Abstract

The notion of rationality is crucial to Computer Science and Artificial Intelligence, Economics, Law, Philosophy, Psychology, Anthropology, etc. Most if not all of these disciplines presuppose the agent's capacity to infer in a logical manner. Theories about rationality tend toward two extremes: either they presuppose an unattainable logical capacity, or they tend to minimize the role of logic, in light of vast data on fallacious inferential performance. We analyze some presuppositions in the classical view of logic, and suggest empirical and theoretical evidence for the place of inferential heuristics in a theory of rationality. We propose (1) to outline a new theory of rationality that includes the key notion of logical capacity as a necessary but realistic factor, (2) to expand the notion of inference to include non-deductive inference, specifically non-monotonic, and (3) to emphasize the logical role of inferential heuristics and constraints such as cognitive economy.