In Defense of Sensitivity

The sensitivity condition on knowledge says that one knows that P only if one would not believe that P if P were false. Difficulties for this condition are now well documented. Keith DeRose has recently suggested a revised sensitivity condition that is designed to avoid some of these difficulties. We argue, however, that there are decisive objections to DeRose’s revised condition. Yet rather than simply abandoning his proposed condition, we uncover a rationale for its adoption, a rationale which suggests a further revision that avoids our objections as well as others. The payoff is considerable: along the way to our revision, we learn lessons about the epistemic significance of certain explanatory relations, about how we ought to envisage epistemic closure principles, and about the epistemic significance of methods of belief formation. Quite a while ago now, Robert Nozick suggested that one knows that P only if one’s belief that P is sensitive to the truth, that is, only if one would not believe that P if P were false. In spite of the fact that sensitivity seems an appropriate condition for knowledge, its troubles are well documented. Keith DeRose has recently proposed a condition on knowledge that assigns sensitivity a leading role while it recruits supporting players that will help sensitivity avoid the difficulties it faces when it stands alone. We argue, however, that there are decisive objections to DeRose’s proposal. Yet rather than simply abandoning his proposed condition, we uncover a rationale for its adoption, and then revise it so that it avoids objections like those we offer. The payoff is considerable: along the way to our revision, we learn lessons about the epistemic significance of certain explanatory relations, about how we ought to envisage epistemic closure principles, and about the epistemic significance of methods of belief formation. 1. DEROSE’S PROPOSAL DeRose introduces his conditions on knowledge as a means to solving a problem. The problem arises for theories according to which one knows that P only if one sensitively believes that P. Again, this seems a reasonable requirement for knowledge. For one thing, as Nozick points out, the sensitivity condition, unlike the justified-true-belief account, allows us to get the right result in cases like those described by Gettier. Suppose, for example, that I am justified in believing that (J) Jones, who works in my office, owns a Ford. From this, I legitimately infer that (S) someone in my office owns a Ford. This belief is true, but it’s true not because Jones owns a Ford, but because Brown, who works in my office but is a stranger to me, owns a Ford. I have a justified true belief that S, but I do not know that S. The sensitivity condition, however, yields the right result in this case: I don’t know that S because I would believe that S even if it were true that no one in my office, including Brown, owns a Ford. But certain cases suggest a problem for these theories. The problematic cases are those in which we judge that I know that P even when I do not sensitively believe that P. DeRose provides two such cases. First, take my belief that (F) I don’t falsely believe that I have hands. This belief is insensitive, for I would believe that F even if F were false. Nevertheless, we judge that I know that F. DeRose’s second case centers around the belief that (D) I’m not an intelligent dog who’s always incorrectly thinking that I have hands. Again, this belief is insensitive, but we judge that I know that D. These cases suggest a problem: while sensitivity seems to be a reasonable condition on knowledge, some insensitive beliefs seem to count as knowledge.