1. THE SKEPTIC'S ARGUMENT

Nichols, Stich, and Weinberg (henceforth NSW) focus on what they call Cartesian skeptical arguments. These arguments are characterized by their use of ‘skeptical hypotheses’, which describe purportedly undetectable, cognitively debilitating states such as dreaming, hallucination, and victimization by the Cartesian demon. Roughly, a hypothesis is ‘skeptical’ if (a) its truth is inconsistent with some propositions we ordinarily take ourselves to know, and yet (b) the hypothesis is compatible with all our evidence for those ordinary propositions (NSW Reference Nichols, Stich, Weinberg and Stich2012: 225).Footnote ¹ To illustrate, consider the following brain-in-a-vat version of the skeptic's argument:

1. 1. I don't know that I am not a handless brain in a vat.

2. 2. If I don't know that I am not a handless brain in a vat, then I don't know that I have hands.

3. 3. Therefore, I don't know that I have hands.

Although Descartes never mentions brains in vats, the origin of this argument can be traced to his Meditations, if not earlier. This argument has also played an important role in the work of some leading 20th Century philosophers, including Bertrand Russell and G.E. Moore. Many contemporary epistemologists, such as Keith DeRose (Reference DeRose, DeRose and Warfield1999: 2), Stephen Schiffer (Reference Schiffer1996: 317), and Stewart Cohen (Reference Cohen1999: 62), provide essentially the same formulation of skepticism. Suitably articulated, the skeptic's argument will lead us to deny much of our putative knowledge of the world around us.

By abstracting away from the details, we can reach a more general formulation of skepticism. For simplicity, let's say that O represents some ordinary proposition about the external world that we think we know and H represents a suitably chosen skeptical hypothesis that is inconsistent with O. The general structure of the skeptic's argument is:

1. 1. I don't know that not-H.

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

3. 3. Therefore, I don't know that O.

Let's call this the skeptic's argument. DeRose thinks it is “clearly valid … and each of its premises, considered on its own, enjoys a good deal of intuitive support” (Reference DeRose, DeRose and Warfield1999: 2–3). Cohen agrees that “both of these premises are intuitively quite appealing” (Reference Cohen1999: 62).

A variety of epistemologists have argued that our inclination to ascribe knowledge decreases when skeptical hypotheses are raised.Footnote ² Barry Stroud maintains that skepticism “appeals to something deep in our nature and seems to raise a real problem about the human condition” (Stroud Reference Stroud1984: 39). Colin McGinn takes skepticism to be a universal feature lurking in human thought (McGinn Reference McGinn1993: 107–8). Presumably skepticism is enduring and significant because many people find the skeptic's reasoning intuitive. In the Meditations, for instance, Descartes appeals to the ease with which we acknowledge that the Meditator must know that he is not dreaming if he is to know that he is sitting by the fire. We empathize with the Meditator and recognize that we share his epistemic handicap.

The skeptic's argument easily generalizes. Not only can we use it to show that some person S does not know some ordinary proposition O, we can use it to show that nobody knows anything (or virtually anything). Considering a skeptical hypothesis can quickly lead to rampant skepticism: we do not know various ordinary things because we cannot eliminate (or at least have not eliminated) defeating counter-possibilites that are consistent with our evidence.

The brand of skepticism that makes use of skeptical hypotheses had been widely discussed in recent epistemology, and it is this type of skepticism that NSW target. The next section will outline their meta-skeptical challenge.

2. THE META-SKEPTICAL ARGUMENT

According to NSW, the skeptic treats the premises of her argument as intuitively obvious and therefore she offers little or no defense for them. Following NSW, let's call any pre-theoretical belief according to which we do not know the falsity of a skeptical hypothesis a skeptical intuition. (For instance, you have a skeptical intuition if it seems to you that you do not know that you are not a brain in a vat.) I will grant to NSW that in order for skepticism to get its teeth, skeptical intuitions must be widely shared.Footnote ³ The ‘brain in a vat’ argument wouldn't be compelling if most people simply denied the first premise of the argument. If we think we know that we are not handless brains in vats, then we wouldn't be taken in by the skeptic's reasoning. More generally, if people do not have skeptical intuitions, then skepticism is not as threatening as many philosophers have thought.

NSW attack the skeptic's argument on these grounds. Their empirical research suggests that skeptical intuitions are “far from universal” (NSW 2012: 224). In particular, NSW show that many of the skeptical intuitions invoked by epistemologists vary with factors like cultural background and level of education. This raises a significant problem for skeptical arguments: if the intuitions needed to motivate skepticism are not widely shared, then “skepticism may be much less interesting and much less worrisome than philosophers have taken it to be” (NSW 2012: 224).

