the expired prescription for a few years just in case the patient is unaware or forgot that his prescription has expired. To prevent the patient from abusing the refill i.e. trying to get another prescription for the same medication from another doctor, a doctor is going to do a search on the patient’s name in m-pills when he logs into m-pills. If the patient already has an unexpired prescription, m-pills will bring that situation to the attention of the doctor by displaying a warning message saying “Patient XXX already has an existing prescription, do you still wish to prescribe one?” The doctor then can decide whether or not he still wants to issue a prescription to that particular patient. The further decomposition of flows in Figure 1 produced a set of features that are presented in Table 1 below. Table 1. Decompositions of information flows in m-pills

Feature 1 2 3 4 5 6 7 8 9 10

Definition Doctor’s username & password Doctor’s endorsement Patient’s info & medication info Authentication code Formatted prescription info Encrypted authentication code Patient’s username & password Patient’s verified username & password List of patient’s prescriptions Current price info

216

N.-C. Lee et al. Table 1. (continued)

11 12 13 14 15 16 17 18 19 20 21 22

Updated price info Price info of a particular type of medication Selected prescription Search result Order info Estimated pick up time Patient’s phone number Verified estimated pick up time Location of pharmacy Verified Doctor’s username & password Amount due Order confirmation

These developed features in Table 1 were given to several medical doctors who used e-pills before for validation. Some features were modified and added based on the comments from these subjects. The final table was sent back to these subjects for verifications again. There is no further modification.

3

Prototype of M-Pill Business System

Based on the features developed in Table 1, a prototype m-pills system was developed. The m-pills system is a mobile prescription system that not only prevents doctor shoppers but also saves patients’ time in getting medication. The nine attributes of a system are components, interrelated components, boundary, purpose, environment, interfaces, input, output and constraints. The components of our system are doctor’s PC, patient’s mobile device, pharmacies’ databases, m-pills web server and all of the components are interrelated to each other. The boundary of the system is that the system can only provide electronic prescriptions, not medical records. The purpose of the system is to facilitate a means for patients to obtain his prescribed medication more conveniently using his mobile device and also to prevent doctor shopping. The environment for the system is doctor and pharmacy acceptance, government support and funding, and patients’ support. The interfaces for the system are the doctor’s PC, patient’s phone, pharmacies’ databases. The inputs to the system are the prescription the doctor creates through m-pills; the patient’s preference i.e. cheapest price, nearest pharmacy or pharmacy loyalty for the medication search; the pries for the prescribed medications from different pharmacies and the pharmacy location. The outputs of the system are authentication code, the pharmacy location as a result of the patient’s preference, the filled prescription to the pharmacy, payment options for the patient. The constraints on the system are that the patient must have a mobile device with internet capability, and that the doctor and pharmacy must also have access to the internet. In the system analysis phase, the process modeling such data flow diagrams were used to construct users’ required features in Table 1.

An M-Pill Framework in the Electronic Healthcare

217

In the m-pills system design phase, we focused on database design and user interface design. A relational database model was created using Microsoft Access based on the m-pills entity relationship diagram in Figure 1. Designing the user interface is the most important aspect of the m-pills system. There are two limitations of mobile devices: “limited attention span and the device constraint.” [7] They recommend that all interface designers need to focus on these limitations. The scholars favors simple and easy to use interface. Wu, Shu, and Lin [8] also find out that “compatibility, perceived usefulness, and perceived ease of use significantly affect healthcare professional behavioral intent.” As a result, a good interface design would solve all these issues when designing the m-pills prescription. Some key features of the m-pills system are user-friendly, meaningful functionality, and fun-to-use. Details of these features are auto text complete, windows-like display, and help menu. The Visual Basic.net was used to design and implement menu driven interface and Microsoft Access for relational database. The system was built targeting to offer userfriendly interfaces as well.

