Backward Induction and Model Deterioration

The issue of how players' model of a game may evolves over time is largely unexplored. We formalize this issue for games with perfect information, and show that small-probability model deterioration may upset the complete-model backward induction solution, possibly yielding a Pareto-improving long run distribution of play. We derive necessary and sufficient conditions for the robustness of backward induction. These conditions can be interpreted with a forward-induction logic, and are shown to be closely related to the requirements for asymptotic stability of the backward induction path under standard evolutionary dynamics.