Practical Engagement and the Skeptical Challenge — Epoche B2
Why the Dream Argument Tells Against Pragmatic Encroachment The complaint that contextualism is too linguistic Epistemic contextualism holds that the truth conditions of a sentence of the form 'S knows that p' vary with the context of the person making the attribution. Against it a familiar complaint is lodged: that it makes skepticism a fact about conversations. Whatever changes when the dream hypothesis of Descartes's first Meditation is raised [1] , the complaint runs, it is not merely what we may truthfully say — something changes in the believer's relation to the world, in what she can rely on and act from, and a semantic theory of 'knows' does not reach it. That complaint names a real position, though it is usually made without naming it. The view that what a subject knows depends on what practically turns on her being right — on the stakes rather than on the attributor's conversation — is pragmatic encroachment , defended by John Hawthorne [2] , Jason Stanley [3] , and Jeremy Fantl and Matthew McGrath [4] . It is the theory the complaint is groping for. This essay argues that the dream argument is the worst possible case for that theory, and a good case for the attention-based contextualism it was meant to improve on. The reason is that the dream hypothesis is practically inert: it does not change what turns on being right, because it bears on every available option identically. A theory that makes knowledge sensitive to stakes therefore predicts that the dream hypothesis leaves knowledge undisturbed, which is the opposite of what its advocates want from it. The claim would be false if entertaining the dream hypothesis could be shown to alter the ranking of any two courses of action. Lewis's machinery David Lewis's 'Elusive Knowledge' (1996 [5] ) gives contextualism a definite form. To know a proposition is for one's evidence to eliminate every possibility in which it is false — every possibility, that is, except those one is properly ignoring . The exception clause is not a hedge; it is the whole content of the theory, and Lewis states rules governing when ignoring is proper. Some are prohibitions. One may never ignore actuality. One may not ignore what the subject believes, or ought to believe, to be the case. One may not ignore a possibility that saliently resembles one that is not being ignored. Others are permissions: one may ignore possibilities in which reliable processes fail, in which normal methods misfire, and one may ignore what is generally ignored in the conversation. The last rule is the Rule of Attention, and it is blunt. A possibility that is being attended to is not being ignored, properly or otherwise, and so must be eliminated by evidence. Since one's evidence cannot rule out that one is dreaming — the hypothesis is constructed so that all one's evidence is consistent with it — the moment the possibility is attended to, most of what one claimed to know is lost. Lewis accepts this consequence without complaint, and states it as a result rather than an embarrassment: epistemology destroys knowledge, and it does so for as long as we do epistemology. Step out of the seminar and knowledge returns, because the possibility ceases to be attended to and may properly be ignored again. It is worth noticing that this already delivers the phenomenon the complaint claimed contextualism could not reach. The account is not that we change what we say while our epistemic position stays fixed; it is that attending to a possibility changes whether the subject knows. The instability is in the knowing. The rival, stated properly Pragmatic encroachment makes a different move. Its data are Keith DeRose's bank cases [6] : a subject with the same evidence about a bank's Saturday hours knows in the version where little turns on it, and does not know in the version where a mortgage payment does. What differs is not the conversation, since we may hold the attributor fixed, and not the evidence. What differs is what is at stake for the subject. So knowledge, on this view, depends partly on practical factors, and Fantl and McGrath give the connection a slogan: if one knows that p, one is warranted in treating p as a reason for action. The attraction for the complaint under discussion is obvious. Encroachment locates the shift in the subject and her situation rather than in a speaker's conversational standards, and it ties knowledge directly to what one may act on. If skepticism is felt as a disturbance in the relation between belief and action, encroachment looks like the theory that says so. Why it cannot take the dream case Consider what raising the stakes does in a bank case. It makes one option worse in one branch of the possibilities: if the bank is shut and the payment is missed, the loss is large; if it is open, nothing is lost by having checked. The stakes matter because they are differential . Being wrong about p costs more under one course of action than under another, and that is what makes further inquiry rational. The dream hypothesis has no such structure. Suppose it is true. Then the bento one is deciding whether to buy is dreamt, the money is dreamt, the hunger is dreamt, and so is the outcome of every alternative — buying, not buying, buying something else. The hypothesis applies uniformly to every branch of the decision, so it discounts every outcome by the same factor and leaves the ranking of the options exactly as it was. Whatever it was rational to do if the world is real, it remains rational to do if the world is dreamt, because from the inside the dreamt and the real bento call for the same action. An account on which knowledge tracks practical stakes must therefore say that raising the dream hypothesis does nothing to what one knows. Nothing turns on it differentially, nothing about any decision changes, and by the encroachment theorist's own diagnostic the subject remains in a low-stakes situation. But the dream argument is not felt as inert, and its force is what t