Probabilistic representation of complexity. (Q1400985)

A remarkable aspect of human behavior is individuals' ability to cope with complexity. The role of complexity in everyday life is almost self-evident: people engage in complex social and economic interactions where it is pointless to examine every possible contingency. While its importance and pervasiveness is beyond dispute, complexity is an elusive concept in a formally model. Difficulties arise from modeling complexity in a way that precludes systematic errors and arbitrage opportunities that undermine the equilibrium analysis. Here an agent's behavior is explained as a result of a coherent probabilistic model representing his attempt at identifying patterns or regularities, while at the same time recognizing that his model cannot explain everything. This paper proposes an approach that produces interesting cognitive limitations due to complexity, yet behavior is rational in the sense of being consistent with learning from data and displays no exploitable systematic errors.