A Problem with Bayesian Conditionalization

Abstract

Bayesianism is often divided into a static and dynamic part. The static part is a theory of rational belief at an instant, and the dynamic part is a theory of rational belief change. In this paper I will begin by looking closely at the rule of conditionalization, which is central to the dynamic side of Bayesianism. I will argue that the dynamic side of Bayesianism needs to make distinctions central to traditional non-Bayesian epistemology; Bayesians may consider this unfortunate, since Bayesianism is supposed to be a way to avoid many of the problems that arise in classical epistemology. I will then give a counterexample that shows we can be rational even if we do not conditionalize in situations in which conditionalization is applicable. This counterexample also raises problems for the view that conditionalization is a way to manage our beliefs in pursuit of some ideal. Contrary to Bayesian doctrine, I will argue that we generally cannot use conditionalization to manage our beliefs. The reason for this is that conditionalization is an externalist requirement that aims at the goal of being in a good epistemic situation, but rationality contains only internal requirements; thus conditionalization is not required in order to be rational. I will then look at whether we can develop an internalist version of