Existing economic equilibrium models assume that agents, participating in interactive systems, have precise knowledge of the system, the environment in which it operates, and the behavior of the other agents. When such knowledge is not available, or when differential information is possessed by different agents, the models assume that uncertain parameters are drawn from prior probability distributions which are known to the agents. Since the parameters and distributions assumed known are the correct ones, these models may be considered objective. While some of the data available in interactive situations is objective, the vast majority is not. For example, when one models competition with uncertainty about production costs, any probability distribution over these parameters must be highly subjective. Probability distributions over future prices and competitors' actions, which compound uncertainties about basic primitives with uncertainties about other agents' responses to such uncertainties, become even more subjective. A second criticism of economic equilibria is the lack of dynamical systems describing the learning processes that lead agents to equilibrium. Without such a process it is difficult to justify equilibrium analysis as a predictive model of agents' actual behavior. This project aims to address the above two issues. Some well- established dynamic equilibrium concepts will be expanded to allow economic agents to hold subjective beliefs. The research will concentrate on (1) developing further theoretical methods in the theory of Bayesian and Bounded Bayesian forecasting, and (2) applying these methods to some well-established equilibrium concepts in economic theory. These include equilibrium analysis in strategic environments, rational expectations equilibrium, and equilibrium analysis of more general social dynamics.
