Journal of Pattern Recognition Research

This paper presents and tests a new learning model of boundedly rational players interacting with nature. According to the model, players look back on the history of choices made by nature and search for patterns. After updating their beliefs about the likelihood of the reappearance of each pattern, the players choose the best response to more likely patterns. A probabilistic decision rule governs the tendency of the players to choose (or not to choose) certain actions, along with the influence of the payoff for each action. The parameters of the model are estimated using data from a new experiment, and the model is compared with other existing learning models.