Project Abstract Despite their widespread application, there are well known examples in which economic theory, decision theory, and game theory fail to describe human behavior realistically. Since these theories are used in public policy analysis, the prescriptions of such analyses can be questioned on this ground. This research program aims to further understanding of the conditions under which the standard theories fail and to develop new theories for use in such conditions. The research is divided into four projects. The first of these is a collection of laboratory experiments concerning two person zero sum games. This is the class of games for which the theory of games was initially invented, and for which it might be expected to match reality best. Past laboratory experiments on such games have indicated that there are senses in which the theory describes reality well and senses in which it does not. The experiments in this project are designed to further clarify of this issue with the eventual purpose of contributing toward a modified theory that fits reality better. The second project is concerned with the further development of a novel theory in which economic actors are assumed to be involved in such a complicated world that they have no choice but to engage in either imitative or experimental modes of behavior. The analysis of an earlier prototype mathematical model produced long run patterns that were delicately balanced between stability and instability. Since it is important to know when an economic system is stable, in this project variations on the theoretical prototype are studied in order to see when the instabilities are robust and whether such model instabilities are related to instabilities in real economies. The work involves both mathematical analysis and computer simulations. The third project is an attempt to merge the classical theory of games with the more recently developed evolutionary approach to game theory. The two approaches appear to describe reality well in different circumstances; the hope is that a hybrid theory might do better than the two theories separately. Project four stays within the confines of classical game theory. It is an attempt to generalize and see what is behind a peculiar and poorly understood pattern of equilibrium behavior that has turned up in a diverse collection of applications: from models of auctions to models of labor market screening. In these models, individuals randomize their actions in order to keep them secret from opponents, but they do so in ways that are internally highly structured.