Using several vignettes, NSW tested whether or not experimental subjects were willing to ascribe knowledge to someone who has been confronted with a skeptical hypothesis. One of their cases involves two college roommates, neither of whom is a brain in a vat, discussing the brain in a vat scenario. One of the roommates, named George, appeals to certain perceptions he has in order to justify his belief that he is not a brain in a vat. The participants were asked whether George “really knows” or “only believes” that he is not a brain in a vat. NSW report that 80% of those participants with more philosophical training denied knowledge (“only believes”), whereas only 45% of those with less philosophical training denied knowledge (2012: 241). This indicates that people with less exposure to philosophy are more likely to claim that George knows he is not a brain in a vat, and presumably are more willing to deny the first premise of the skeptic's argument. In contrast, people with more philosophical training are more likely to have skeptical intuitions.

Another case used to test for skeptical intuitions was based on a scenario described by Fred Dretske (Reference Dretske1970):

Although the majority of participants maintained that Pat “only believes” the animal is a zebra, people with a lower socio-economic status (measured by level of education) were significantly more likely to say that Pat “really knows” (31%) than people with a higher socio-economic status (11%). This suggests that skeptical intuitions vary across different socio-economic groups. Those with a lower socio-economic status are more willing to deny the skeptic's conclusion than people with a higher socio-economic status.

In their cross-cultural studies, NSW presented their participants with another variant of Dretske's zebra case:

Using this case, NSW found a significant difference between Western and Indian Subcontinent participants. More precisely, only 32% of Westerners said that Mike “really knows” whereas 50% of participants from the Indian Subcontinent were willing to ascribe knowledge to Mike.

What conclusions can be drawn from these studies? These data suggest that whether or not people have skeptical intuitions partly depends on factors such as level of philosophical training, socio-economic status, and cultural background. This is bad news for the skeptic, according to NSW, because her argument relies crucially on skeptical intuitions. If many people do not have skeptical intuitions, there will be widespread disagreement about whether the skeptic's argument is plausible. If many people reject one (or both) of the premises in the skeptic's argument, then skepticism is much less interesting and much less worrisome than philosophers have claimed. Thus, NSW conclude that their data should make us question the central place that debates about skepticism have occupied in Western philosophy.

3. A CRITIQUE OF NSW'S META-SKEPTICAL ARGUMENT

How might we reply to the meta-skeptical argument? One response is to say that people who do not have skeptical intuitions are using the word ‘know’ incorrectly, so their judgments have no bearing on whether skepticism is correct. However, it seems arbitrary – if not question begging – to simply assume that people who lack skeptical intuitions are misusing the word ‘know’. Why privilege one set of intuitions over another?Footnote ⁴ Another response is to say that people who do not share our intuitions (whichever those happen to be) have a different concept of knowledge, so their judgments have no bearing on whether skepticism is correct from the standpoint of our epistemic judgments.Footnote ⁵ While this view might ultimately be correct (I leave that open), it does not work as a reply to NSW's meta-skeptical argument. This type of conceptual relativity would reduce skepticism to the view that we cannot ‘know’ facts about the external world only according to the concept of knowledge possessed by some (possibly small) group. If this were correct, then skepticism really would be much less interesting and much less worrisome than philosophers have thought. Skepticism would only be threatening from the standpoint of some group's epistemic intuitions; but epistemology typically aspires to general conclusions, not local ones (see Hannon Reference Hannon2015).

I will propose a different way to resist NSW's meta-skeptical argument. The best way to introduce my strategy is to reformulate NSW's reasoning, which can be summarized as follows:

The Meta-skeptical Argument

1. 1 If the intuitions that are crucial to the skeptic's argument are not widely shared, then skepticism is much less interesting and much less worrisome than philosophers have thought.

2. 2 Skepticism relies crucially on the intuitions tested by NSW.

3. 3 The intuitions tested by NSW are not widely shared.

4. 4 Therefore, skepticism is much less interesting and much less worrisome than philosophers have thought.

I will grant the first and third premise, as well as the argument's validity.Footnote ⁶ What I'll deny is the second premise. Skepticism does not rely on the intuitions tested by NSW. As a result, NSW miss their intended target, and so their meta-skeptical argument fails.

My argument begins with a brief story. Suppose you are teaching Epistemology 101 and that you are discussing Descartes's Meditations. You present the class with Descartes's evil demon hypothesis and then articulate the following skeptical argument:

1. 1 You don't know that you are not deceived by an evil demon.

2. 2 If you don't know that you are not deceived by an evil demon, then you don't know that you have hands.

3. 3 Therefore, you don't know that you have hands.

