Abstract

This dissertation develops the theory of imperfect rationality in the practical and theoretical domains. I characterize imperfect rationality in terms of the quality of reasoning on which an agent's actions and beliefs are based, which I call their rational worth. Perfectly rational actions and beliefs are based on the best reasoning available to the agent: they are based on all the agent's evidence, on an appropriate weighting of all the relevant values, and on the right inductive and deductive principles. Imperfectly rational actions and beliefs are based on good reasoning that falls short of the best reasoning: they are based on a portion of the agent's evidence, on a subset of the relevant values, and on approximate or heuristic reasoning. The idea of rational worth, I argue, helps both to describe and to improve upon the tools that imperfectly rational agents use to think about what to do and what to believe. These tools include a plurality of deliberative questions that imperfectly rational agents use to guide themselves toward better ways of falling short of the ideal of perfect rationality. They also include a practical normative verdict available to agents who can assess a few possible rationales for acting one way or another but who cannot say what an ideal evaluation of all their evidence would support doing. Finally, they include substantive heuristics that result in beliefs and actions that are more, but still not perfectly, rational; for example, I argue that imperfectly rational agents can legitimately engage in a form of hindsight bias as a reasonable second-best strategy. I aim primarily at understanding how imperfectly rational agents can and should cope with their distinctive predicament, but in each case, I also draw lessons for our understanding of the ideal of perfect rationality.