Imperfect Play in Games of Incomplete Information: Theory and Experiments

我们建议继续对量子响应平衡(QRE)进行理论和实验研究，QRE是博弈中均衡行为的统计模型。均衡的关键特征是，个体并不总是对对手的策略做出最好的反应，而是以更高的概率打出更好的策略，而不是更差的策略。在这里，我们建议将QRE应用于三个不同的应用领域：拍卖、投票和信息聚合。计划了几个不同的系列实验，包括大规模实验(100多名受试者)，以测试关于大群体中的信息聚合和战略投票的理论预测。在理论方面，我们建议将该模型扩展到动态设置，以包括学习和重复玩，并概括错误结构，以允许跨动作的关联。该提案还有一个计算部分。具体地说，我们计划以几种方式扩展我们目前计算QRE的算法，以允许在某些类型的较大博弈中更有效地计算均衡，并允许在具有更一般误差结构的模型中进行估计。

We propose to continue our theoretical and experimental research studying Quantal Response Equilibria (QRE), a statistical model of equilibrium behavior in games. The key feature of the equilibrium is that individuals do not always play best responses to the strategies of their opponents, but play better strategies with higher probability than worse strategies. Here we propose to apply QRE to three different areas of application: auctions, voting, and information aggregation. Several different series of experiments are planned, including large-scale experiments (100+ subjects) to test theoretical predictions about information aggregation and strategic voting in large groups. On the theoretical side, we propose to extend the model to dynamic settings to incorporate learning and repeated play, and to generalize the error structure to allow for correlation across moves. There is also a computational part of the proposal. Specifically, we plan to extend our current algorithms for computing QRE in several ways, to allow more efficient computation of equilibria in certain kinds of larger games and to allow for estimation in models with more general error structures.