In my experience, a sufficiently large classroom of undergraduates will rarely meet this argument with unanimous approval. Drawing on NSW's data, we may reasonably suppose that a non-trivial number of students will not find this argument immediately compelling. Of those students who do not find this argument compelling, imagine that one of them exclaims, “Come on, Professor! I know that I am not deceived by an evil demon.” According to NSW, this student does not have a skeptical intuition. As a result, she rejects the skeptic's first premise (as outline in §1). Does this imply that skepticism relies on a dubious assumption (for this student)?

If you are like me, you will try to get this student to see the force of the skeptic's argument. You will probably do this by Socratic questioning. For instance, you could ask: “How do you know that you aren't deceived? What evidence could you provide us with? Wouldn't you think you weren't deceived even if you in fact were?” And so forth. With a little extra work, many resilient students eventually – and perhaps grudgingly – concede they don't really know they're not deceived (or at least they come to appreciate the difficulty in maintaining knowledge in a non-dogmatic way). The student might also try to resist the second premise. She might say: “Okay, I might not know that I'm not deceived, but of course I know that I have hands!” Again, with a bit of effort you can probably make this student uneasy about her claim. You might ask: “What is your evidence for having hands? Aren't your perceptual experiences consistent with the possibility that you are deceived? If so, why think that your perceptual experiences count as evidence for having hands? How could you know that you have hands if you don't know that you're not a brain in a vat?” And so on.

The length of questioning will of course depend on a combination of how resilient the student is and how incisive the inquirer is. But it is plausible, and certainly true in my experience, that after a sufficient amount of Socratic questioning, skepticism will start to look genuinely puzzling – or at least not obviously false – to people who initially reported lacking skeptical intuitions. Of course, resilient students will not necessarily be led to endorse skepticism (few of us do), but the argument will now seem credible, puzzling, and philosophically significant to many of them.Footnote ⁷

Let me explain how the Epistemology 101 example reveals the central flaw in NSW's meta-skeptical argument. NSW present data showing that many people confronted with a skeptical hypothesis will continue ascribing knowledge. In other words, for many people the ascription of knowledge does not become more stringent once a skeptical hypothesis is made salient. The flaw in NSW's argument, however, is this: the mere fact that some people do not immediately become less willing to ascribe knowledge does nothing to undermine skepticism. Why not? It is because the skeptic's argument does not rely on people's immediate judgments about cases. Thus, contrary to the second premise of the meta-skeptical argument outlined above, skepticism does not rely crucially on the intuitions tested by NSW.

It will be useful for our discussion to distinguish between two broad types of intuitions: surface intuitions, which are first-off judgments, and robust intuitions, which are judgments we hold after some reflection or consideration.Footnote ⁸

Admittedly, this is a rough distinction. But whatever the merit of this distinction, the skeptic's argument certainly depends on something more than just our immediate reactions to scenarios involving skeptical hypotheses. It is robust intuitions, not surface ones, that matter here. When a philosopher says competent speakers would use ‘knows’ in certain ways, she is not necessarily predicting that such speakers will respond in these ways off the cuff. The fact that a student in your Epistemology 101 class is not immediately perturbed by a skeptical hypothesis does not threaten the significance of skepticism because no premise in the skeptic's argument relies essentially on this type of intuition. What makes skepticism tempting is that many people who are prodded to think reflectively about knowledge tend to find the skeptic's argument intuitive and difficult to refute. The driving force behind skepticism is our tendency to find the premises in the skeptic's argument plausible, but whether we find them immediately plausible is inessential. What matters is that people can be brought to feel the “tug” of skepticism. For simplicity, I will call the intuitions on which skepticism depends ‘robust skeptical intuitions’.Footnote ⁹

One might ask, in what sense are robust intuitions intuitions? The term ‘intuition’ is used in both ordinary language and philosophy with a variety of extensions. While my focus is on the role of intuitions in distinctively philosophical inquiries, there is nevertheless little agreement about how to characterize these intuitions. Some philosophers equate them with beliefs; for example, David Lewis writes,

Others, such as Peter van Inwagen (Reference van Inwagen1997: 309), claim that one has an intuition that p if one is disposed to believe p. By contrast, George Bealer (Reference Bealer, DePaul and Ramsey1998) claims that intuitions are marked by special phenomenal characteristics, whereas Antti Kauppinen (Reference Kauppinen2007) argues that philosophical intuitions are those mental states subjected to critical examination as part of reflective participation in traditional philosophical discourse.

