The Disappearance of Ignorance

Keith DeRose’s new book The Appearance of Ignorance (TAI) is a welcome companion volume to his 2009 book The Case for Contextualism (TCC). Where TCC focused on contextualism as a view in the philosophy of language, TAI focuses on how contextualism contributes to our understanding of (and solution to) some perennial epistemological problems, with the skeptical problem being the main focus of six of the seven chapters. DeRose’s view is that a solution to the skeptical problem must do two things. First, it must explain how it is that we can know lots of things, such as that we have hands. Second, it must explain how it can seem that we don’t know these things. In slogan form, DeRose’s argument is that a contextualist semantics for knowledge attributions is needed to account for the “appearance of ignorance”—the appearance that we don’t know that skeptical hypotheses fail to obtain. In my critical discussion I will argue inter alia that we don’t need a contextualist semantics to account for the appearance of ignorance, and in any case that the “strength” of the appearance of ignorance is unclear, as is the need for a philosophical diagnosis of it.


Introduction
Keith DeRose's new book The Appearance of Ignorance (henceforth TAI) is a companion volume to his 2009 book The Case for Contextualism (henceforth TCC). Where TCC focused on arguments for and against contextualism drawn from the philosophy of language, TAI focuses on arguments for and against contextualism drawn from epistemology, with the skeptical problem being the main focus in six of the seven chapters. As one would expect, there is a degree of overlap here; as DeRose is keen to emphasize, it is hard to keep the philosophy of language out of epistemology. But it is safe to say that the reader expecting a deeper degree of engagement with epistemological issues than in TCC will not be disappointed. Taken as a whole, TAI is a salutary reminder that DeRose's contributions to 20 th and 21 st -century epistemology are hardly confined to a thesis about the meaning of the word 'knows'.
I am aware of two models for writing a critical commentary on a substantial piece of work like TAI. The first involves highlighting some central claims without making much of an attempt to show how they fit together, and then subjecting each claim to individual critical scrutiny. The second involves reconstructing the central line of argument and then subjecting that argument to sustained critical scrutiny. Each model has its benefits and costs. The first model allows for in-depth engagement with the nitty gritty of the target work, but runs the risk of missing the wood for the trees. The second model focuses on the wood at the expense of the trees. For better or worse, this commentary will follow the second model. I will start by outlining what I take to be DeRose's central argument in TAI. In slogan form, DeRose's argument is that a contextualist semantics for knowledge attributions is needed to account for the "appearance of ignorance"-the appearance that we don't know that skeptical hypotheses fail to obtain. I will then argue inter alia that we don't need a contextualist semantics to account for the appearance of ignorance, and in any case that the "strength" of the appearance of ignorance is unclear, as is the need for a philosophical diagnosis of it.

The Two Tasks for a Theory of Knowledge
The central aim of TAI is to deal with what DeRose calls the "argument from ignorance" (AI).
Let "O" be a proposition about the external world one would ordinarily take oneself to know (e.g. "I have hands", "That is a zebra") and "H" a suitably chosen skeptical hypothesis (e.g. "I am a handless brain in a vat (BIV)", "That is a mule cleverly disguised to look like a zebra").
AI can be stated as follows: 1.
I don't know that not-H.

