非对称理性下的策略学习
Asymmetric Rationality: Protocol Design and Behavioral Anomalies
AI总结:
本文针对完全理性的私人信息发送者与有限理性、有记忆约束的接收者间的动态互动,刻画最优决策协议,量化操纵空间与防范成本,揭示相关行为模式是系统性均衡反应,为策略环境程序理性提供框架。
摘要:
本文分析了完全理性、拥有私人信息的发送者与有限理性、无信息且存在记忆约束的接收者之间的动态互动。发送者控制信息流,接收者则设计决策协议,利用有限状态空间进行学习并提供激励。我们刻画了最优协议,量化了操纵空间及防范操纵的激励成本。研究表明,原本看似不稳定或受心理驱动的独特行为模式——如信息脱离、基于相同信息的观点极化、决策点附近的犹豫不决(弃权(不执行))——是对非对称理性和信息的系统性均衡反应。该模型为策略环境中的程序理性提供了一个富有表现力的框架。
英文摘要:
We study how asymmetric rationality in strategic interactions can generate seemingly anomalous patterns of belief updating and behavior. A boundedly rational receiver designs a finite-state decision protocol to learn from a fully rational, privately informed, and biased sender who controls the flow of information. The receiver must allocate scarce memory between learning and incentive provision. We identify a simple class of protocols sufficient for optimal design and show how their structure generates information disengagement, indecision, and polarization. The receiver hesitates before deciding but becomes unresponsive to further evidence afterward. Identical receivers following the same protocol can reach opposing decisions despite receiving the same sequence of signals, with the probability of disagreement remaining bounded away from zero as their payoffs approach the optimum. These behaviors, which might otherwise appear erratic or psychologically driven, arise from the strategic use of limited memory.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。