Whether or not philosophically relevant intuitions constitute a single epistemic and psychological kind is not my concern. There is a range of conceptions, from thin conceptions that treat intuitions as generic kinds of mental states (i.e., beliefs) to thick conceptions that place additional constraints (see Weinberg and Alexander Reference Weinberg, Alexander, Booth and Rowbottom2014). This plurality of conceptions raises a question: which intuitions are required to motivate skepticism? As I've argued, skepticism does not rely crucially on the intuitions elicited by NSW's experimental study. In saying that the skeptic's argument is ‘intuitive’, philosophers are not appealing to gut reactions about a specific case but rather to pre-theoretical judgments that require reflection. Skepticism is threatening because people can be brought, via Socratic dialogue, to worry about it.Footnote ¹⁰

This does not seem to be the notion of an intuition that NSW have in mind. They define an ‘intuition’ as “a spontaneous judgment about the truth or falsity of a proposition” (2012: fn. 2). This characterization seems befitting of what I am calling a ‘surface’ intuition. However, this characterization of an intuition is somewhat in tension with NSW's claim that the intuitions undergirding skeptical arguments are supposed to be “shared by everyone (or almost everyone) who thinks reflectively about knowledge” (2012: 224, emphasis mine). This latter characterization is closer to what I am calling a ‘robust’ intuition. Thus, we have on the table two broad conceptions of intuitions: spontaneous (or surface) intuitions and reflective (or robust) intuitions. If NSW contend that skeptical intuitions are the product of thinking reflectively about knowledge, then they seem to agree that robust intuitions must (or at least can) be part of the evidence for skepticism. If, however, they believe the evidence for skepticism must be comprised of surface intuitions, they are mistaken.

This is not to say that skepticism cannot be motivated by surface intuitions. As the data presented by NSW show, many people from distinct groups are immediately willing to deny knowledge when a skeptical hypothesis is made salient (e.g., Westerners, people with philosophical training, etc.). I am not claiming that robust intuitions are necessary to motivate skepticism; in fact, surface intuitions might be sufficient to render skepticism interesting and worrisome for people who have those intuitions. My claim is that surface intuitions are not a necessary component of the skeptic's argument.

To illustrate this, imagine a world in which people are initially willing to attribute knowledge in the face of skeptical hypotheses, but also these people can be led to feel skeptical worries after some further questioning and reflection. In other words, imagine these people do not have surface skeptical intuitions but they have robust skeptical intuitions. This is clearly a world in which skepticism is interesting and worrisome. The fact that the inhabitants of this world are not immediately willing to deny knowledge does not diminish the significance of skepticism. Similarly, the threat of skepticism is not lessened by the mere fact that some students in Epistemology 101 do not find the skeptic's argument immediately tempting. Analogously, the fact that some survey participants are not immediately willing to deny knowledge in the context of a survey does not show that skepticism is any less significant than philosophers have thought. The skeptic's argument does not crucially rely on the intuitions tested by NSW, despite their claim to the contrary. It is more plausible that skepticism can be motivated in a variety of ways.

Before concluding this section, let me briefly mention another direction my argument might be taken. So far my discussion has been framed in terms of different kinds of intuitions: I have argued that skepticism does not crucially rely on the type of intuitions tested by NSW, but rather on another sort (which, following Kauppinen, I have called ‘robust’ intuitions). However, this might actually be underselling my central point. Instead of capturing the idea in terms of different types of intuitions, it might be better to say that skepticism does not rely on intuitions at all; rather, it relies on the experience of epistemic worry. The experience of epistemic worry might be more affective than intellectual, and it needn't manifest as anything even remotely like an intuition that some particular proposition is true.Footnote ¹¹

This proposal allows for a more expansive discussion of the question of just what is required for skepticism to get a hold of us; however, I will not pursue this topic here. Although I have claimed that skepticism is ‘intuitive’ on the grounds that people can be brought, via Socratic dialogue, to worry about skepticism, my argument does not turn on framing the issue this way. My main point is that skepticism does not rely crucially on the intuitions tested by NSW. In what follows I will continue using the distinction between ‘surface’ and ‘robust’ intuitions, though I recognize that the main points in this paper could be reformulated in terms of the experience of epistemic worry rather than having a robust skeptical intuition.

4. OBJECTIONS AND REPLIES

In response to my argument, NSW might say, “But surely there is something about the psychology of our participants that explains why only certain groups of people are willing to attribute knowledge in the cases we have tested.” This is true as far as it goes. My argument does not try to show that the results obtained by NSW do not call for an explanation of some sort. My point is that their results do not diminish the threat or significance of philosophical skepticism, contrary to the conclusion of their argument. Explaining why some of their survey participants are less willing to ascribe knowledge is not needed to undermine their meta-skeptical argument.

NSW might also claim they are only concerned with versions of skepticism that do in fact rely on surface intuitions. But this would commit them to attacking a straw man, which would trivialize their meta-skeptical argument. No popular, intriguing, and compelling form of skepticism relies essentially on premises that people must find immediately plausible. All that matters, as I've argued, is that people can be led (e.g., via Socratic questioning) to feel the pull of the skeptic's reasoning.