4

Discussions and Conclusions

The current paper developed a mobile pill system by considering the usability features. It first developed a conceptual framework for m-pill, and then breakdown the m-pill information flows from this conceptual model to determine what features need to be concentrated on. The findings from the current research indicate that developing user-friendly m-pill features is crucial to the success of m-pill system development. There are major findings in the current study. First, in the current research, we are trying to build the gap by trying to utilize the mobile technologies so that physicians can access patient’s medical history anywhere. Physician can access and update patients’ record instantly, reducing the times to enter the same information twice. The system also prevent misread medication name because handwritten prescription sometimes is hard to read. Meanwhile, m-pills system saves patients’ time to get medication [3]. Second, the developed usability features in the current paper help an m-pill development company when consider distinguishing its system from its competitors, and enhance the user-friendliness for mobile pill interfaces’ design. Third, m-pill systems have usability limitations including limited size, display window, processing power, and bandwidth, comparing with e-pill system [6] [7]. The m-pill system development company needs to gain popularity by considering these features. Fourth, m-pills systems have security concern, lack of technical support, and organizational support [4]. The systems lacks of technical support and organizational support because mobile healthcare are fairly new. One limitation is that this prototype does not have evaluation. Future study is to evaluate this prototype this developed m-pills system, and investigate the importance of factors influencing m-pill adoptions by developing a survey so as to quantitatively measure these features.

218

N.-C. Lee et al.

References 1. Coolest Products of 2010. Health Management Technology 31(12), 8–12 (December 2010) 2. E-Prescribing on the Rise. Chain Drug Review 32(6), 77–87 (2010) 3. Lee, Y.E., Benbasat, I.: Interface Design for Mobile Commerce. Communications of the ACM 12(46), 49–52 (2003) 4. Lu, Y., Xiao, Y., Sears, A., Jacko, J.: A review and a framework of handheld computer adoption in healthcare, vol. 74(2), pp. 409–422 (2005) 5. Medvedeff, D.: Early Experiences in e-Prescribing. Health Management Technology 24(12), 12–36 (2003) 6. Tarasewich, P.: Designing Mobile Commerce Applications. Communications of the ACM 46(12), 57–60 (2003) 7. Wei, J., Ozok, A.: Development of a Web-based Mobile Air Travel Ticketing Model with Usability Features. Industrial Management and Data Systems 105(9), 1261–1277 (2005) 8. Wu, J., Wang, S., Lin, L.: Mobile computing acceptance factors in the healthcare industry: A structural equation model. International Journal of Medical Informatics 76, 66–77 (2007)

Online Design Discussion Sites: Emerging Resource for Creative Design Moushumi Sharmin and Brian P. Bailey University of Illinois at Urbana-Champaign Department of Computer Science Urbana, IL, USA {sharmin2,bpbailey}@illinois.edu

Abstract. Online design discussion sites are a popular networking platform drawing thousands of designers from all around the globe. These sites show promise as emerging design resource by enabling designers to learn about design techniques and tools, share ideas and to seek (receive) critique to help refine their ideas, and to learn from other designers’ experience. In this paper, we report results of a study that aims to understand designers’ experience and activities in online design discussion site and how they relate to design. We have analyzed member profiles (N=12164), posts submitted by one-time posters (N=2854), and post-contents (N=1004) for a month and conducted follow-up interviews (N=5) to investigate what roles such sites play in supporting design. Our findings reveal that while designers find such sites extremely promising, ineffective representation of ongoing activities, lack of visibility of contribution, and difficulty in finding needed information hinder (even discourage) participation. We propose actionable implications that can lead into better site design. Keywords: Creative Design, Design Community, Online Design Discussion Site, Design Critique.

1 Introduction Online discussion sites provide a place for informal conversation for people with similar experience [1] as well as offer opportunities for building and maintaining relationships. Online discussion sites serve as effective sources of communication [2], information gathering and entertainment [3]. Online design discussion sites (ODDS) are a small but important subset, attracting many designers especially students and freelancers - offering an informal environment for communication and collaboration. Design researchers consider creative design to be a social process [4, 5] and creative outcomes are attributed to the discussion and collaboration among designers as it promotes divergent thinking [6]. ODDS thus show promise in supporting design creativity by allowing designers to share and receive critique on their ideas from a diverse group of designers, making design sociable. For many freelance and student designers who don’t have access to a design team or colleagues to assist in the design process, ODDS are the only source of critique. A.A. Ozok and P. Zaphiris (Eds.): Online Communities, HCII 2011, LNCS 6778, pp. 219–228, 2011. © Springer-Verlag Berlin Heidelberg 2011

