From E=K to Scepticism?

Volume 58 • Number 233 • October 2008 CONTENTS ARTICLES A New Defence of Anselmian Theism Solving the Problem of Easy Knowledge Forbidden Ways of Life Intentional Action and the Praise–Blame Asymmetry Perceptual Knowledge and the Metaphysics of Experience A Dilemma for Particularist Virtue Ethics Yujin Nagasawa Tim Black Ben Colburn 577 597 618 Frank Hindriks 630 Michael Pace Rebecca Stangl 642 665 DISCUSSIONS From E = K to Scepticism? A Problem for Guidance Control Modal Knowledge. Counterfactual Knowledge and the Role of Experience Clayton Littlejohn 679 Patrick Todd and Neal A. Tognazzini 685 C.S. Jenkins 693 Daniel Devereux Mark LeBar Garrett Cullity Noël Carroll 702 711 720 732 741 CRITICAL STUDIES Meno Re-examined Development and Reasons Pyrrhic Pyrrhonism On the Aesthetic Function of Art BOOK REVIEWS The Scots Philosophical Club and the University of St Andrews The Philosophical Quarterly Vol. , No.  ISSN – October  doi: ./j.-...x DISCUSSIONS FROM E = K TO SCEPTICISM? B C L In a recent article Dylan Dodd has argued that anyone who holds that all knowledge is evidence must concede that we know next to nothing about the external world. The argument is intended to show that any infallibilist account of knowledge is committed to scepticism, and that anyone who identifies our evidence with the propositions we know is committed to infallibilism. I shall offer some reasons for thinking Dodd’s argument is unsound, and explain where his argument goes wrong. I. INTRODUCTION According to Timothy Williamson, your evidence consists of all and only the propositions you know, ‘(E=K)’ hereafter.1 According to Dylan Dodd, (E=K) engenders scepticism.2 Of course, Williamson is not a sceptic, but if Dodd is right, he cannot have it both ways. Anybody who claims that (E=K) is true must be prepared to accept the sceptical consequences. Dodd’s argument can be stated as follows: If (E=K), infallibilism is true, i.e., S knows that p only if the probability of p on S’s evidence (hereafter PS p) =  For virtually everything we believe, our evidence underdetermines our beliefs (i.e., it is all but certain that if S believes that p, PS p < ) If infallibilism is true, you cannot know that p if p is underdetermined by the evidence Therefore if (E=K), virtually nothing we believe amounts to knowledge. . . . C. Dodd’s defence of () is straightforward. On the assumption that you know that p, (E=K) establishes that p is part of your evidence. When p is part of your evidence, the epistemic probability of p on your evidence =  (i.e., PS p = ). Thus (E=K) is incompatible with fallibilism, expressed as S can know that p even if PS p < . F. T. Williamson, Knowledge and its Limits (Oxford UP, ). D. Dodd, ‘Why Williamson should be a Sceptic’, The Philosophical Quarterly,  (), pp. –. 1 2 ©  The Author Journal compilation ©  The Editors of The Philosophical Quarterly Published by Blackwell Publishing,  Garsington Road, Oxford  , UK, and  Main Street, Malden,  , USA  CLAYTON LITTLEJOHN Dodd defends () by appeal to intuition. Let p be the proposition It is sunny outside. On any non-sceptical view, you should be able to know that it is sunny outside, if you have just undergone the right sort of veridical experience and judged this as a result. Or suppose p is the proposition I shall not be able to afford to go on a safari this summer. On any non-sceptical view, I should be able to know that this is so, since I also know that I live on a philosopher’s salary and have nothing put away in savings. Although it seems that I know such things, Dodd remarks (p. ) that ‘when one first thinks about this, it is counter-intuitive to say that there is literally no chance at all that my veridical experience of the sunny sky is a hallucination’. Concerning beliefs about the future, he notes that on Williamson’s view it follows from the claim that I know I shall not be able to afford an African safari that there is no chance that I shall win a lottery which would give me the money for an African safari. This cannot be right, he says (p. ), because there is no denying that there ‘just is a chance that my lottery ticket will win, and if it wins, I shall probably be able to afford the safari’. Not much needs to be said in defence of (). This is true as a matter of definition. It appears, then, that (E=K) engenders scepticism. Appearances being what they are, they can mislead. I do not think Dodd has shown that (E=K) carries with it any troubling sceptical implications. In §II, I shall explain why the argument from (E=K) to scepticism should be treated with suspicion. In §II, I shall offer an explanation of where his argument goes wrong. II. DOUBTS In this section, I shall raise some preliminary doubts concerning the argument from (E=K) to scepticism. First, it would be surprising if Dodd’s argument worked, because it does