This of course raises the question of whether or not people can generally be led to feel skeptical anxiety as a result of Socratic questioning. NSW might argue that absent any specific evidence to the contrary, the most likely hypothesis is that a difference in surface intuitions will be more or less mirrored upon reflection. Thus, merely pointing out that they did not test the results of ‘Socratification’ is not enough for me to show the skepticism really is important and worrisome.

I will use the label ‘Skeptical Anxiety Hypothesis’ to refer to the hypothesis that people can generally be led to deny knowledge as a result of skeptical considerations. The question at hand is: how confident should we be that the Skeptical Anxiety Hypothesis is correct? At present, this is an open empirical question. Earlier I drew on my experience teaching Epistemology 101 to support this hypothesis, but I do not want to rest too heavily on these reflections. (Indeed, this strategy has gotten armchair philosophers in trouble with experimental philosophers.) I will briefly discuss some potential concerns with relying on data collected in the classroom; then I will consider some additional reasons in favor of the Skeptical Anxiety Hypothesis.Footnote ¹²

One concern is that I am dealing with a highly self-selected group, namely, students who chose to take a philosophy course. However, this is not a worry in my own case because I teach at a Catholic university where all students are required to take two philosophy courses. But one might still be concerned that there is a selection bias, namely, this is a group of students who chose to attend a Catholic school. I do not find this argument very persuasive. While this might show that I am dealing with a self-selected group, the claim that there is a connection between the choice to attend a Catholic school and the inclination to feel skeptical anxiety seems tenuous. I see no reason to think students attending such a school would be more inclined to have robust skeptical intuitions than students at non-Catholic schools.

Another concern is that my sample of students does not include those who are unmoved by skeptical considerations enough to even bother wanting to debate the professor about it. In other words, there might be a non-trivial number of students who are silent on this topic precisely because they are unaffected by the skeptic's reasoning (in addition to those who are unmoved and yet remain silent for other reasons, such as shyness).

I share this worry. Nevertheless, it is telling that a large number of students can be led to feel the pull of the skeptic's reasoning even in spite of their initial and vocal resistance to skepticism (e.g., the student who says, “Come on, Professor! I know that I am not deceived by an evil demon”). This provides us with some reason to think that if we could get non-vocal students to engage in Socratic dialogue, then they, too, might be led to feel skeptical anxiety.

Putting aside the problem of sample bias, another worry is the risk of an experimenter bias. The ‘subjects’ (the students) might know full well what result the ‘experimenter’ (the teacher) wants from them, and will thus be on average more inclined to give that result than they would if they did not know that.Footnote ¹³

I acknowledge this concern, so I do not want to rely too heavily on my own experience in the classroom. What else can be said in favor of the Skeptical Anxiety Hypothesis?

One casual observation is the wide success gained by certain sorts of sci-fi films and books that importantly turn on inducing skeptical anxiety. As John Turri observes,

Maybe these films would be less popular if audience members weren't led to feel some level of skeptical anxiety as a result of skeptical possibilities. This evidence isn't very conclusive, however. Can anything else explain the persistent and widespread appeal of skepticism?

Perhaps certain psychological factors render us vulnerable to skeptical doubts. Turri (Reference Turri2015) presents evidence that skeptical arguments gain potency from an interaction between two factors: first, people evaluate inferential belief more harshly than perceptual belief (we view perception as a more likely source of knowledge than inference); second, people evaluate inferential belief more harshly when its content is negative (i.e., that something is not the case) than when it is positive (i.e., that something is the case). Thus, the skeptic will likely instill doubt in us if she gets us to focus on an inferential belief with negative content.Footnote ¹⁴

Turri tested this hypothesis using several vignettes, one of which is similar to the zoo scenarios described by NSW. The basic storyline features an avid zoo patron, Michelle, who regularly visits The Big Cat Exhibit, and on every occasion for the past 10 years the animal in this exhibit has been a jaguar. Turri tested the effect of two factors on knowledge denial. First, he varied whether Michelle judges the animal based on perception (i.e., while looking at the animal) or based on an inference unaided by observation (i.e., Michelle can't visit the zoo but she infers from past experience). Second, participants in the Positive condition were asked whether Michelle knows that the animal is a jaguar (as opposed to a leopard); in the Negative condition they were asked whether Michelle knows that it is not a leopard.

