The skeptical argument that we are considering proceeds as follows.
The burden of sensitivity theories is to explain why the relevant skeptical premise and the associated skeptical thought are false. To this end, sensitivity theorists develop accounts of knowledge, or at least partial accounts of knowledge, that are intended to do that job. The burden of contextualist theories is to explain how the skeptic's claims and ordinary knowledge claims can all be true. To this end, contextualists develop accounts of knowledge language intended to do that job. The next sections consider these strategies in more detail.
1. sensitivity theories and the denial of closure
Sensitivity theories deny premise 1 of SA, claiming that ordinary knowledge about the world does not require knowing that various skeptical hypotheses are false, or that various skeptical possibilities are not actual. The burden of this approach is to make that move plausible, and this is no easy task. For one, the skeptical premise is intuitively plausible as it stands. How could one know that one has two hands while not knowing that one is a handless brain in a vat? But second, the premise gains support from various ‘closure principles’ in the neighborhood.
Let's define a closure principle as a proposition that knowledge is closed under some operation. In other words, a closure principle says that if you start from knowledge, and perform some operation on that knowledge, you get more knowledge as a result. For example, it is plausible that knowledge is closed under known deduction: If you know that p, and you know that p entails q, then you know that q. One might quibble with this principle, insisting that knowledge requires believing that q on the basis of deducing it from p. But presumably the principle could be fixed up accordingly – for example, we can say that knowledge is closed under ‘competent’ deduction. The important point here is that there seem to be plausible closure principles in the neighborhood, and it seems that such principles will be satisfied by o and h in SA above. For example, it seems that if you know that you have two hands, and if you know that having two hands entails that you are not a handless brain in a vat, then you know (or can know by deducing it) that you are not a handless brain in a vat. Put another way, it seems that premise 1 of SA is well supported by plausible closure principles.4
So sensitivity theorists incur a heavy burden. They try to discharge it by providing an account of knowledge that explains why, contrary to first appearances, premise 1 and nearby closure principles are false. More exactly, sensitivity theorists propose a necessary condition on knowledge that is supposed to do this job. This approach has been championed by, among others, Fred Dretske and Robert Nozick.5 Here I will ignore the details of their respective views and focus only on the work that is supposed to be achieved by making ‘sensitivity’ a necessary condition for knowledge.
The most straightforward way to understand the sensitivity condition is in terms of a subjunctive conditional:
Sensitivity. S knows that p only if: If p were false, S would not believe that p.6
Intuitively, if one knows that p, then one would not believe p anyway if p were false. On the standard semantics for subjunctive conditionals, we can understand Sensitivity this way:
S knows that p only if: In the closest possible world where p is false, S does not believe that p.
The idea that knowledge requires sensitivity is intuitively plausible. For example, suppose that you now know that you have two hands. Intuitively, if you did not have two hands you would not believe that you did. If you did not have two hands, you would see that you didn't. On the other hand, suppose your belief that you have two hands is not ‘sensitive’. Suppose, for example, that your brain is being manipulated so that you would believe that you have two hands even if you did not. Intuitively, in that case you do not know that you have two hands, and precisely because your belief that you do is not sensitive.
Placing a sensitivity condition on knowledge yields a distinctive strategy for responding to the skeptical argument above. First, the sensitivity theorist can accept premise 2 of SA and can explain why it is true. In short, one does not know that skeptical hypotheses are false because one's beliefs to that effect are not sensitive. For example, you now believe that your are not a handless brain in a vat, fed ordinary experiences by a supercomputer stimulating your severed nerve endings so as to simulate an ordinary life. But if you were, you would still believe that you were not. Your belief that you are not a brain in a vat is not sensitive and therefore not knowledge.
More importantly for present purposes, the sensitivity theorist can reject premise 1 of SA, along with supporting closure principles in the neighborhood. Consider: We have just seen that your belief that you have two hands is sensitive. But we have also just seen that your belief that you are not a handless brain in a vat is not sensitive. Assuming that your belief that you have two hands satisfies other conditions on knowledge, it follows that you know that you have two hands even though you do not know that you are not a handless brain in a vat. Premise 1 of SA is false. By the same reasoning, relevant closure principles are false as well.
We will consider two problems for this anti-skeptical approach below. But first I want to consider a reaction to the approach that is natural but misguided. The reaction is to think that the approach begs the question against the skeptic in some unacceptable way. At most, one might think, the sensitivity theorist achieves only a ‘stand-off’ with the skeptic, since the approach assumes that ordinary beliefs about the world, such as that I have two hands, are sensitive. That is, the approach assumes that one is not a handless brain in a vat, and that is not an assumption that the skeptic is willing to concede.
