Cases used to motivate contextualism or the denial of what Jason Stanley calls “intellectualism” are often referred to as “stakes-shifting cases” or “standards-shifting cases.” Typically, such cases tend to go schematically, like this: LOW, for whom not much hinges on whether p is true and for whom various counterpossibilities to p are not salient, says of some S who is fairly well situated epistemically with respect to p, “S knows that p.”. HIGH, for whom much is at stake or for whom various counterpossibilities have been made salient, says of either the same subject, or some subject equally well situated epistemically with respect to p, “S doesn’t know that p.”. According to those who are trying to draw epistemological lessons from these cases, LOW’s attribution and HIGH’s denial are both true.
I don’t want to get into the lessons people draw from this. My concern is how we should view the cases themselves – as data or as something like the conclusions of arguments.
The CASES-AS-DATA approach:
The usual use of the cases is as data. Either the data involves the intuitive truth of the attributions and denials presented in the cases (e.g. Stanley: “In Low Stakes, our reaction is that Hannah is right; her utterance of ‘I know the bank will be open’ is true.”) or the data involves the intuitive accuracy of the cases with respect to how we talk; intuitively ordinary and competent language users will normally talk in the ways suggested in the cases (e.g. DeRose: “As we can sense, perhaps with even more certainty than that with which we judge the truth-values of the claims, speakers do in fact use ‘know(s)’ in the way described, and appropriately so.”).
Much of the critical literature attempts to explain why our intuitive reactions cannot teach us the lessons they’re supposed to teach us: 1) we might wrongly think that the attributors speak truly in the cases, 2) even if we’re right in thinking that they speak truly, we can’t trust that fact to be consistent with stipulated features of the cases – e.g. that the subject maintains belief or the same epistemic standing across the cases, and 3) even if normal attributors might talk that way, talking that way does not indicate that what they say is true.
The CASES-AS-CONCLUSIONS approach:
The Cases-as-Data approach is not the only way to use the cases, however. They can also be conclusions of arguments in the following sense: there may be true principles that entail that there are cases of the required sort – cases in which LOW truly says that S “knows” and HIGH truly says that S “doesn’t know” but in which S’s epistemic standing and belief remain constant. Here it is irrelevant whether there is reason to think that our intuitive reactions to such cases might be misguided. The mere fact that there must be such cases gets us to the desired conclusions.
What might a principled argument that there are such cases look like? Here’s a four-step strategy-type.
FIRST, start with an assumption of something like fallibilism – the claim that S can know p even if S’s epistemic position is lacking in some crucial way (perhaps there is a chance that p is false or p is less than maximally justified for S or S’s evidence doesn’t entail that p is true). It doesn’t matter for the general strategic outline exactly what fallibilism is. The only thing this strategy needs out of fallibilism – whatever your favored conception is – is that,
SECOND, if S fallibly knows that p then there is some other subject, T, with the same epistemic standing with respect to p who is in some important sense not rational to act on p or not justified in acting on p or not justified in taking p for granted in reasoning or not acting properly if basing her action on p, even though S is so rational (or justified or proper), because the stakes for T are higher. Why is this a reasonable thing to say? Because it is reasonable to say that, on some very important senses of “rational,” “justified,” and “proper,” what a subject is rational or justified or proper to do is determined by two factors: 1) the subject’s epistemic standing with respect to some relevant propositions and 2) what the various costs and benefits are of acting in certain ways. And that seems to indicate that, if one’s epistemic standing is lacking in some relevant ways, what one is rational or proper or justified in doing can be changed by tweaking the costs and benefits of acting while holding one’s epistemic standing constant. Perhaps this is false: perhaps it is literally impossible for a LOW subject to fallibly know that p while a HIGH subject with the same epistemic standing with respect to p isn’t, say, rational to act as if p. This might have to be fought over. But fighting over this possibility is not the same as fighting over whether intuitions about the cases are trustworthy or can teach us lessons.
THIRD, argue for a principled connection between knowledge (or knowledge-attributions) and actionability. These can be anti-intellectualist-friendly (e.g. S knows that only if S is rational to act as if p) or contextualist-friendly (e.g. S truly says “S knows that p” only if S is rational to act as if p). But, in conjunction with the SECOND step and fallibilism, the result will be that “S knows that p” is true in one case and “S doesn’t know that p” is true in the other, even though S’s epistemic standing remains constant. The result is not motivated by intuitions about whether various attributions or denials are true or whether normal speakers would talk this way, or whether it is the case that if normal speakers would talk this way, then that counts as evidence that they speak truly.
FOURTH, if one is so inclined, argue for further third-person knowledge-actionability links (e.g. something like S truly says “T knows that p” only if what S reasonably takes T’s epistemic standing to be is sufficient for S to act as if p). If the principles are chosen carefully enough, we can generate the third-person cases that are so helpful for contextualists.
Some ways of executing this strategy might be less promising than others, and some steps might be more fraught with difficulty. There can always be worries about the knowledge-actionability links. But the important point is that the worries about those links need to be addressed at the arguments for those links, NOT at the intuitions about the cases, because the cases are not used in the arguments for the links. The existence of the cases, rather, follows from fallibilism, reasonable connections between fallibilism and what possible epistemic twins are justified or proper or rational or reasonable in doing, and reasonable links between knowledge (or “knowledge”) and actionability.