2.
If I don't know that not-H, then I don't know that O.

I don't know that O.
While DeRose ultimately wants to reject AI, he thinks that it is worth taking seriously. Here he is making the case for the plausibility of each premise of AI: AI does present us with a puzzle because … each of its premises is initially plausible, when H is well chosen. For however improbable or bizarre it may seem to suppose that I am a BIV, it also seems that I don't know that I'm not one. We have something approaching (though maybe not quite) a paradox: two plausible premises entail a conclusion that we find incredible. While DeRose sets out to resolve this nearparadox, not just any old resolution will do: [W]e should hope for a better treatment of the argument than simply choosing which of the three individually plausible propositions-the two premises and the negation of the conclusion-seems least certain and rejecting it on the grounds that the other two are true.
In seeking a solution to this puzzle, we should seek an explanation of how we fell into this skeptical trap in the first place, and not settle for making a simple choice among three distasteful ways out of the trap. We must explain how two premises that together yield a conclusion we find so incredible can themselves seem so plausible to us. (p. 2) DeRose is therefore wary of any solution to the skeptical problem that merely consists in rejecting one of the two premises. Any such solution fails to explain why AI seems so plausible in the first place, and so doesn't explain the appearance of ignorance. So, for DeRose, there are two tasks a solution to AI must accomplish. It must, first, tell us how the conclusion of AI is to be resisted and, second, explain the appearance of ignorance.
This conception of the skeptical problem puts a constraint on the appropriate methodology for solving it. It isn't enough to come up with a theory of knowledge, motivate it via non-skeptical considerations such as fit with intuitions about ordinary cases of knowledge, then apply it to AI, yielding the result that one of the premises is to be rejected. This would complete the first task of telling us how the conclusion of AI is to be resisted, but it wouldn't even begin on the second task of explaining why both premises-including the one to be rejected-seem so plausible. 2

DeRose's "Picture" of Knowledge
In this section, I will outline how I take DeRose to think the first task should be carried out. It will be helpful to work with this concrete instance of AI, which I'll call BIV: 1. I don't know that I'm not a handless BIV.

2.
If I don't know that I'm not a handless BIV, then I don't know that I have hands.
3. I don't know that I have hands.
One way of resisting the skeptical conclusion of BIV would be to develop a theory of knowledge on which one of the premises is false. On this approach, the aim would be to come up with necessary and sufficient conditions on knowledge (or at least approximations to such conditions) according to which I do know that I have hands. But this way of going has the drawback that most (or all) candidate sets of necessary and sufficient conditions seem subject to counter-example. DeRose's response to this is to invoke a distinction between "theories" and "pictures". In Naming and Necessity, Kripke doesn't see himself as defending a new theory 6 of reference but rather as presenting a new picture of reference. The problem with theories is that, as Kripke tells us, they tend to be false. The nice thing about pictures is that they are more SAFETY: One's belief that p is safe iff there are no nearby possible worlds in which one believes that p but p is false.
DOUBLE SAFETY: One's belief that p is doubly safe iff there are no nearby possible worlds in which one's belief as to whether or not p is the case fails to match the fact of the matter.
Imagine that, for whatever reason, Catriona is very prone to abandoning her belief that she has hands when she takes drugs, and she takes drugs all the time. But she has no tendency to falsely believe that she has hands. Catriona's belief is safe, because there are no nearby worlds in which she believes that she has hands even though she doesn't. But her belief isn't doubly safe, because there are plenty of nearby worlds in which she has hands yet doesn't believe that she does (worlds in which she took drugs). So DeRose's double safety account would say Catriona doesn't know that she has hands, even though her belief is safe.
As is familiar, the safety condition is the basis of a simple account of how it is that we can know common-or-garden propositions. Take the common-or-garden proposition "I have a hand.", There are no nearby possible worlds in which I believe that I have a hand yet my belief is false. So my belief is safe. DeRose's double safety account can offer a very similar account of how it is that we can know common-or-garden propositions, with the slight difference that it will rule that people like Catriona don't know that they have hands. Setting aside this minor difference, DeRose's account of how we can know common-or-garden propositions is identical with the standard safety account. 3 As is also familiar, one can extend this account to explain how we can know the denials of skeptical hypotheses (Sosa 1999 how it is that we can know that we're not BIVs (see especially pp. 220-222). 5 These beliefs are vacuously safe precisely because there simply are no nearby worlds in which we are BIVs. 6 This leads DeRose to develop a more complex account of how we can know that we're not BIVs. This account holds that, in order to know that one is not a BIV, one must: i. Truly believe that one is not a BIV; and ii. Doubly safely believe that one is not a BIV; and iii.
Have no good reason to think that one is a BIV.
DeRose bases (iii) on a defense of a form of epistemic conservatism: the epistemic conservative thinks (put roughly) that, absent good reason to think that things are otherwise to how they seem, one is justified in believing what seems to one to be the case. 7 While one could endorse epistemic conservatism across the board, DeRose seems to want to restrict it to whatever "seemings" are relevant to our belief that we're not BIVs (p. 229). So his view is that, absent good reason to think that we're not BIVs, we are justified in believing that we're not. In order to turn this into knowledge, our beliefs that we're not BIVs have to also be (i) true and (ii) 5 It strikes me that DeRose is assuming that the two tasks for a theory of knowledge are a little more onerous than stated in Section 2. A theory of knowledge must not only tell us which premise of AI is to be resisted, it must also do so on the basis of a genuinely explanatory account of what knowledge is.
The problem DeRose sees with a standard safety account of how we know that we're not BIVs is that it offers no explanation at all, because the safety condition is vacuously true in this case. 6 While DeRose is clearly displaying internalist sympathies here, it is important to note that he defends these sympathies rather than simply assuming them.

