The reinforcement heuristic in normal form games

We analyze simple reinforcement-based behavioral rules in 3 × 3 games through choice data and response times. We argue that there is a large overlap between reinforcement-based heuristics (win-stay, lose-shift) and the more “rational” behavioral rule of myopic best reply. However, evidence from response times shows that choices in agreement with the common prescription of those rules are comparatively fast, and choices of the form “lose-shift” occur more frequently for larger differences with bygone payoffs. Both observations speak in favor of reinforcement processes as a cognitive shortcut for apparent myopic best reply, and advise caution when interpreting behavioral results in favor of optimizing behavior.