Knowledge, More or Less

Citation preview

Knowledge, More or Less John Bacon Noûs, Vol. 17, No. 4. (Nov., 1983), pp. 663-668. Stable URL: http://links.jstor.org/sici?sici=0029-4624%28198311%2917%3A4%3C663%3AKMOL%3E2.0.CO%3B2-5 Noûs is currently published by Blackwell Publishing.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/black.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.org Mon Jun 4 08:33:36 2007

CRITICAL STUDIES

Knowledge, More or Less Part I of a critical study of Midwest Studies i n Philosophy V : Studies i n Epistemology (Minneapolis, MN : University of Minnesota Press, 1980), 568 pp., $35.00 (cloth), $15.00 (paper). JOHN BACON UNIVERSITY OF SYDNEY

T h e title Studies in Epistemology notwithstanding, only half of the papers in the closely packed volume V of the Midwest Studies in Philosophy [7] are on epistemology and the definition of knowledge. I discuss these below. I n Part I1 of this critical study, Michael Devitt will review contributions to realism and the philosophy of language. I n addition, the volume includes Jonathan Bennett's "Spinoza's Vacuum Argument," Panayot Butchvarov's "Adverbial Theories of Consciousness," L. Jonathan Cohen's "Is a Criterion of Verifiability Possible?", David M. Rosenthal's "Keeping Matter in Mind," and Peter Unger's "The Problem of the Many." While each of these papers is of interest in its own right, they did not fit into either part of our review. The post-Gettier disillusionment has its hedgehogs and its foxes. Embracing the moral that knowledge cannot be a simple matter, some enterprising souls are venturing bewilderingly complex definitions. Others, more cautious, have gone back to refining the ingredients, such as belief, justification, and other necessary conditions for knowledge. But first, the hedgehogs.

1. DEFINITIONS O F KNOWLEDGE

Three philosophers contribute full-fledged definitions of knowledge to this collection: Mark Pastin, Hector-Neri Castaiieda, and implicitly Roderick Chisholm. (I use 'definition' loosely for the production of necessary and sufficient conditions o r something stronger.) Pastin's and Castaiieda's accounts are relativized. I n fact, Pastin does not actually try to define knowledge per se but rather the attribution of knowledge. I turn to his account first. a. Pastin's basic idea in "The Multi-Perspectival Theory of Knowledge" is that (MP:A)

A's attribution at tlof knowledge of P to Sat tzis true iff (1) S believes P a t tz, (2) Pis true, and (3) S at tzis able to inform A at t l about P. (105)

T h e central notion here of capacity to inform is defined tentatively as Sat t2caninform A at tlabout P: there is a property +such that (1)S has *att2,and (2) if A should justifiably believe at tl that S has +at t2, A would alsojustifiably believe P at tl. (103)

I$ here may not be a "funny" property, i.e., one trading on paradoxes of material conditionals o r vacuous abstraction. Although Pastin points out that this latter definition is not yet adequate, a particularly simple class of counterexamples is got by letting 4 be + a n d P be 5 has +l . T h e definiens then (assuming existential generalization into a belief follows from 5 has context, which may be controversial). Thus S can inform A about any of S's properties +whatsoever. need not be "funny." This is way too broad. Pastin accordingly tightens u p his definition:

+'

+

+

Sat t2cancognitively inform A at tlabout P: there is a property +such that (1) S has and believes P at t2,and (2) these two facts about S at t2,jointly but not separately, would if justifiably believed by A at t1 suffice for A justifiably to believe P at tl. (105f)