DeRose's Contextualist Semantics
As I read him, DeRose's substantive picture of knowledge is primarily intended to carry out the first task for a theory of knowledge. It serves to explain how it is that we can know the denials of skeptical hypotheses, and so how the conclusion of AI can be resisted. His contextualist semantics for knowledge attributions is primarily intended to carry out the second task. It serves to explain how it can seem like we don't know much if anything, and so why AI can seem so plausible. So the task for contextualism is to explain the appearance of ignorance.
We can start with the basic idea behind contextualism and contextualist solutions to the skeptical problem. The contextualist holds that the word 'knows' is context-sensitive; it can mean different things in different contexts. In one context (a context governed by ordinary standards), to say "I know that I have hands" might mean that I know by ordinary standards that I have hands. In another context (a context governed by very high standards), to say "I know that I have hands" might mean that I know by very high standards that I have hands.
Because I may know by ordinary standards but not by very high standards, both of these utterances might be true. The contextualist also holds that, while the word 'knows' is contextsensitive, ordinary speakers are often ignorant of this fact (in much the same way that ordinary speakers are often ignorant of the fact that words like 'flat' and 'empty' are context-sensitive). 9 This yields the basic contextualist solution to BIV. The contextualist says that, in some contexts-contexts governed by very high standards-the first premise is true. Because the second premise-"If I don't know that I'm not a handless BIV, then I don't know that I have hands"-is true in all contexts, it follows that, in contexts governed by very high standards, the conclusion is true. 10 That is, in such contexts it is true to say that we don't know that we're not handless BIVs, and that we don't know that we have hands. But in other contexts-contexts governed by more ordinary standards-the first premise is false. Indeed, in such contexts it is true to say that we do know that we're not handless BIVs, andthat we have hands.
Before moving on to the ways in which DeRose goes beyond the basic contextualist solution to BIV, it is worth noting that the basic contextualist solution already does quite a good job of explaining the appearance of ignorance. Because ordinary speakers are often ignorant of the fact that the word 'knows' is context-sensitive, it is no surprise that considering arguments like BIV (and more generally arguments of the form AI) can lead them to think they know very little, if anything. DeRose's development of the basic contextualist solution builds on this. The central idea of his solution is contained in this passage (from Chapter 1, which is a reprint of his 1995 classic, "Solving the Skeptical Problem"): When it is asserted that some subject S knows (or does not know) some proposition P, the standards for knowledge (the standards for how good an epistemic position one must be in to count as knowing) tend to be raised, if need be to such a level as to require S's belief in that particular P to be sensitive for it to count as knowledge (p. 27).
DeRose calls this the "rule of sensitivity". The idea is that, when you assert that S knows (or does not know) that p, the standards tend to rise to a level such that S's belief that p must be sensitive, where S's belief that p is sensitive just in case, if p were false, S wouldn't believe that p.
This goes beyond the basic contextualist solution to the skeptical problem in (at least) two respects. First, it provides some insight into the mechanism by which considering skeptical arguments might lead to an increase in standards, and so generate the appearance of ignorance.
When the skeptic asserts "You don't know that you're not a handless BIV", this tends to raise the standards to a level such that your belief must be sensitive in order to count as knowledge, and because your belief is not sensitive-if it were false, you would still believe it-your belief does not count as sensitive in this context.
Second, it provides more of an explanation of the appearance of ignorance. Some (e.g. Nozick) hold that sensitivity is a necessary condition on knowledge. That is, they hold that: SENSITIVITY: S knows that p only if S's belief is sensitive, where S's belief that p is sensitive iff, if p were false, S wouldn't believe that p.
DeRose holds that sensitivity isn't a necessary condition on knowledge, but that there is a conversational rule to the effect that asserting that someone knows (or does not know) some p tends to put into place standards for attributing knowledge such that their belief that p must be sensitive in order to count as knowledge. In slogan form: sensitivity is not a necessary condition for knowledge, but rather for knowledge attribution. Because of the operation of the rule of sensitivity, it is no surprise that we can seem to not know the denials of skeptical hypotheses, and so why considering arguments like BIV can lead us to deny we know much if anything.
Much more could be said about the details of DeRose's contextualist semantics. In particular, one of the main themes in TAI is that contextualism-or at least DeRose's version of it-is far less "even-handed" than many suppose (see especially Chapter 4). More generally, TAI clears up a wide range of misunderstandings, confusions and mistakes that are often made in discussions of contextualism in general, and contextualist solutions to skepticism in particular. But we already have what we need for my purposes. To recap: DeRose thinks that a solution to AI should do two things. First, it should explain how we can know common-orgarden propositions and the denials of skeptical hypotheses which these propositions entail.
Second, it should explain how it can seem like we don't know such things. His substantive picture of knowledge carries out the first task; his contextualist semantics for knowledge attributions the second. Put together, this seems an effective "proof by example" that contextualism is not an "evasion" of epistemology (Kornblith 2000) or irrelevant to epistemological (in particular skeptical) concerns (Sosa 2000). I will now turn to critical evaluation of DeRose's package of views.

