Learning in experimental 2×2 games

In this paper we introduce four new learning models: impulse balance learning, impulse matching learning, action-sampling learning, and payoff-sampling learning. With this models and together with the models of self- tuning EWA learning and reinforcement learning, we conduct simulations over 12 different 2×2 games and compare the results with experimental data obtained by Selten & Chmura (2008). Our results are two-fold: While the simulations, especially those with action-sampling learning and impulse matching learning successfully replicate the experimental data on the aggregate, they fail in describing the individual behavior. A simple inertia rule beats the learning models in describing individuals behavior.