Unfortunately, this is still too broad. Let I$ be illness and P be 'S is ill and has a belief1 , and suppose that S is in fact ill and believes P at tz. Again, the definiens follows: S is able to inform A cognitively about P, satisfying condition (3) for knowledge-attribution. Conditions (1) and (2) already hold ex hypothesi, so by (MP:A) A's attribution of knowledge that P to S would be true in this case. Notice that many other unfunny properties could have been taken besides illness. Also, it does not matter who A is: the example shows that on Pastin's definitions anybody's knowledge-attribution of P would be justified in such a case, even if they knew nothing of S's existence. Finally, the same trick does not verify A's attribution of knowledge that S is ill to S: thus conjunctive simplification fails in knowledge contexts for Pastin. It will be interesting to see how he reinforces his promising theory. b. Castalieda's definition of knowledge in "The Theory of Questions, Epistemic Powers, and the Indexical Theory of Knowledge" outdoes Chisholm in its complexity, if that is possible. It takes u p three-quarters of a page, exclusive of explanatory notes, and is as yet only tentative, at that (227). We begin to appreciate the foxes' motive for proceeding piecemeal: what can you d o with a definition too difficult for anyone but its fond parent to understand or to apply readily? (In fairness, Castarieda's epigraph from Sellars is well taken: "The real danger of oversimplified models is not that they are oversimple, but that we may be satisfied with them. . ." (193)) Without actually giving Castarieda's definition of knowledge, I shall concentrate on one of its most intriguing features, indexicality. For Castarieda knowledge is a genus o r determinable with many species or determinate forms. 'Know' is accordingly indexical, o r context-relative (210): for each context i, knowledgeiis a species of knowledge. T h e relevant epistemic contexts, according to Castarieda, are sextuples comprising (1) a set of agents making u p the epistemic community, (2) a set of questions at issue, (3) a set of possible answers to (2) whose epistemic status is in question, (4) a set of presupposed beliefs of (1) about (3), (5) a set of abnormal respects, and (6) a network of assumed structural inference rules deemed relevant to (3)in view of (4) and (5) by part of (1). I n speaking of species of knowledge, Castarieda seems implicitly to recognize a genus, knowledge simpliciter (although he explicitly denies this [210]).

KNOWLEDGE, MORE OR LESS

665

Even if English 'know' is always indexical, we are free to coin 'knows,' for 'knowsiat some i.' (For convenience I shall abbreviate 'know,' simply as 'know.') I n these terms, we are free to say that (at common t) S knowsi that p but does not knowj that p (cf. 210), and even that S knows that p but does not knowj that p. What is ruled out is that S knowi that p but not knowi that p. Since Castaneda affirms the truth of knowledge (227[vi] 1),it is also ruled out that S knowi that p while X knowj that not-p, or that S know that p while X know that not-p. Thus indexicality or context-relativity need not imply Protagorean relativism. Castaneda in fact escapes the apparent relativism of (4) and (6)above by requiring the parts relied upon by the knower to be true or valid (227[vi]2). An important aspect of Castaneda's epistemic contexts is (2), the set of questions. This is meant to incorporate the Plato-Popper-Powers insight that knowledge is the power to answer questions. It is Castaneda's idea that a declarative sentence may be queried with respect to any components that are substituends for a variable of some sort (213; my interpretation). Interrogation thus behaves syntactically rather like second-order abstraction. Each declarative sentence generates a set of questions it answers. Typically, only some of these questions will be at issue in a given epistemic context. What I find interesting about this idea, apart from Plato's, Popper's, and Powers' support for it, is that it offers a n alternative device to abstraction o r quantifier scope (Thomason and Stalnaker [9]) to account for the de dictolde re distinction (terminology pace Castaiieda and Devitt). This is not altogether surprising, in view of the similarity between Castaneda's interrogation and abstraction. But Castaiieda takes the ambiguity out of the knowledge-sentence and locates it in the context or setting. Take Larry Powers' example in which at t (7) Andrea does not know that there is a four-letter word ending in 'eny' (she is inclined to answer "No" to the question "Is there a four-letter word etc.?") but (8) Andrea knows that there is a four-letter word ending in 'eny' (she answers a confident "Yes" to the question "Is 'deny' a four-letter etc.?") Castaneda's diagnosis is that, while 'know' is univocal here, it is indexical, referring to two different contexts, say j and i, corresponding to the two different questions (210f). Thus in (7) 'know' is implicitly 'knowj,' while in (8) 'knows' is 'knowsi.' I suggest, on the other hand, that while 'know' is univocal, (7) and (8) are syntactically ambiguous as to quantifier scope (cf. Castaiieda 217f). (7) is to be read de dicto, but (8) as the de re There is a word which Andrea knows to have four letters ending in 'eny.' Ultimately, of course, readings must be determined by contexts, perhaps the very sort of contexts Castaiieda proposes. But in this case, I suspect that the context-dependency extends wider syntactically than the single word 'know.' Castaneda's proposed definition of knowledge deserves further exploration, which may perhaps be hoped to reduce its complexity. c. Chisholm. "A Version of Foundationalism" continues chisholming away at the account of knowledge based on epistemic preferability already essayed in [2], [3], [4], [5], and [6]. T h e account comprises a network of