Alternative Explanations of the Appearance of Ignorance
DeRose's picture of knowledge combines a standard safety account of how we can know common-or-garden propositions with a slightly more complex epistemic conservative account of how we can know the denials of skeptical hypotheses. To epistemologists of a more externalist bent, this might look a little like overkill: why do we need to add something "extra" to explain how we can know the denials of skeptical hypotheses? But let's set this to one side.
Perhaps the right response to this is to say that, while there may be a form of cognitive achievement that only requires safe belief, there are other forms of cognitive achievement that require a little more (see n. 8). What is more important is that DeRose doesn't think that even the combination of the safety account and the epistemic conservative account is sufficient to deal with the problem posed by AI. This is because he thinks his contextualist semantics for knowledge attributions is required to account for the appearance of ignorance. One can challenge this in two ways. The first is by challenging whether one needs his contextualist semantics to account for the appearance of ignorance. The second is by challenging whether there is much of an appearance to account for in the first place. In this section, I will look at the first challenge. In the next, I will look at the second.
One odd feature of TAI is that, while it spends a lot of time (especially in Chapter 1) arguing that the skeptical "response" to AI (scare-quotes because the response is to just accept the argument and its conclusion) is to be rejected, it spends little time arguing that nonskeptical, non-contextualist responses to AI are inferior to contextualism. This absence is made more troubling by the fact that, as we have seen, DeRose combines his contextualist semantics with a substantive picture of knowledge. This prompts a question: what, exactly, does the contextualist semantics add to the sort of explanation of the appearance of ignorance that can be given by a combination of a standard safety and an epistemic conservative account of knowledge?
We can start by looking at the resources that the standard safety account and the epistemic conservative account have, taken individually, to account for the appearance of ignorance. In "How to Defeat Opposition to Moore," Ernest Sosa argues that, not only does the standard safety account have the resources to explain how it is that we know the denials of skeptical hypotheses (the first task for a theory of knowledge), it can also explain the appearance of ignorance (the second task). 11 The basic thought is that, while knowledge requires safety rather than sensitivity, it is very easy to confuse safety with sensitivity. It is therefore not difficult to see how a skilled skeptic could lead us to deny that we know the denial of skeptical hypotheses (or for that matter to deny that we know anything at all) by appealing to reasons for thinking that our beliefs are insensitive. It takes skilled philosophical diagnosis to reveal that knowledge requires safety, not sensitivity, and so the skeptic's reasons should be rejected.
To Sosa's explanation we can add the following: the standard safety account is an externalist account of knowledge, and to the extent that skeptical arguments rely on internalist assumptions and/or sympathies (e.g., that there is something lacking from our epistemic position if we lack an awareness of the facts in virtue of which our beliefs are justified/qualify as knowledge, or that in order to know/be justified you have to be able to provide reasons in support of your belief), it is no surprise that many of us (especially those of us who haven't read the recent epistemological literature) can be brought by skeptical argument to question our beliefs. So the externalist seems to have a more general explanation of the appearance of ignorance, which the defender of the standard safety account can appeal to.
What about the epistemic conservative account? I am not aware of a "standard" epistemic conservative explanation of the appearance of ignorance and the "allure" of skepticism, but let me venture a suggestion. At the base of the epistemic conservative pictureat least, as DeRose understands it-is an appeal to what we can call "Reidian common sense." Here's Thomas Reid articulating what I mean by Reidian common sense: The sceptic asks me, Why do you believe the existence of the external object which you perceive? This belief, sir, is none of my manufacture; it came from the mint of Nature; it bears her image and superscription; and, if it is not right, the fault is not mine: I even took it upon trust, and without suspicion. Reason, says the sceptic, is the only judge of truth, and you ought to throw off every opinion and every belief that is not grounded on reason.
Why, sir, should I believe the faculty of reason more than that of perception?-they came both out of the same shop, and were made by the same artist; and if he puts one piece of false ware into my hands, what should hinder him from putting another? (Reid 1764: 168-169) On this picture, certain beliefs-like the belief that we're not being constantly deceived by our senses-are not supported by reasons, but rather are parts of our nature. The skeptic's mistake is to ask for reasons for them, as if they need to be given.
Perhaps the epistemic conservative is right that such reasons don't need to be given, and so right that the skeptic's argument should not move us to abandon our most basic beliefs.
Yet, if DeRose is right, we can be moved to at least consider abandoning them due to skeptical argument. (If DeRose doesn't think we can be moved to at least consider abandoning them due to skeptical argument, I don't know what the appearance of ignorance amounts to.) It would seem to me that the Reidian commonsensical picture provides a pretty good explanation why we might be drawn to take these arguments seriously, even though we have no reason to: we have nothing to pit against them other than our trust that their conclusions are false. Now, it may be that we can mount a philosophical defense of the propriety of pitting this trust against the skeptic's arguments. But it would hardly be a surprise if many of us, lacking any awareness of that defense, were inclined to go along with the skeptic. So it seems like the epistemic conservative account can do a pretty good job of explaining the appearance of ignorance all by itself. The appearance of ignorance is generated by the sorts of suspicions about the propriety of trusting to "our nature" or "constitution" that philosophical defenses of epistemic conservatism are meant to dispel.
I take this to at least make it plausible that, taken individually, the defender of the standard safety account and the defender of the epistemic conservative account have ample resources to explain the appearance of ignorance. If we combine these accounts, as DeRose does, we get the power of both these explanations. Indeed, we may get more than that. One can see DeRose's combination of the standard safety account and the epistemic conservative account as in part motivated by the felt inadequacy of either account taken in isolation. The standard safety account seems to run into trouble when it comes to justifying our most basic beliefs (consider the problems of easy knowledge, and bootstrapping). 12 The epistemic conservative account seems too weak as an account of knowledge in general: surely knowledge requires a little more than the absence of reason to doubt! For someone who hasn't considered the possibility of combining these views, whether because they think they can't be combined, or because they have never considered doing so, it is, again, no surprise that skeptical argument might take hold, and the appearance of ignorance be generated.

