My grad seminar this semester is on pragmatic encroachment–syllabus here, for those interested in what we are reading: Pragmatic Encroachment Seminar. Everything we are doing is stuff that gives me a reason to get up in the morning to do more epistemology, and I love getting to teach that kind of material!
This week we did chapters from Jason Stanley’s book Knowledge and Practical Interests, which is wonderful stuff! Got clear on the form of the view, from the programmatic proposal on page 89, where the idea is that the kind of invariantism Jason wants is one that claims four conditions necessary for knowledge and jointly sufficient (though I whine about the first condition, that knowledge of p requires that ~p is not a serious epistemic possibility, noting that any account which builds an account of epistemic possibility on top of the usual dual rule will find this condition introducing a hopeless circularity, as well as whining about the fact that there’s nothing here to address the Gettier problem).
And then we come to the arguments about costs and benefits of such a view, compared with contextualism, and indirectly with other views. The arguments concern a standard bank case, with Hannah the agent considering whether to go to the bank. Jason rightly notes the costs of his version of invariantism (noting that when the stakes shift higher between Thursday and Friday, his view might have to say that Hannah knew on Thursday that the bank is open on Saturdays, but didn’t know this on Friday). Jason tries to disarm this admittedly unintuitive result by showing that other views have other counterintuitive results. Especially, he wants to show that some standard versions of classical invariantism, which deny pragmatic encroachment, have unintuitive consequences.
The one I’ll remark on here is the one about what Jason reports reliabilists must say about fake barn country.
Here’s what Jason says about reliabilism:
Reliabilism captures this intuition [that in fake barn country one doesn’t know that the the object one is observing is a barn], because, in Ginet’s fake barn country, perception is not a reliable method. Nevertheless, we cannot say, of a person in fake barn country who is looking at a genuine barn:
“Poor Bill. He doesn’t know that is a barn. But if there were fewer fake barns around then he would know that is a barn.”
Several things to say here, but I’ll do some border skirmishing before getting to the main battle. First, the idea of perception being unreliable in fake barn country is false and insensitive to the literature. What’s unreliable is, at most, perception-of-barns-in-fake-barn-country. And the relevance of this depends on which version of reliabilism one is talking about. Certainly no plausible version would hold that all perceptual beliefs are undermined with respect to knowledge by the presence of fake barns, and some reliabilists think it doesn’t undermine knowledge about barns either (see, e.g., Lycan and Sosa–except, of course, that only one of these is a reliabilist, on which point see the next paragraph…).
Second, if there is a problem here, it is a problem for any approach to the entire Gettier problem, not one for reliabilism alone. While it is true that reliabilism can be a theory about what it takes to turn true belief into knowledge, most reliabilists are willing to admit that their view requires a Gettier condition in addition to a reliability condition. So, this isn’t really about reliabilism per se. Instead it’s about what an adequate response to the Gettier problem will look like, and the quote about poor Bill applies to any (responsible(!)) view equally: no false lemma views, defeasibility views, modal views involving safety and sensitivity, etc.
But more important, why think the quote is troubling? Well, it is troubling, since it is obviously false. But no theory of knowledge in the post-Gettier situation needs to endorse it. In the Ginet case, the Wisconsinites have replaced nearly all barns with fake barns, and so it is false that if there were fewer fake barns, Bill would know. No, he wouldn’t: there doesn’t have to be even a preponderance of fakes to undermine the knowledge claim. In fact, it won’t take a very large percentage at all to cause problems here, even though the case was originally constructed in an exaggerated way because all that was needed at the time is to show that no inference through some false claim was needed for a Gettier-like situation to arise.
What is true is that it there were zero or nearly no fake barns, he would know. What the threshold is here–how many fake barns can leave knowledge intact–is vague and need not concern us. The point is that a bit fewer fake barns doesn’t help: you have to eliminate nearly all of them.
For this to be a cost to any theory, it would have to be intuitively false. But it isn’t. As I read it, it is an obvious truth. So I don’t see what the intuitive cost is in fake barn cases for standard invariantist views.
That is not to say, of course, that there aren’t other costs, such as having to explain away any purported data about what people ordinarily say about bank, train, and plane cases–the cases used to motivate both contextualism and the sorts of invariantisms that embrace pragmatic encroachment. If intuitive costs are being measured, that is probably a better place to focus on than with Gettier cases.