Modeling bounded rationality