The Disappearance of Ignorance
So far I have argued that the non-contextualist part of DeRose's overall picture might be enough to explain the appearance of ignorance. In this section, I want to say something about the appearance of ignorance itself. In particular, I want to address three issues: problem, in short, is that if the appearance of ignorance isn't that widespread, robust, or rational, why should a solution to AI have to account for it?
We can start with DeRose's take on how widespread the appearance of ignorance is.
DeRose presents a decidedly mixed picture here. On the one hand, he tells us that, in his experience, undergraduates can very easily be led to see the power of AI (pp. 55-56). On the other, he reports a study he carried out with Joshua Knobe in which ordinary adults didn't generally find AI particularly persuasive (Appendix B). More generally, DeRose seems willing to grant that, for many, AI holds little power at all, whether intuitive or not. But, DeRose tells us, this doesn't matter: It is important to emphasize … that not only does the value of studying skeptical arguments not depend on their being so powerful as to truly constitute "paradoxes," but that these arguments can be well worth studying even if one's initial reaction to them places one toward the extreme Kelly end of the continuum at which one judges them to be in- The idea seems to be that, to the extent that AI is powerful, that power needs to be explained.
Turning now to how robust the appearance of ignorance is, DeRose says this about the durability of the appeal of AI's first premise (the premise that says we don't know the denials of skeptical hypotheses), which he takes to be the weak link in the argument: My attempts to ascertain non-philosophers' assessments of what I take to be AI's weak link have if anything increased the extent to which I thought an account of that premise's appeal was desirable. That there are settings where that appeal seems very strong would tend to show that the premise (and then the argument of which it is the weakest link) can be very powerful, even if there are other settings where that appeal is quite diminished.
(And perhaps when it is presented by an actual skeptic, or a teacher doing a good job of playing the role of a skeptic, its power is significantly greater still than it is when presented by a philosophy professor who is trying to be as neutral as possible about the matter. (p.

