Individual Differences in EWA Learning With Partial Payoff Information
将经验加权吸引力学习扩展到只知道未选策略可能收益的博弈中，为每个玩家单独估计参数以研究异质性，发现三种假设均提高预测精度，且玩家分为两个子群。
We extend experience-weighted attraction (EWA) learning to games in which only the set of possible \nforegone payoffs from unchosen strategies are known, and estimate parameters separately for each \nplayer to study heterogeneity. We assume players estimate unknown foregone payoffs from a strategy, \nby substituting the last payoff actually received from that strategy, by clairvoyantly guessing the actual \nforegone payoff, or by averaging the set of possible foregone payoffs conditional on the actual \noutcomes. All three assumptions improve predictive accuracy of EWA. Individual parameter estimates \nsuggest that players cluster into two separate subgroups (which differ from traditional reinforcement \nand belief learning).