Behavioral Implications of Causal Misperceptions

This review presents an approach to modeling decision making under misspecified subjective models. The approach is based on the idea that decision makers impose subjective causal interpretations on observed correlations, and it borrows basic concepts and tools from the statistics and artificial intelligence literatures on Bayesian networks. While these background literatures used Bayesian networks as a platform for normative and computational analysis of probabilistic and causal inference, in the framework proposed here graphical models represent causal misperceptions and help analyze their behavioral implications. I show how this approach sheds light on earlier equilibrium models with nonrational expectations and demonstrate its scope of economic applications.