definitions and a set of epistemic principles couched in the defined terms. Both of these components are changed in the present version. The following epistemic concepts are redefined: self-presenting, evident, basis, directly evident, makes evident, nondefectively evident. Eight new epistemic principles are announced. But for one or two, it is not clear whether these augment or supersede the eight principles of [6]. A basic departure from earlier versions is a switch from propositional attitudes to the attribution, in particular the selfattribution, of properties. Chisholm does not explain this switch, but reasons may be gleaned, e.g., from Lewis [a]. However, as Chisholm notes, all the principles and definitions of this article could be stated the old way in terms of propositional attitudes, with some loss of generality. For the sake of readability, I follow that policy here. Chisholm does not actually give his definition of knowledge here, although he alludes to it. We may assume that it is verbally the same as in [6]: nondefectively evident true belief. But both 'nondefectively evident' and its ingredient 'self-presenting' have been redefined, apparently for the worse. Nondefective evidence no longer implies evidence (it would if evidence were closed under adjunction and entailment). The problem is that the 'or evident and' condition has been dropped from the new definiens (563 n. 7): I suggest putting it back in. 'Nondefectively evident' then means, as in ([6]: 267), 'certain or evident and entailed by a conjunction each conjunct of which has a basis that is not the basis of any falsehood.' Taking 'basis' intuitively, I am impressed with Chisholm's definition u p to this point, and can think of no counterexamples. 'Basis' has been redefined. e is now a basis for h iff e is true and selfpresenting, and (necessarily (cf. [6]: 262)) if e is true and h epistemically in the clear, then h is evident (568). The condition of truth on e has been added because the new definiens of 'self-presenting' no longer implies truth: 'ipsofacto (necessarily?) believed if considered and true, and possible to consider while true' (549). Indeed, it appears that all impossible propositions are now selfpresenting for Chisholm, if we take 'while' in the sense of 'whenever.' Zen fans please note. 'Epistemically in the clear' is in turn defined by means of 'confirms,' and the latter in terms of Chisholm's basic notions. I leave these definitions aside, for by now 'basis' has become so complex that I have lost confidence of understanding it. It may be worth the effort, though, in view of the basic appeal of Chisholm's approach to knowledge. Chisholm defines entailment as necessary belief-transmission: p entails q iff, necessarily, whoever believes p also believes q (550, again adjusting from properties to propositions). This seems much too stringent to be of use. I would venture to affirm that for any two propositions, there is a possible world where someone believes the one but not the other. If I am right, then Chisholm's entailment collapses into identity. Chisholm concludes his paper with a persuasive argument that his system is indeed foundationalism. Considered in this light, what I have always found interesting in Chisholm's foundationalism is that he does not just rely on a single justification relation. 2. EPISTEMOLOGICAL TIDBITS

a. Sosa and Alston. Now let us turn our attention more fleetingly to the remaining epistemological papers in the collection under review. Ernest Sosa ("The Raft and the Pyramid: Coherence versus Foundations in the Theory of Knowledge"), as well as Chisholm in the paper just discussed, address them-

KNOWLEDGE. MORE O R LESS

667

selves to the clarification of foundationalism. Curiously, Sosa defines coherentism as a species of foundationalism. While his sympathies clearly lie with foundationalism in the narrower sense, unlike Chisholm he does not offer an actual foundationalist theory here. He does suggest, however, that intellectual virtues may prove as important for epistemology as the moral virtues for ethics. William Alston ("Level-Confusions in Epistemology") seeks to show how arguments for foundationalism and skepticism have been vitiated by such confusions as between believing and believing that one knows. Sextus, Descartes, and Chisholm come in for raps. b. Bonjour, Audi, and Goldman. Laurence Bonjour and Robert Audi defend internalist conceptions ofjustification, according to which the links in a justificatory chain (tree) must in some sense be available to the subject's consciousness. Bonjour ("Externalist Theories of Empirical Knowledge") is concerned with the justification of belief, but he devotes a good part of his discussion to a critique of Armstrong's externalist account of noninferential knowledge in terms of reliability [I]. He acknowledges that this approach can seem plausible if we limit basic beliefs to perception and introspection. Most of Bonjour's proffered counterexamples accordingly involve clairvoyance, which presumably should not be excluded on purely epistemological grounds. Audi ("Defeated Knowledge, Reliability, and Justification") develops necessary internalist conditions for perceptual knowledge, failing which he holds that such knowledge is defeated. Both these papers are serious efforts to stem the rising tide of externalism and reliabilism. Alvin Goldman ("The Internalist Conception of Justification") tries to give an ultimately externalist account of the general form of "doxastic decision principles" that will dojustice to the essential insights of internalism. A doxastic decision principle is a function from a subject's consciously accessible cognitively relevant states to doxastic attitudes. Although Goldman does not work out details here, something of the sort is clearly needed in applied epistemology. c. Johnson and Ackerman. Oliver A. Johnson ("The Standard Definition") and Diana Ackerman ("Natural Kinds, Concepts, and Propositional Attitudes," 472) would escape Gettier by requiring that every essential link in the justificatory chain for a knowledge claim be true. As Johnson puts it, the traditional understanding of tjustified true' has been tjustified to be true.' This may be on the right track as far as it goes; i.e., it may be a necessary condition for knowledge (as Harman holds, 163, P2). Castaiieda's purported counterexample of a woman who learns to draw correct conclus~onsfrom circumstances apprising her of the falsity of beliefs (203) is wide of the mark: the false beliefs themselves need not be in her justificatory chain. Still, it is very doubtful that Johnson and Ackerman's proposed condition is sufficient. Ackerman's version is not armored against defeaters. In case Johnson would build indefeasibility into justification, he still needs an account of the former. d. Pappas, Ginet, and Harman. George S. Pappas, Carl Ginet, and Gilbert Harman attack the problem of undermined knowledge. Pappas ("Lost Justification") argues that knowledge whose justification has been forgotten becomes immediate knowledge, rather than necessarily lapsing into mere true belief. I would hold, to the contrary, that it remains justified as before, provided the knower remembers that it was justified. Ginet ("Knowing Less by Knowing More") and Harman ("Reasoning and Evidence One Does Not Possess") take u p cases in which knowledge might be undermined by additional knowledge. (E.g., I know that my cat Inka is in the room--or do I, being about to discover of her unknown identical twin Stinka that the latter has just wan-