This reaction is natural enough, but it misconceives the nature of the anti-skeptical project under consideration. The purpose of that project is not to persuade a non-believing skeptic, or to otherwise refute the skeptic in a way that is rhetorically satisfactory. Rather, the project is to reject something in the skeptical reasoning under consideration, whether or not a ‘real’ skeptic would be satisfied with that rejection. Put another way, the project is to critique the skeptical argument rather than to convince or persuade a skeptical person (ourselves or someone else). In the context of this project, we are looking for a response to skepticism that is theoretically adequate, as opposed to rhetorically or pragmatically adequate. And what would theoretical adequacy require? Just what the sensitivity theorist pretends to provide: an account of knowledge (in this case a partial account) that explains where the skeptical argument goes wrong, and thereby explains how knowledge is possible.7
So that is what sensitivity theorists pretend to do. Are they successful? Here I will review two objections that have been raised against the approach in this regard.
The first objection is that a rejection of premise 1 comes at too high a theoretical cost. One way to put the objection is this: Premise 1 of SA and relevant closure principles seem more plausible than the account of knowledge that rejects them. On this way of thinking, the rejection of closure principles should be seen as a reductio of the sensitivity condition rather than a consequence of it. Putting the objection this way, however, amounts to little more than a conflict of intuitions. A more persuasive statement of the objection calls attention to what Keith DeRose calls ‘abominable conjunctions’.8 Specifically, it would seem that the sensitivity theorist is committed to embarrassing claims such as the following:
- a.I know that I have two hands, but not that I am a handless brain in a vat.
- b.I know that I am sitting in front of a computer, but not that I am merely deceived into thinking I am sitting in front of a computer.
- c.I know I am not in a vat, but I don't know that I am not a handless brain in a vat.
DeRose's point is evident. The conjunctions in a–c border on absurdity, and therefore count heavily against any theory that entails them.
A second objection against sensitivity theories is that they cannot accommodate clear cases of inductive knowledge. This sort of objection has been pressed by Jonathan Vogel, who offers the following two examples.
The Hole-In-One. Sixty golfers are entered in the Wealth and Privilege Invitational Tournament. The course has a short but difficult hole, known as the ‘Heartbreaker’. Before the round begins, you think to yourself that, surely not all sixty players will get a hole-in-one on the ‘Heartbreaker’. (‘New Relevant Alternatives Theory’ 165)
The Rookie Cop. Suppose two policemen confront a mugger, who is standing some distance away with a drawn gun. One of the officers, a rookie, attempts to disarm the mugger by shooting a bullet down the barrel of the mugger's gun. . . . Imagine that the rookie's veteran partner knows what the rookie is trying to do. The veteran sees him fire, but is screened from seeing the result. Aware that his partner is trying something that is all but impossible, the veteran thinks (correctly as it turns out) [that the] rookie missed. (‘Tracking’ 212)
Both of these cases have the following structure: there is a close world where a highly improbable possibility is actual. Put another way, in each case almost all close worlds are p-worlds, but there are some not-p-worlds close to the actual world. Notice that in this sort of case, sensitivity is violated. For example, in the nearest world where all sixty players will get a hole-in-one, you still believe that they won't. In the nearest world where the rookie cop does not miss, the veteran cop still believes that he does. But it is absurd to think that there is no knowledge in the two cases.
Finally, the following case from John Hawthorne suggests that the problem is ubiquitous.
The Disappearing Desk. Suppose that there is a desk in front of me. Quantum mechanics tells us that there is a wave function that describes the space of nomically possible developments of the system that is that desk. On those interpretations of quantum mechanics according to which the wave function gives probability of location, there is some non-zero probability that, within a short while, the particles belonging to the surface of the desk remain more or less unmoved but the material inside the desk unfolds in a bizarre enough way that the system no longer counts as a desk. Owing to its intact surface, the system would be reckoned a desk by normal observers. (4–5)
Sensitivity theorists therefore face two formidable objections. First, their account seems to entail counter-intuitive results, such as the denial of plausible closure principles and DeRose's abominable conjunctions. Second, their account seems inadequate to accommodate at least some cases of inductive knowledge. Perhaps the best way to answer these objections is to take a holistic approach. The idea here is that any account of knowledge will have costs and benefits in the face of SA and related problems. One might argue, then, that the theoretical benefits of sensitivity theories outweigh the costs, relative to competitors. Perhaps this is the best way to understand the case put forward by Dretske and Nozick, and more recently by Kelly Becker.9 The following sections put the competitors more clearly in view.
