The last few years have seen increasing mutual awareness and interaction between computer science and interactive decision theory (e.g., game theory, economics). This has been spurred by parallel developments in the two disciplines: in computer science, the advent of distributed systems; in interactive decision theory, the analysis of increasingly complex social systems, leading to the realization that traditional rationality assumptions make extremely large demands on the information processing capabilities of the agents. Computer scientists started using game-theoretic and similar interactive models for the study of distributed computing systems and economists turned to computer science (finite automata, Turing machines, and actual computer programs) to model bounded rationality. This project concentrates on three central problems which arise in this connection: (1) Optimization with limited memory-the study of the effect of limited capacity on the optimization of rational agents. The explicit modelling of the cost of information processing. (2) Common knowledge-to introduce the idea of "formal common knowledge" which would, practically speaking, achieve all the desirable coordination consequences that would be achieved under common knowledge. (3) Social interaction with bounded rationality-to formulate the presence of irrationality in social systems and define their equilibria. That is, the project initiates a serious study of the behavior of interactive systems when some or all of the agents are bounded rational.//
