题目：Multiple Selves

主讲人：Kim-Sau Chung，Associate Professor, University of Minnesota

Abstract

We show that (a version of) common strong belief of rationality and three epistemic axioms characterizes a new solution concept called agent-form rationalizability (AFR). We argue that AFR is an appropriate solution concept to capture forward induction reasoning in games where players may or may not be time-inconsistent. When players are all time-consistent, AFR collapses to extensive-form rationalizability (EFR), a solution concept that Battigalli and Siniscalchi (2002) argue is an appropriate solution concept to capture forward induction. This shows that forward induction reasoning, at least as captured by EFR, does not really rely on the strong assumption of full control. Replacing the assumption of full control with our three epistemic axioms suffices to obtain the same predictions.