2. attributor contextualism
Contextualist responses to SA grant that the skeptical reasoning is sound, and that the skeptical conclusion is true, in most cases where they are expressed. But contextualists deny that this has widespread skeptical consequences. The guiding idea is that the truth-conditions of knowledge attributions vary across conversational contexts, with the following anti-skeptical effect: Although the skeptic typically says something true when she says, ‘You don't know’, ordinary speakers (in ordinary conversational contexts) typically say something true when they say, ‘I do know’.
Here is a simple illustration of how this might work. Suppose that in contexts where the skeptic is making her arguments the standards for ‘knowledge’ are very high. For example, in these contexts it requires a great deal of evidence, of very great quality, for sentences of the form ‘S knows that p’ to come out true. But suppose also that in ordinary contexts standards are much lower. In these ordinary contexts one requires much less evidence for sentences of the form ‘S knows that p’ to come out true. If these suppositions hold, then our knowledge language will work exactly as the contextualist suggests: the skeptic will be right when she claims ‘S does not know that p’, but ordinary speakers will often enough be right when they claim ‘I do know that p’.10
In addition to this anti-skeptical effect, contextualists claim two further advantages of their view. First, it explains why skeptical arguments can seem so convincing. Skeptics and non-skeptics alike have long noted a puzzling dynamic: skeptical arguments can seem persuasive while we are engaging them, but then their power fades as soon as we cease from philosophizing.11 The contextualist response to skepticism has a ready explanation for this phenomenon. Namely, in contexts where we are considering skeptical scenarios and the like, the standards for knowledge get raised unusually high, and so the skeptic is right when she claims ‘We do not know’ in those contexts. On the other hand, the standards for knowledge operative in ordinary contexts are not so high, and so when we return to these in our non-philosophical lives we are right when we claim ‘We do know’ in those contexts. In each context, the claims that we make seem true because they are true!
Second, contextualism achieves its anti-skeptical effect without denying plausible closure principles. Relative to skeptical contexts, we ‘know’ neither that ordinary propositions about the world are true nor that skeptical scenarios are false. Relative to ordinary contexts, however, we ‘know’ both that ordinary propositions about the world are true and that skeptical scenarios are false. We may not be able to say that we know that skeptical scenarios are false, since mentioning skeptical scenarios tends to move us out of an ordinary conversational context and into one that is skeptical, thereby raising the standards for ‘knowledge’ and making the saying false. The point remains, however, that closure principles hold across all contexts: in no single context is the claim ‘I know that I have two hands’ true and ‘I know that I am not a handless brain in a vat’ false.
Some philosophers have argued that contextualist theories concede too much to the skeptic, however. As we have seen, the contextualist is happy to say that the skeptic is right relative to skeptical context – when the skeptic claims ‘You don't know that you have two hands’, or ‘No one knows he is not a brain in a vat’, these claims are true in the contexts where they are made. But many philosophers would like to deny just that. That is, many philosophers want to say that the skeptic is wrong when she makes such claims. One way to press the point is to consider Moore's statements when he says things like ‘Of course I know that here is a hand’, and ‘Of course I know that the world has existed for more than five minutes’. Many philosophers want to say that Moore is right when he opposes the skeptic in these ways, even though Moore's claims are made in a context where he is engaging skeptical arguments.
DeRose has argued that the contextualist need not make the concessions at issue.12 It is consistent with contextualism, DeRose points out, to hold that skeptical considerations fail to raise the standards for knowledge as high as skepticism requires, so that knowledge claims come out true, and skeptical claims come out false, even relative to standards operative in philosophical contexts. One way this might happen is if non-skeptics have ‘veto power’ over skeptical attempts to raise the standards for knowledge too high. The thought here is that, in general, raising and lowering standards requires more than just willing it to be so. That being the case, contextualists can posit conversational mechanisms that prevent skeptical standards from coming into play even in philosophical contexts. A second option open to contextualists is to posit ‘gaps’ in the truth-values of knowledge claims, so that competing standards in philosophical contexts result in knowledge claims being neither true nor false.13
DeRose is surely right to point out these options for contextuaism. But many philosophers will not be satisfied. On the ‘Gaps View’, our knowledge claims continue to come out not true relative to philosophical contexts, although skeptical claims come out not true as well. Some will think that this still concedes too much. What about the ‘Veto Power View’? Here our knowledge claims come out true and skeptical claims come out false, relative to both ordinary and philosophical contexts. However, it is not clear that this view still gives us a contextualist response to the skeptical argument. The real anti-skeptical work, it would seem, will require a theory of knowledge that explains why skeptical claims are false, and how non-skeptical claims can be true, across the board.