Epistemic Risk and Epistemic Utility for Credences

Abstract

AbstractIn this chapter, I note Sophie Horowitz’s (2017) observation that a credal version of Kelly’s Jamesian argument from the previous chapter will not deliver permissivism in the credal case. This is because, in that case, it is generally agreed that all legitimate measures of epistemic utility must have a particular property—in the jargon, they must be strictly proper. But, for any strictly proper epistemic utility function and any evidential probability, the evidential probability expects itself to be epistemically better than any other credence. So it renders itself the unique rational response to the evidence, which gives impermissivism. I conclude that, if we are to secure permissivism for rational credences using epistemic utility theory, we cannot adopt Kelly’s approach.