The result was that a belief's source (perceptual vs. inferential) and content (positive vs. negative) had a clear effect on knowledge denial in contexts where skeptical possibilities were explicitly raised and considered. Specifically, we see high rates of knowledge denial in cases of negative inferential belief. This helps explain the force of the skeptic's argument. The skeptic can easily get us to focus on what seems like an inferential belief with negative content; for instance, “Do you know that your car has not been stolen today, even though you haven't checked?” As a result of these two factors, we are prone to doubt that such beliefs count as knowledge.

This not only provides us with evidence supporting the Skeptical Anxiety Hypothesis, it also sheds some light on NSW's findings. In the zoo cases they tested, participants were asked to evaluate a perceptual belief with positive content. Since these scenarios did not focus the participant's attention on a negative inferential belief, people were less prone to deny knowledge. In the brain in a vat case, NSW asked participants to evaluate a belief with a negative content, but their vignette stipulates that George's belief is formed on the basis of perception, not inference. By focusing on a perceptual belief rather than an inferential belief, NSW fail to sow seeds of skeptical doubt. As Turri's study reveals, knowledge denial is much higher in cases of negative inference (71%) than negative perception (27%).

Allow me to venture a final explanation for why people would be more inclined to attribute knowledge after reading NSW's vignettes than they would after Socratic questioning. In NSW's survey, each vignette mentions a possibility of error that is stipulated to be unrealized; for example, participants were told that George is not really a brain in a vat and that Pat and Mike are not actually looking at a cleverly painted mule. The readers of these cases were thus supplied with a crucial fact about the situation that the subjects being evaluated (George, Mike, and Pat) did not have: the survey-takers knew the possibility of error was unrealized. However, this is precisely the sort of fact that the skeptic will deny that we know. We simply cannot tell whether or not the actual world is one in which a skeptical hypothesis is realized. In this way, we suffer the same epistemic disadvantage as the characters in NSW's examples. I suspect the survey-takers who attributed knowledge in NWS's study were inclined to do so partly because they evaluated the subject's belief in the light of their own awareness that the subject's world is one where the relevant skeptical hypothesis is false (via stipulation).Footnote ¹⁵ In contrast, when we ourselves are reflecting on what we know (e.g., ‘Is that really a zebra?’), it would beg the question to simply stipulate that the actual world is a non-skeptical world. We lack such information, so we are less inclined to say that we know.Footnote ¹⁶

5. THE STATE OF THE ART AND THE WAY FORWARD

Although I have offered some considerations in favor of the Skeptical Anxiety Hypothesis, settling this question requires more empirical data. Collecting these data goes beyond what can reasonably be done in this paper, so I will conclude with some reflections on how a future study might test this hypothesis.

Suppose Turri has correctly identified certain psychological mechanisms that lead people to deny knowledge. One way to empirically test the Skeptical Anxiety Hypothesis would be to redesign NSW's survey in a way that gets participants to focus on inferential beliefs with negative content. If nearly all survey-takers denied knowledge under these circumstances, this would indicate that people could easily be led to find skepticism intuitive.

However, this strategy is unlikely to fully succeed. Turri's study shows that only 71% of respondents denied knowledge in the example involving a negative inferential belief. This suggests the skeptic will not instill doubt in many people even after getting them to focus on an inferential belief with negative content. Does this indicate that many people cannot be led to feel skeptical anxiety?

I think not. All this would suggest is that more prodding might be needed to elicit robust skeptical intuitions. What sort of prodding? Precisely the sort that characterizes Socratic questioning. For this reason, I doubt the simple survey method used by NSW and other experimentalists can adequately test the Skeptical Anxiety Hypothesis. In order to create the conditions needed to elicit robust intuitions, we can imagine a researcher going through a survey participant's answers together with her, asking for the reasons why she answered one way rather than another, making sure she did correctly understand the scenario described, varying examples, teasing out implications, pointing out similarities and disanalogies with other cases of knowledge, and trying to get the subject to reflect on whether her initial response really is what she wants to say. But, as Kauppinen (Reference Kauppinen2007: 106) points out, then we are no longer doing experimental philosophy. Rather, we are returning to the Socratic method, where one engages in serious dialogue with test subjects. A researcher eliciting robust intuitions would be engaging in a style of questioning similar to that used in my example of Epistemology 101, and this is pretty much how philosophy has always been done.Footnote ¹⁷

Could we design an experiment to elicit robust intuitions? Kauppinen doubts that any amount of experimental ingenuity will allow us to design a study to test whether people have robust intuitions. I think this is overly pessimistic. Although the survey methodology used by NSW is inadequate for empirically addressing this question, there are other methodological tools available that could conceivably be used in a future study. I'll now briefly consider a wider range of possibilities for empirically testing the Skeptical Anxiety Hypothesis.Footnote ¹⁸