220

M. Sharmin and B.P. Bailey

Due to the popularity of the online networking and socializing sites, a large number of researchers have focused on understanding collaborative content-creation [7, 8, 9] and users’ socializing behavior in this setting [10]. In addition to the facilities that any online discussion site may provide, ODDS present an opportunity to share early ideas and receive critique from a diverse group of designers – allowing the designer to consider different perspectives and thus making these sites valuable resource for creative design. We believe research focusing on understanding designers’ experience and activities in ODDS and how these ongoing activities relate to design will lead to better design support sites - enhancing the utilization of ODDS as a resource for creative design. We extend research in this area by taking a first step towards this goal – aiming to offer insight about current practices and to identify areas that need to be addressed to increase the effectiveness of such sites. In this paper, we report results from a study that investigated activities, contents, and user experience in a popular ODDS aiming to understand the role of such sites as a design resource. Our study offered several important insights - (1) requests for critique and information about design techniques account for 70% of the total activity; (2) novice (mainly student) designers utilize these sites to seek critique and to learn about design techniques and tools; (3) expert designers (mainly freelance) utilize these sites to find collaborators and to promote their work; (4) critique and information requests receive notable number of responses within a short period of time (on an average, 7.8 and 5 replies respectively within the same day); (5) contribution is highly skewed with only 6.1% members contributing 83.28% of total content; (6) contribution is a function of members’ need, rank, and experience with the site; and (7) members’ experience varies as a function of involvement with the site and the information provided on their profiles. We also offer actionable implications that can improve the design of ODDS.

2 Related Work Internet facilitates information sharing using various systems including online networks [11]. Online discussion sites are successors of online bulletin boards which created the culture of mass communication [1,12]. Bulletin boards and early discussion sites concentrated on information sharing and problem solving in specific domains such as sports, politics, religion, technical hot lines, etc. With the rapid increase of Internet users, discussion sites became a part of the online culture [13]. Recent years have seen tremendous growth in the popularity of online design discussion communities (ODDS) (e.g., core77, highend3d, about.com). ODDS show potential to support creativity in design by providing a platform for exchanging ideas, refining ideas based on critique from a diverse group of designers through open communication and collaboration. However, ODDS look exactly like any other discussion group and provide little support for identifying the ongoing design activities - hindering the effective utilization of these sites as a design resource. Little is known about the designers’ needs and experience surrounding ODDS. In our research we take a first step towards filling this gap by offering insight about existing practice and identifying areas that deserve attention.

Online Design Discussion Sites: Emerging Resource for Creative Design

221

Researchers examined the pattern of participation in the online setting, especially sites that create encyclopedic material [9], provide entertainment resources [7, 8], and/or offer social networking capabilities [10]. Work on online movie databases examined factors that influence contribution [7]. Considerable amount of work also focused on the contribution pattern in Wikipedia [14, 15, 9]. While both entertainment and content-creation communities try to motivate users to contribute by creating an opportunity for self-satisfaction, ODDS additionally offer an opportunity to enhance the quality of designs by receiving critique on ongoing designs from a diverse group of designers. The difference in ongoing activities and users’ expectation makes it difficult to directly apply lessons from other communities to encourage effective participation in design discussion sites. Our work attempts to extend existing research on online communities by offering deeper understanding about ongoing activities and factors that influence designers’ experience in these sites.

3 Methodology The purpose of our research was to gain insight about activities and motivations for participation in ODDS and whether and how they relate to design. Our study combined both quantitative and qualitative research methods to investigate the activities and learn about user experience of core77, a thriving online design discussion site [16]. For our study, we have chosen core77, but we believe that our findings are generalizable to other ODDS. We have collected and analyzed member profiles (N=12164) along with post history (rank, number of posts, number of replies) to examine participation pattern and to identify factors that influence members’ experience in the site. We have also analyzed all contents posted in a month (N=1004, 177 threads receiving 827 replies) to investigate types of posts and their relation to different design activity. Additionally, we have analyzed all posts (N=2854) submitted by the one-time posters (members who posted just once) to identify underlying factors, if any, that motivated the only post but hindered or discouraged further active participation. To better understand members’ experience, we have conducted semi-structured interviews with five core77 users who have varying level of involvement with the site - ranging from administrator to semi-active users. See table 1 for a sample of the interview questions. Table 1. Sample of the questions asked during the interview Participation What factors influence your participation in the site? Does your participation pattern change with time? What factors influence this change? Motivation What motivates you to contribute in the site? From your perspective, what are the potential benefits of active participation in the site? Challenge From your perspective, what are the main obstacles that hinder active participation?

