UC Irvine

This dissertation is a systematic exposition of the pragmatic case for Bayesian conditionalization. In contrast to traditional epistemologies which begin with notions of belief, justification, or knowledge, I begin with the notion of an update, and then ask the question: How would an agent want to update in various situations? To answer the question, I retrieve and generalize a value of information theorem due to Peter M. Brown. In many situations, agents will want to update by conditionalization or by some other method that approximates it. One of the main contributions is an exposition of what it means to approximate conditionalization, with special attention to the case of Jeffrey conditionalization and, by extension, some non-partitional updates. The second half of the dissertation catalogs situations in which an agent wouldn't want to update by conditionalization, with new implications for accuracy-centered epistemology and social epistemology.