One way to proceed would be to stage the right sort of Socratic intervention, record it, and play it for the subjects, then ask the subjects to evaluate the participants in the dialogue. For example, the subjects might listen to a conversation between two interlocutors, Claudia and Brett. Claudia tries to convince Brett that neither of them knows that they have hands, since neither of them can prove that they are not a brain in a vat. At first, Brett might resist Claudia's reasoning; for instance, he might say, “But it's so unlikely that we are brains in vats!” Claudia might reply in the following way: “However unlikely it might seem, you don't know that we aren't. After all, winning the lottery is highly unlikely but you don't know that you'll lose.” After a variety of considerations have been discussed, the dialogue could end in one of two ways: either Brett concludes that he doesn't know that he has hands (scenario 1) or he maintains that he knows that he has hands (scenario 2). Participants who are given scenario 1 would be asked, “Do you agree that Brett doesn't know he has hands?” Participants given scenario 2 would be asked, “Do you agree that Brett knows he has hands?”

Another way to proceed is to hypothesize at greater length as to just what it is that subjects who do not feel skeptical anxiety are failing to grasp, such that once they do grasp it, they will become worried. Once we generate a set of “missing links” that might bring the unworried to feel skeptical anxiety, it would be easy enough to design an experimental instrument for online delivery that would take subjects through those links. For example, some people might not grasp the epistemic closure principle that has been touted as a linchpin in the most compelling skeptical arguments (e.g., Dretske Reference Dretske1970; Stroud Reference Stroud1984). This principle states that, necessarily, if you know that p, and you know that if p then q, then you also know that q (or at least you are in a position to know this). Perhaps some participants do not realize that if they claim to know they have hands, then, plausibly, they must also know they are not a brain in a vat.Footnote ¹⁹ Others might claim knowledge because they take themselves to have something close enough to knowledge such that there is no point in refraining from a knowledge attribution (see BonJour Reference BonJour2010: 73). To these people we might point out that an envatted brain is wildly wrong about their world, not merely basically-right-but-short-of-metaphysical-certainty.Footnote ²⁰

Additional ways to induce reflection in one's subjects could be as minimal as the use of time delays before they can answer, or making subjects justify their answers, or even just making them think they'll have to justify their answers. One could also use tasks like the Cognitive Reflection Test (Frederick Reference Frederick2005) to see whether untutored-but-basically-more-reflective subjects are more likely to deny knowledge compared with untutored-but-basically-unreflective subjects.

Having considered a few ways in which a future study might test the Skeptical Anxiety Hypothesis, I will now highlight three limitations of experimental philosophy as it is often practiced. The first two limitations concern the debate about skepticism, whereas the third limitation is much broader and more significant.

First, the data presented by NSW have no bearing on the question of skepticism's significance. This is because the argument for skepticism does not rely essentially on the intuitions they test.

Second, it is doubtful that gathering the relevant data is merely a matter of redesigning surveys. Even if Turri is right that ascriptions of knowledge are more stringent in cases involving negative inferential beliefs, his study reveals that in such cases many people remain willing to ascribe knowledge in the face of skeptical possibilities. For this reason I doubt that one could simply redesign NSW's survey to elicit robust skeptical intuitions. I think the kinds of surveys that have been stock-in-trade for experimental philosophers are an inadequate tool to test my hypothesis. Rather, we must get the subject to engage in further reflection. Merely providing them with a brief vignette and then asking whether the protagonist “really knows” or “only believes” is not sufficient.Footnote ²¹

Third, and most significantly, experimental philosophy is likely to encounter similar limitations on a variety of philosophical issues. It is doubtful that skepticism is the only philosophical view motivated by robust intuitions, rather than surface ones. Determining which views or arguments rely on surface intuitions and which do not is surely a matter to be taken up one case at a time. This is not a task I have space for here, but there is no reason to regard the debate about skepticism as unique in the sense that it is the only area of philosophy where robust intuitions, rather than surface intuitions, play an important evidential role.

For this reason, I think adopting the Socratic method would lead to a change in some of our philosophical practices. This is especially so when we consider the widely used “method of cases,” which is the use of intuitive judgments elicited by intuition pumps (i.e., thought experiments) as evidence for or against philosophical theories. The method of cases is useful for eliciting surface intuitions, but to get at the deeper, more philosophically relevant intuitions, philosophers have to supplement the method of cases with Socratic engagement. Many philosophers do this already, but many do not.Footnote ²² It is not enough to simply present one's readers with a hypothetical case that is supposed to elicit an intuition I, and then use the content of I as a premise in an argument for or against a philosophical theory. Rather, one must also get the subject to participate in further reflection.

