Abstract

Most existing accounts of what is required for an agent to be rational--in economics, decision theory, game theory, and philosophy--are based on the idea of optimizing: an agent is said to be rational if and only if she selects the best alternative in every choice situation. ;This kind of perfect rationality requirement has recently, however, been shown to be too strong. Several impossibility results, from several disciplines, have shown perfect rationality to be strictly impossible. Since what is impossible cannot be obligatory, perfect rationality has ceased to be suitable even as a normative ideal. To take its place, we have only a disparate and tangled set of descriptions of various kinds of bounded rationality, i.e. kinds or degrees of rationality that may reasonably be expected of bounded agents. ;This situation gives rise to the main questions which this dissertation begins to answer: under what conditions can a limited, finite agent, which is incapable of perfect rationality , be said to be rational? In other words, which of the standard necessary conditions for rationality remain acceptable after the bounded abilities of agents are taken into account? How can we demarcate rational deliberation or action from arbitrary choice, once we have given up the demand for perfect rationality? ;In Chapter 2, I argue for some minimal conditions which any sensible rationality-criterion must include, basing this on the standard accounts of rationality. In Chapter 3, I argue that there are few significant differences between the several kinds of rationality , that therefore existing work on all of these kinds may be drawn upon in constructing an account of bounded rationality, and that such an account ought in turn to be applicable to each kind of rationality. In Chapter 4, I canvass the assorted arguments against the possibility of perfect rationality in order to extract a coherent delineation of just what is impossible, and therefore also of what is possible. In Chapter 5, I survey the existing work on bounded rationality, in order to extract the material for the unified framework for theories of bounded rationality constructed in Chapter 6. Finally, in Chapter 7 I assess the degree of success of the framework I have proposed