| As an epistemic agent, my ultimate goal is to match my doxastic attitudes to the world. Matching isn't an all-or-nothing affair. Instead I face a gradational criterion of success: the closer I come to fitting my beliefs to the world, the better. I then have reason to follow a given epistemic constraint on my credences only insofar as I have reason to think it will help me in my quest for accuracy. Truth, in other words, is the highest epistemic good, and closeness to truth is epistemic utility. In matters both pragmatic and epistemic, an agent ought to maximize her utility, and I exploit the standard apparatus of decision theory throughout the dissertation. However, while rational agents may have vastly different practical utility functions, they mostly agree about which doxastic states are preferable to others. Even if Hume is right that it is not pragmatically irrational to prefer the destruction of the world to the scratching of his little finger, it's epistemically irrational to prefer a credence of 1/2 to a credence of 3/4 in a true proposition. Because the space of reasonable epistemic utility functions is so limited, cognitive decision theory is a powerful formal tool for answering normative questions in epistemology. This dissertation develops the decision-theoretic approach. Chapter 1 explores how the quest for accuracy ought to determine derivative norms on doxastic states. I argue that if a kind of epistemic attitude can be taken as primitive---i.e., not derived from a more fundamental kind of attitude---then it should come packaged with a sufficiently robust notion of accuracy. Unlike full-belief and credence, comparative confidence has no prospects for its own measure of accuracy. Chapter 2 examines which epistemic utility functions are rationally permissible. I attack the most popular measures of inaccuracy---quadratic scoring rules---and provide considerations in favor of logarithmic rules. Chs. 3 and 4 apply cognitive decision theory to the problem of peer disagreement. |
