A Contextual Accuracy Dominance Argument for Probabilism

A central motivation for Probabilism—the principle of rationality that requires one to have credences that satisfy the axioms of probability—is the accuracy dominance argument: one should not have accuracy dominated credences, and one avoids accuracy dominance just in case one satisfies Probabilism. Until recently, the accuracy dominance argument for Probabilism has been restricted to finite credal states. One reason for this is that there are several impossibility results that apply when defining the accuracy of infinite credal states. In this paper, I offer a fully general accuracy dominance argument for Probabilism that allows for the possibility that not all sets of credences can be measured for accuracy. The normative core of the argument is the principle that one should not have credences that are accuracy dominated in some "scorable" epistemic context by alternative credences that do not have this defect. Beyond this novel rationality principle, the argument offers two important theoretical tools to accuracy-first epistemology: the idea of an epistemic context being scorable and a general contextualizing strategy for extending the arguments of accuracy-first epistemology to the infinite setting.