dered into the next room?) Ginet contrasts such a case with would-be knowledge that fails to make the grade at all owing to a defeater: I really knew that Inka was in the room, until that fatal glance through the door. Harman's intuition diverges here: "we are reluctant to say that. . . the person knows" (164). Accordingly, Harman would make essential to reasoning (includingjustificatory chains) the disbelief in obtainable counterevidence to the justification of one's conclusion. By his P2 (see 92c above), if the disbelief is false, the reasoning cannot yield knowledge. e. Hunter, Perry, a n d Lehrer. F. M. Hunter and John Perry examine doxastic concepts. While Hunter's paper "Believing" contains many interesting and useful observations about how we use 'believe,' he is curiously defeatist about getting on top of the concept. "The question what believing is presupposes that it is something, whereas we could now say, it is not anything" (250). I hope this is not the last word. Perry ("Belief and Acceptance") seeks to show that a three-place sentence-acceptance-at-t predicate is not definable in terms of a five-place 'believes-true' predicate ("at t, X believes that 'S' is true for cu at 70). Finally, Keith Lehrer ("Coherence and the Racehorse Paradox") poses some interesting lottery-like paradoxes to show that probability and informational content alone are insufficient to pick out most reasonable hypotheses. The missing ingredient, he urges, is coherence, based on negative relevance. It is evident from this inadequate summary that Studies i n Epistemology is a rich resource for recent research in the field. At the same time, it seems to typify what must sureIy be a transitional impasse. The Gettier problem exposed a deep anomaly in traditional epistemology. It is not yet clear which responses are paradigm-patchwork and which may contain the seeds of renewal. REFERENCES [ l ] D. M. Armstrong, Belief, Truth and Knowledge (Cambridge: University Press, 1973). [2] Roderick M. Chisholm, Theory of (Knowledge (Englewood Cliffs: Prentice-Hall, 1966). "On the Nature of Empirical Evidence," Experience and Theory, eds., Foster [3] , and Swanson (Amherst: University of Massachusetts Press, 1970): 103-134. revision of [3],Empirical Knowledge: Readings from Contemporary Sources, eds., [4] , Chisholm and Swartz (Englewood Cliffs: Prentice-Hall, 1973): 224-249. [5] , 2nd ed. of [2] (1976). revision of [4], Essays on Knowledge and Justification, eds., Pappas and Swain [6] , (Ithaca: Cornell University Press, 1978): 253-278. [7] Peter A. French, Theodore E. Uehling Jr., and Howard K. Wettstein (eds.),Studies in Epistemology: Midwest Studies in Philosophy 5(Minneapolis: University of Minnesota Press, 1980). [8] David K. Lewis, "Attitudesde dicta anddese,"PhilosophcialReview87(1979):513-533. [9] Richmond H. Thomason, and Robert C. Stalnaker, "Modality and Reference,"Nous Z(1968): 359-372.