Christopher Nicholson

Chapter One of this dissertation examines the scope of the epistemic imperative to pursue accurate belief, beginning by arguing that the accuracy that comes from believing self-fulfilling prophecies has no epistemic value. It then extends that argument’s reasoning to present a theory of the distinction between belief and decision: pure beliefs concern propositions whose truth values are entirely independent of an agent’s actions, pure decisions concern propositions whose truth values are entirely dependent on an agent’s actions, and there many cases in between, rendering the distinction between belief and decision a spectrum, not a bright line. The chapter examines a variety of cases to illustrate its point, both original ones and well-known ones drawn from the literature on epistemic consequentialism. Chapter Two presents a theory of the act of suspending judgment about a proposition, exploring the question of how the concept fits within a notion of degrees of belief. It considers the merits of reducing suspension of judgment to the absence of belief, to middling precise credence, and to imprecise credence, concluding that each reduction fails to capture all paradigmatic cases such as agnosticism about God, jurors’ suspensions of judgment about defendants’ guilt, and Descartes’ suspension of judgment at the beginning of his Meditations. The chapter concludes by presenting an affirmative theory: one who suspends judgment about a proposition chooses not to use their belief about it in all the ways people normally use beliefs. By doing this, they allow the rest of their beliefs to evolve in the way they would have if the agent held no belief about the proposition they have suspended judgment on. Chapter Three presents a theory of how agents learn new concepts. It begins by describing the problem: the Bayesian theory of learning by conditioning upon evidence explains how people can improve their existing beliefs but is silent on the question of how they can acquire new beliefs about new propositions. The chapter outlines a theory of conceptual learning that uses the same basic ingredients as learning by conditioning and can work hand-in-hand with it. The approach revolves around the use of a catch-all hypothesis that all one’s other hypotheses are wrong. Observing confounding evidence one’s regular hypotheses cannot account for raises confidence in one’s catch-all hypothesis near 1, driving them to modify their original partitions of hypotheses and observations, and their credences in propositions based on those partitions, in the most minimal way sufficient to lower credence in their catch-all hypothesis toward 0. The chapter presents and refutes influential arguments against the use of a catch-all hypothesis. It closes by describing the mechanisms by which new hypotheses and observations change an agent’s confidence.