not require anything as strong as (E=K). It involves only the assumption that knowledge of p’s truth is sufficient for p’s inclusion in S’s evidence. I shall assume, although this is a controversial assumption, that evidence is propositional. If this is correct, then in order to avoid the sceptical worries Dodd raises, if p is to be included in S’s evidence, it must satisfy further requirements beyond merely being known. Whatever these requirements amounted to, they would have to establish that no proposition p could be part of S’s evidence if S could properly concede that there is the slightest chance that p is mistaken. If mere knowledge of p’s truth is insufficient for p’s inclusion in our evidence, then we possess much less evidence than Williamson suggests. It might seem that if we had less evidence for our beliefs concerning the external world than initially seemed available, we would not thereby be better placed to deal with sceptical doubts concerning those beliefs. It is possible, of course, for me to know that p and to acquire later additional evidence which makes me epistemically worse off as a result. This happens when misleading evidence is acquired, or evidence is found which undermines the evidence which initially supported p. However, the propositions I am considering are propositions which (E=K) assures us are part of our evidence, and these are not invariably propositions which would undermine the justification for our beliefs or mislead us. If ©  The Author Journal compilation ©  The Editors of The Philosophical Quarterly FROM E = K TO SCEPTICISM?  it is really (E=K) which generates untoward sceptical consequences, denying (E=K) should help us to deal with these sceptical worries. Denying (E=K) and insisting that Williamson’s standards for having evidence are too permissive should not increase our confidence that our beliefs constitute knowledge, given what our evidence is. Of course, not everyone believes that evidence is propositional.3 A reason for denying (E=K) might be the thesis that some evidence is non-propositional (e.g., perceptual states which lack conceptual content). But it would surely still be odd to deny that if one account of propositional evidence says that nothing more than knowledge of p’s truth is necessary for p’s inclusion in your evidence, and a second account insists that knowledge is not enough, the first faces greater sceptical challenges than the second. If this suggestion is mistaken, then if the problem that Dodd raised for Williamson’s view is genuine, it is a problem for every view on which any evidence is propositional. Someone might say that the advantage of a view on which knowledge of p’s truth is insufficient for p’s inclusion in S’s evidence is that on this view it is not a necessary condition on having knowledge of p’s truth that PS p = .4 I do not see that this is an advantage. It is a consequence of Williamson’s view that a necessary condition for knowing that p is that PS p = . It is for him a rather trivial consequence of the fact that he is using the concept of knowledge to explicate the concept of epistemic probability that nothing more than S’s coming to know that p is necessary for it to be the case that PS p = .5 However, it is not on his view a necessary precondition for coming to know that p that there is evidence independent of p on which PS p = . While it is his view that S does not know that p unless PS p = , Williamson (p. ) denies that p must be absolutely certain for S in order for S to know that p or for it to be the case that PS p = . While scepticism plausibly follows from the claim that S knows p only if either p is absolutely certain for S or S had evidence prior to accepting p which logically entails p, Williamson denies that either of these claims is necessary for S’s knowledge of p or for it to be the case that PS p = . It is also worth noting that the argument that knowledge of p’s truth is insufficient for p’s inclusion in S’s evidence would also need to be supplemented with an explanation of how a remark such as ‘It is sunny outside, and while Bob knows that this is true he has no reason to believe it is not overcast’ could ever turn out to be correct. On the assumption that pieces of evidence are reasons for belief, and that knowledge that p is true is insufficient for p’s inclusion in one’s evidence, such a remark must seem reasonable. Yet it seems unreasonable. The natural explanation is simply that if you know that p, this is all you could need for p’s inclusion in your evidence. The remark is unreasonable because it is self-contradictory to assert that p is known to S, but that S has no reason to believe the obvious lightweight consequences of p. This explanation, however, is ruled out if knowledge of p’s truth is insufficient for p’s inclusion in one’s evidence. A final worry concerns Dodd’s