222

M. Sharmin and B.P. Bailey

4 Study Results 4.1 Post Categories – Conversational, Design Technique and Information, Informational, and Critique Designers participate in ODDS for a variety of reasons, including informal conservation, query about design tools and techniques, seeking critique about ongoing ideas or just sharing information about design events. In core77, there are 21 different topic categories from which we identified 12 most active categories, such as general design discussion, sketching, portfolio, etc. in terms of number of posts and views. Analysis of the posts indicated that posts related to critique and design technique dominate 10 of the 12 most active categories. Overall, requests for critique and question regarding design techniques and tools accounted for over 70% of the posts in the 12 category studied. This clearly indicates the predominant utilization of these sites as a resource for creative design as opposed to a place for informal networking. We have analyzed the post-contents to identify types of posts and whether and how they relate to different types of design activity. Content analysis revealed that posts can be broadly categorized into four types – (i) Conversational: Informal experience sharing - targeted to initiate discussion (e.g., good examples of “design language”); (ii) Design Technique and Information: Query seeking information on tools and techniques (e.g., How do I do good transition surfaces?); (iii) Informational: Posts containing links of design events, jobs, blogs, or articles (e.g., design job websites); and (iv) Critique: Posts seeking or providing critique on a design, often including a link to the project or portfolio or images along with a brief description of the project (e.g., feedback please...maybe a critique here and there). Figure 1 presents a screen capture of one such category indicating that different types of posts get represented in the same manner providing no information about the type or content.

a

b

c Fig. 1. Posts from one category of core77. Posts can be categorized as a) conversational; b) informational; and c) design techniques and tools, but the current interface represents all these similarly, providing no feedback to the members about the post types.

Online Design Discussion Sites: Emerging Resource for Creative Design

223

Table 2. Post categories and activities surrounding posts by one-time posters in core77 Category Design Tech. and Info. Conversational Critique Informational

Average (Reply) 5.06 12.23 7.82 1.58

Stdv (Reply) 10.58 18.86 7.90 2.48

Average (View) 2210.26 5458.22 2844.88 1168.11

Stdv (View) 4584.07 13570.98 2620.85 1009.13

Our interviewees expressed that not all designers are interested in all types of posts and consider it difficult to find posts of interest. Similar finding has also been reported in [17] for online Q and A sites, where some users preferred to look at informational vs. conversational posts. The high cost associated with finding posts that demand attention and can be benefitted by a members’ response limits the utilization of such sites. As a result, to ensure effective participation, it is utmost important to distinguish between different types of posts. While categorization and filtering of posts based of types will allow the members to find posts of interest effectively, categorization solely based on content is difficult and often imperfect as the original and subsequent posts can fall into multiple categories. Providing option for indicating the post type along with content-based categorization may lead to better representation of posts. 4.2 Pattern of Participation: A Function of Need, Rank, and Expertise Posts are prompted to satisfy information need: 75% of one-time posters posted on the same day they have become registered members. This behavior may occur due to the fact that users are “required to register” to post. 60% of the one-time posters were seeking information on specific design tool or technique while 28.6% engaged in conversational topics. Interestingly, only 5.6% of the one-time posters were requesting critique – emphasizing members’ tendency to become familiar with the site before posting a critique request. See table 2 for the community response for posts submitted by one-time posters – reflecting that even new members receive significant number of replies and views from the community. Interestingly, 30% of these one-time posts also replied to other designers’ posts, indicating the presence of passive participation from a significantly large number of users. Ranking as an ineffective mechanism to encourage contribution: Like many existing online communities, Core77 uses a ranking model to categorize its members. The current ranking model utilized is as follows: Administrator, Moderator, Full-self Realization, Step four, Step three, Step two, Step one, and then members without any rank. See table 3 for a rank-based distribution of members and their contribution. Moderators and administrators not only maintaining a clean and welcoming environment but also are the most active users (23% posts are contributed by them who are 0.13% of the total members). Most of the other members take a semi-active role, posting when needed. Positions such as moderators and administrators are usually awarded to members who have high rank (calculated based on number of posts). While ranking is used as a mechanism for encouraging participation, the current ranking is proved ineffective as 93% users in the site have no rank, with 39.5% never posting anything and an additional 23.5% posting just once. High effort required for finding thread of interest is reported as one of the main reasons hindering contribution.

224

M. Sharmin and B.P. Bailey Table 3. Contribution statistics of members in core77

Rank Category

Post (Required)

Post (Avg.)

Administrator Moderator Full selfrealization Step four Step three Step two Step one No rank Spammer Banned

>=600

3275.0 1608.3

200~599 100~199 50~99 20~49