Enrique Munoz de Cote
Abstract: The cornerstone game theoretic assumption, i.e. (VNM-)rationality, implies that all players that satisfy four "rationality" axioms have a utility function that describe their goals. This suggests that when rational agents (humans, animals or machines) repeatedly face the same strategic situation, one can expect them to exhibit equilibrium behavior. However, many experimental studies say otherwise. Could this disagreement be caused by the type of rationality suggested by GT? Could the problem be associated with physical constraints (be them human, animal or machine) or is it a problem of (mis)information?
In this talk I will give an overview of recent work on cognitive models used to model opponents. I will sketch pros and cons of the most representative models, map these to the LSG tournament and close with still unanswered questions on opponent modelling.