Abstract

According to the ”Lockean thesis”, rational full belief corresponds to partial belief (or credence) above a certain threshold. Within the branch of formal epistemology known as epistemic utility theory, a version of the Lockean thesis can be derived from the popular assumption that rational agents strive to maximize the ”expected accuracy” of their beliefs (Dorst 2019).

In this talk, I discuss the prospects of generalizing the utility-theoretic Lockean thesis to settings where agents may be ”imprecise”, in the sense of having credences represented not by single (probability) functions, but by sets of such functions (e.g., Levi, 1974; 1980). In this context, there is no guarantee of a unique ”expected accuracy” for agents to maximize, meaning that a new decision rule is required to ground the relation between rational full belief and imprecise credence.

Various decision rules for imprecise agents have been proposed (neat overview: Troffaes, 2007), but these have not yet been discussed in relation to epistemic utility theory. This talk is an early-stage attempt to bridge this gap. I consider a selection of decision rules for use with imprecise credences (most extant, one novel-ish), and discuss their applicability to a basic class of "doxastic decision problems”: three-way ”choices” between believing, disbelieving, or suspending on a given proposition.