contention that (E=K) commits us to scepticism on the ground that it is incompatible with (F). Why think that denying (F) as he 3 4 5 Thanks to an anonymous referee for pressing this point. Thanks to an anonymous referee for raising this worry. See Williamson, Knowledge and its Limits, p. . ©  The Author Journal compilation ©  The Editors of The Philosophical Quarterly  CLAYTON LITTLEJOHN formulates it leads to scepticism? On one standard account of epistemic modals, the truth-conditions for epistemic modals (e.g., ‘might’, ‘must’, etc.) are specified in terms of knowledge. On this view, p is epistemically necessary for S iff ¬p is obviously inconsistent with something S knows.6 When p is known to S, ¬p is as obviously inconsistent with something known to S as anything could be. So if p is known by S, p is epistemically necessary for S. Besides, given that there is this connection between epistemic possibility and probability, it cannot be simultaneously true that p is necessary for S and that PS ¬p > . An argument shows that (F) is incompatible with this account of epistemic modals and the denial of scepticism: . . . If S knows that p, p is epistemically necessary for S If p is epistemically necessary for S, it is not the case that PS p <  Therefore it is not the case that S knows p if PS p < . This conclusion is just the denial of (F). The assumptions needed for running the argument against fallibilism as Dodd formulated it are relatively innocuous. This is really what the worry is. You can deny those assumptions. You can say that a proposition might be epistemically necessary for someone while at the same time the proposition’s negation has a non-zero epistemic probability. You can say that your assertion ‘I know that p’ does not negate my assertion ‘You might be mistaken about p’. But whatever reason you might have for denying these claims, it is hard to believe that you have to deny them in order to avoid scepticism, even if they are incompatible with (F). III. DIAGNOSIS I have given some reasons to think there is something wrong with the argument from (E=K) to scepticism. My hypothesis is that Dodd’s argument runs into difficulty because of his formulation of fallibilism. There is a sense in which it is fair to say that infallibilists are committed to scepticism. Williamson is not an infallibilist in this sense. There is a sense in which I think it is fair to say that Williamson is an infallibilist. But on this conception of infallibilism, infallibilism does not entail scepticism. I shall begin by considering two formulations of fallibilism. The first is due to Dodd and the second to Stewart Cohen. F. S can know that p even if PS p <  F. S can come to know that p on the basis of r even if r does not entail p.7 In the previous section, I pointed out that (F) is incompatible with (E=K). It is also incompatible with a standard account of epistemic modals combined with the denial of scepticism. It seems that while Williamson is an infallibilist in the sense that he rejects (F), infallibilism so understood does not clearly entail scepticism. I can see no 6 See K. DeRose, ‘Epistemic Possibilities’, Philosophical Review,  (), pp. –; J. Hawthorne, Knowledge and Lotteries (Oxford UP, ). 7 See S. Cohen, ‘How to Be a Fallibilist’, Philosophical Perspectives,  (), pp. –. ©  The Author Journal compilation ©  The Editors of The Philosophical Quarterly FROM E = K TO SCEPTICISM?  evidence that Williamson has explicitly committed himself to (F), or is committed to (F) in virtue of his commitment to (E=K). It might appear that (F) follows from (E=K), because it might seem that given (E=K), S could not come to know that p unless S had independent reasons for accepting p such that it is possible to have those reasons and be mistaken about p. Combining this assumption with (E=K) yields (F), but only because the assumption is (F). So long as Williamson accepts (F) but rejects (F), there seems no reason to think he is committed to scepticism. Dodd thinks that there are no coherent views on which (F) is false and (F) is true. He says (p. ) that we should reject any view which combines (F) with the denial of (F): [(F) and the denial of (F)] together entail that we can raise the probability of a proposition from less than  to  without learning anything which makes it actually more likely that the proposition is true, but by merely inferring that the proposition is true from the evidence we already have. Williamson might say that when you come to know that p, there is something you learn that explains why PS p = . You learn that p is true! Certainly it seems odd to say that p serves as the reason for accepting p, and odd to think that you could get away with saying that PS p =  unless you either accept (F) or accept that whenever you come to know that p without entailing grounds, p is a reason for its own