One might worry that it is more difficult to apply the Socratic method when writing about philosophy than in conversation. Perhaps it is more difficult, but it is certainly not impossible (and nobody said philosophy was supposed to be easy). For example, Descartes engages us in precisely this kind of reflection in the First Meditation. He does not merely present us with the possibility that we are dreaming in order to elicit a certain intuition to use as a premise in his skeptical argument. Rather, he leads us to reflect on our knowledge through a chain of reasoning. The Meditator starts by considering the large number of falsehoods he has believed during his life, and he reasons that he might cast all his opinions into doubt if he can doubt the foundations and basic principles upon which they are founded. Subsequently, he claims most of what he knows comes through the senses, but he reminds us the senses can mislead us. In response, he acknowledges that insane people might be deceived, but asserts he is not one of them; however, he then remembers that as humans we dream, and that he has himself been deceived by his dreams before. The Meditator concludes that he can doubt composite things but not the simple and universal parts from which they are constructed, like shape, quantity, size, time, etc. On further reflection, however, the Meditator provides reasons to believe that even simple things can be doubted. And so forth.

What we see here is the Socratic method in action. Descartes is slowly pulling us towards the skeptic's conclusion by getting us to carefully reflect on the reasons for our beliefs. This, I submit, is a model for how philosophers should go about doing the business of philosophy.

What is required is a thoughtful and careful examination of intuitions and the specific roles they play in philosophical theorizing. We must ask whether the intuitions tested by experimental philosophers really are part of the evidence for philosophical theories. In this paper, I have tried to clarify the evidential role of both surface and robust intuitions in the argument for skepticism, but this is just one small corner of philosophy. The picture of philosophical practice that NSW mistakenly impute to the skeptic is probably untrue of philosophy more generally. If the survey model used by experimentalists is an inadequate tool in areas beyond the study of skepticism, then armchair-oriented philosophers will have a way to defend traditional analytic philosophy from some of the challenges posed by experimental philosophers.

This isn't to say that all areas of philosophy escape the experimentalist critique. Contrary to Kauppinen (Reference Kauppinen2007), with whom my view is perhaps most similar, I do not go so far as to claim that recent experimental work has contributed nothing to philosophy. Kauppinen thinks that experimental philosophy cannot resolve the central questions posed by “the analytic project” and he says this constitutes a serious problem for work in experimental philosophy. I reject this view for two reasons: first, I have argued that experimental philosophers have a wider range of methodological tools at their disposal than the surveys used by NSW; second, it is plausible that surface intuitions do play an important and perhaps essential role in some philosophical debates. For example, consider intuitions about the footbridge version of a trolley case (Thompson Reference Thompson1985). We are supposed to just see that it is impermissible to push a fat man off a footbridge to stop a train bearing down on five people, killing the fat man but saving five equally innocent others. This (surface) intuition is supposed to count as ceteris paribus evidence against the utilitarian claim that we ought to sacrifice one life to save five. So if this intuition displays wide variability, the case against utilitarianism might be weakened.

Assuming standard philosophical practice sometimes involves an appeal to surface intuitions, the simple survey methodology can (and should) be used to examine the evidentiary foundation for philosophical views that depend on such intuitions. These studies can both challenge and corroborate the claims of traditional philosophy: sometimes the results will undermine the predictions made by armchair philosophers about what ‘we’ find intuitive; sometimes they will empirically verify the claims made by practitioners of armchair philosophy.Footnote ²³

The philosophical relevance of experimental philosophy will thus depend on the extent to which its practitioners elicit the relevant intuitions. The survey methodology used by NSW (and others) will only be useful to the extent that philosophers are interested in developing theories (of knowledge, belief, responsibility, free will, etc.) that accommodate surface intuitions. To the extent that surface intuitions play a central role in philosophical practice, armchair-oriented philosophy is hostage to the data gathered by experimental surveys. (And it may very well be true that practitioners of this type of philosophy too often rest content with the assumption that their own intuitions are representative of those of a broader class of people.)

A final point: my argument does not depend on the contentious claim that robust intuitions are more reliable than surface intuitions. This further distinguishes my view from Kauppinen (Reference Kauppinen2007), who takes a stand on the controversial issue of what counts as the correct usage of the relevant philosophical concepts. No part of my argument assumes that robust skeptical intuitions are more accurate or reliable and thus should be privileged; thus, I can concede to NSW that there is no more reason to think one set of intuitions tracks the truth any better than another set (see NSW 2012: 242). In fact, it might even be true that robust intuitions are defective or imperfect or exhibit an incomplete understanding of the relevant concept.Footnote ²⁴ The skeptic's argument does not require skeptical-friendly judgments to be indicative of genuine competence. So I can reject NSW's meta-skeptical argument while holding fire on the question of whether or not reflective judgments are more likely to be correct.