In this post I tried to spell out some of my concerns about the the basic experimental design of the WSN experiments. In this follow up post, I want to say a few things about the interpretation of those findings. Obviously, the two issues can't be wholly independent, and much of what I said previously falls under the heading of "interpretative limitations due to methodological shortcomings." But I do think that there are a few more important points to be made.
Let me begin with WSN's presumption that if there is cross-cultural variation in intuitions, this would be a serious problem for intuition-driven rationalism. There are at least three relevant possible readings of the claim that there exists cross-cultural variation in intuition. (There are probably more, depending on which quantifiers are included, but these will suffice to make my point.) In increasing order of strength these are as follows:
- For any culture A there is some culture B such that for [all/some] cases of a given type, while most individuals from both A and B have the same intuitions about some of those cases, the relative proportions are higher in A than B.
- For any culture A there is some culture B such that with respect to at least some cases of a given type, individuals from A differ from individuals from B about the preferred response to those cases.
- For any culture A there is some culture B such that with respect to all cases of a given type, individuals from A differ from individuals from B about the preferred response to those cases.
WSN do not clearly distinguish between these three, very different, claims nor do they give much attention to the differences in philosophical significance between them. So which of these theses, if any, do their studies support? Well, the True Temp cases establish at most the truth of (1). In those cases, there was no difference in the majority opinion (both groups preferred internalism), but there was a statistically significant difference in the relative proportion of externalists (EAs tended more strongly toward internalism). The Gettier case establishes at most the truth of (2), as do the remaining cases. Their studies provide virtually no evidence for (3).
But I think it is clear that (1) and (2) differ greatly from (3) in their philosophical import. Consider (1). It would still follow that there is significant, indeed universal, cross-cultural agreement in intuitions. Why cultures would differ with respect to the relative proportion of people who have the majority intuition is an interesting question, but one that hardly seems to impugn the evidential status of intuitions themselves. The massive cross-cultural agreement uncovered by this study is far more striking than the relatively minor differences. In any event, it is unclear why cross-cultural differences in relative proportions (that all we really have here, REALLY!) is more troubling than the simple observation that not everybody within a given population has the same intuitions. What more, if anything is the cross-cultural part supposed to add? (Of course, the cross-cultural component has a rhetorical advantage, since philosophers are in general shy about appearing to be ethno-centric.) Plausibly this would show that some cases are more difficult for members of some cultures to "get their minds around," and this would hardly be surprising given cultural differences of the sort found by Nisbett et al. Indeed, I think the IDR theorist would predict cross-cultural variation of this sort (see the discussion below).
Reading (2). Things are only moderately dicier here for the rationalist. In the majority of cases, I don't see why this would be philosophically surprising or problematic. After all, merely showing that members of culture A, for instance, fail to have the Gettier intuitions about case X doesn't establish that members of A are not Gettierizable. It merely shows that they don't have the Gettier intuitions about case X; it doesn't even show that members of A can't get the Gettier intuition about X, only that they don't get it up front, as it were. Moreover, if they are Gettierizable, then presumably we could eventually get them to see the case as similar to cases where they withold knowledge. After all, if your knowledge attributions are sensitive to factors concerning epistemic luck (you have the Gettier intuitions) but you aren't getting the Gettier intuitions about, say, the Brown/Barcelona case, then there is something about the case you aren't keying into which makes it a case involving epistemic luck. The alternative is to say that the two cases display different kinds of luck and that knowledge is only sensitive to one kind and that just doesn't seem like a charitable take on things. But even if that were so, then I think we would have a fully adequate account of the differences in a way that would be worrisome for the IDR theorist: either they are locally confused about knowledge (or I am) or they have a different concept (see below).