59)
DeRose's view seems to be that even those who do see the appeal of skeptical arguments only see them as appealing in certain settings or contexts. While some of us can be brought to see the argument as very powerful in some settings, in others this power is rather more limited.
Finally, what is DeRose's take on the rationality of going along with the skeptical argument? Throughout TAI (and especially in Chapter 4) DeRose goes to great lengths to dispel the common idea that contextualists propose an "even-handed" solution to the skeptical problem according to which both the skeptic and the anti-skeptic are right "in their own context." While many interpret the contextualist in this way, DeRose's considered view is that, while the skeptic makes conversational moves intended to put very high standards into place, standards such that all (or almost all) of our beliefs fail to meet them-perhaps by exploiting DeRose's rule of sensitivity-there is no need for anyone to go along with this. Skeptics may try to raise the standards, but we can-and maybe should-resist. 13 This rejection of "evenhandedness" presumably goes with a denial that there is any reason to go along with the skeptic who wants to raise the standards, if not the view that it would be positively irrational to go along with the skeptic.
Putting all this together, DeRose holds the appearance of ignorance isn't that widespread, robust, or rational. Somewhat surprisingly, he doesn't seem to see this as a problem: The Chapter 4 of TCC, which is itself a reprint of DeRose (2004a). One can understand his frustration that his position is still so regularly misinterpreted! sorts of arguments for views in epistemology that fail to establish their conclusions. Take, for instance, arguments to the effect that knowledge doesn't require belief (Radford 1966), or that knowledge just requires true belief (Sartwell 1992), or that knowledge (or "knowledge") isn't factive (Kusch 2002, Hazlett 2010. It is unclear why there need be any requirement to explain away any lingering misleading impressions that these conclusions are true. This makes the task for non-contextualist anti-skeptics all the easier. Maybe it is enough to have an explanation of how it is that we know so much without also having much of an explanation of how it is that we can seem to not know anything at all. Maybe we can seem to not know anything at all for much the same reason that it can seem that knowledge doesn't require belief, or only requires true belief, or that knowledge isn't factive.

Concluding Remarks
DeRose's central argument in TAI is that there are two things that a solution to the skeptical problem must accomplish. First, it must explain how we can resist AI's skeptical conclusion: how is it that we can know common-or-garden propositions? Second, it must explain the appearance of ignorance: how is it that it can seem that we don't know common-or-garden propositions, or indeed anything at all? I have argued that one part of DeRose's solution-his substantive picture of knowledge-threatens to render the other part-his contextualist semantics for knowledge attributions-explanatorily redundant. We can explain any appearance of ignorance via the picture of knowledge, without recourse to contextualism.
Further, the appearance of ignorance is-as DeRose seems to freely admit-not necessarily that widespread, not necessarily that robust, and not particularly rational (and perhaps downright irrational DeRose is doing philosophy of language rather than epistemology. Rather, the complaint is that he does so much epistemology his philosophy of language might look a little redundant.