Do players reason by forward induction in dynamic perfect information games?

Abstract: We conducted an experiment where participants played a perfect-information game against a computer, which was programmed to deviate often from its backward induction strategy right at the beginning of the game. Participants knew that in each game, the computer was nevertheless optimizing against some belief about the participant's future strategy. It turned out that in the aggregate, participants were likely to respond in a way which is optimal with respect to their best-rationalization extensive form rationalizability conjecture - namely the conjecture that the computer is after a larger prize than the one it has foregone, even when this necessarily meant that the computer has attributed future irrationality to the participant when the computer made the first move in the game. Thus, it appeared that participants applied forward induction. However, there exist alternative explanations for the choices of most participants; for example, choices could be based on the extent of risk aversion that participants attributed to the computer in the remainder of the game, rather than to the sunk outside option that the computer has already foregone at the beginning of the game. For this reason, the results of the experiment do not yet provide conclusive evidence for Forward Induction reasoning on the part of the participants.