acceptance. However, there are two things to say on Williamson’s behalf. First, Williamson says that it is a trivial consequence of S’s knowing p that PS p = . We are using knowledge to tell us what it is for a proposition to be epistemically necessary. We are using knowledge to tell us what it means to say that there is no chance for S that some proposition is false (i.e., for it to be the case that PS ¬p = ). We are not using an independent conception of what it would take for it to be the case that PS p =  in order to determine whether S knows p, or to determine whether we ought to take back knowledge ascriptions we initially thought were warranted. To think there is something shocking about the claim that PS p =  for some contingent proposition p, while not admitting scepticism, is to read into the notion of epistemic probability something to which Williamson is not committed (e.g., that PS p =  entails that p is absolutely certain for S). Secondly, I observed in the previous section that every non-sceptical view which incorporates the standard view of epistemic modals rejects (F). But this means that given this approach to epistemic necessity and to evidential probability, the only tenable anti-sceptical view is one on which (F) is false and (F) is true. If every nonsceptical view is like this, every sceptical view has to allow for the sort of possibility which Dodd thinks is fatal to Williamson’s view (i.e., situations in which the evidence S had at first was that PS p < , but on accepting p S comes to know that p, and the consequence is that then PS p = ). Dodd might say that I have not dealt with his worry because I have not yet shown that Williamson’s view is compatible with the underdetermination thesis. If anything, I have only reinforced the point Dodd is trying to make, that Williamson must deny the underdetermination thesis. Dodd might add that any view which is incompatible with the underdetermination thesis is committed to scepticism. ©  The Author Journal compilation ©  The Editors of The Philosophical Quarterly  CLAYTON LITTLEJOHN Dodd has formulated the underdetermination thesis in such a way that it is simply the view that for nearly everything we believe the epistemic probability of that belief is less than . But I have argued that if this is all the underdetermination thesis amounts to, then given the standard account of epistemic modals on which (a) nothing more than knowledge is necessary for epistemic necessity, and (b) nothing more than p’s epistemic necessity is necessary for it to be the case that PS p = , only the sceptic can assert that the underdetermination thesis is true. So if Williamson must reject this version of the underdetermination thesis, he is hardly alone in this. Two readings of the underdetermination thesis need to be distinguished: (a) setting on one side our knowledge of p, our remaining evidence for p is logically compatible with ¬p; (b) even if we know that p, given what we know, the epistemic probability of p is less than . Williamson would presumably accept (a) because he accepts (F), but would insist that the underdetermination thesis on this reading is compatible with the denial of (F). As for (b), it is true he would deny this underdetermination thesis, but it is also true that everyone denies this thesis except the sceptic. As a final attempt to show that it cannot be denied that the rejection of (F) comes with sceptical costs, Dodd might put forward two knowledge ascriptions: . . I know I shall not be able to afford to go on an African safari next week I know I have hands. How, Dodd might ask, can I so cavalierly say that my evidence for these beliefs is not underdetermined? After all, it is obvious that The epistemic probability for me that a secret admirer has just slipped a winning lottery ticket under my door is >  . The epistemic probability for me that I am the victim of hallucination is > . . Well, perhaps it is obvious that () and () are true. Perhaps it is obvious that () and () are incompatible with the claim that I know I shall not win the lottery or that what seem to be my hands are not hallucinations. It is obvious that closure is true. So how can it be anything but painfully obvious that I do not know what () and () say I know? This is a good question. I confess that I have no idea how to answer it. What I do know is that this problem is generated by the combination of closure with () and (). It is not generated by assuming (E=K). It is not avoided by denying (E=K). Luckily, we do not have to solve the sceptical puzzle to show that (E=K) is not the source of any serious sceptical worries.8 Southern Methodist University, Texas 8 I would like to express my gratitude to Mike Almeida, Eric Barnes, Chris Buford and an anonymous referee for discussion. ©  The Author Journal compilation ©  The Editors of The Philosophical Quarterly