Furthermore, reading (2) doesn't show anything about the reliablility of the Gettier intuitions unless we have reason to think that case X is indeed a case of knowledge or we have reason to think that members of A are having counter-Gettier intuitions. But nothing in the WSN data give us any reason to believe either of these things. And even if it did, it is still a long step from here to general skepticism about intuitions. After all, few if any contemporary IDR advocates are infallibilists. Incidentally, it is worth bearing in mind here that, while WSN focus on cases which are similar to cases tabled for philosophically substantive purposes, the IDR theorist will insist that when assessing intuitions we also need to consider all of the boring paradigm cases as well. So not only should we consider, say, the Gettier-type cases, but also run-of-the-mill cases that don't have any philosophical interest in and of themselves. If there weren't on balance, massive agreement on these cases, it is hard to see why we would even take the two cultures to possess the same concept.
So it is reading (3) that would be most worrisome. Or would it? I suppose it depends on the details, but at least one natural explanation for "cultural variation in intuitions" in this sense is culturual variation in concepts or conceptual schemes. This would be a genuinely interesting philosophical result, but one that bears most markedly on issues other than the evidentirary status of intuitions. For instance, if we were to find a culture C that was unGettierizable, this might be a reason for thinking that they lack the concept of knowledge altogether. This might, in turn, be a reason for thinking that concepts like knowledge aren't as central to a theory of the world as one might have thought. And that is an interesting result, but not an interesting result about intuitions. (For instance, this might be the most plausible account of the the Machery et al results [couldn't find a link] concerning cultural variation in rigidity intuitions about "water" and its translations.)
In sum: while it sounds bad for IDR that there might be cultural variation in intuitions, this is so (at worst) only when you read the claim as in (3) and the WSN results just simply don't support that reading.
There is a second concern about how to understand the WSN results regarding the True Temp cases. According to the reported results, they gave three different cases of this type. In the first one, which was intended to reflect a more individualistic Western perspective, they found a statistically significant difference between responses given by Westerners (Ws) and East Asians (EAs). They then wondered if they could "undermine" this result by giving cases which were more reflective of an East Asian perspective. As predicted this is what they found. In the two cases of this type, there was no signficant difference between Westerners and East Asians. But we should simply ignore their "prediction." The fact is that in two out of the three cases WSN ran, they found no significant difference in even the proportions of the responses given. Even if they have a hunch as it why the second two cases found no difference, this is little more than an unsubstantiated guess. As a result, their data just simply don't support the claim of cross-cultural variation in the sense of reading (1) above.
OK. Last point. Despite a relatively long discussion of IDR, WSN don't really give us a clear picture of their target. In fact, I think that their attempt to get results that apply very widely has simply caused them to miss almost everyone. So let me suggest a target that most advocates of IDR would be comfortable accepting and would feel uncomfortable if it were wrong.
[P] If two subjects A and B understand a case C in the same way, then (typically) A and B will have the same intuitions concerning C.
But I just don't see how anything in the study supports the denial of [P]. By the very reasoning they use to motivate the study, we have good reason to believe that individuals from different cultures may not understand a written vignette in the same way. So even if we grant that the subjects are having intuitions and even if we grant that these intuitions vary by culture, we don't really have any reason to think that these intuitions are about the same case (or the same case understood in the same way). In order to show that [P] is false, they need to show that the antecedent obtains and this is precisely what we have reason to doubt. Rather, at most the WSN experiments show that the following principle is false:
[P*] If two subjects A and B read a given case C, they will typically have the same intuitions about C.
But whoever thought that [P*] was true? In the discussion of Part I, I have already expressed concern about identifying reading-and-responding with intuiting. But waving that concern, WSN's interpretation of their results effectively relies on the following assumption: if A and B read C this is sufficient for A an B to understand C in the same way (often enough). But I just don't see any good reason for thinking that this is true in the experimental context under consideration. (And no, this doesn't generate some sort of skepticism about whether or not I know whether or not people are having intuitions about the same case understood in the same way. The point isn't that it is really, really hard to know such things, but only that this is a source of disagreement and one that is hard to control in the experimental setting of WSN.)