中介沟通的边界
The Bounds of Mediated Communication
AI总结：
研究发送者-接收者博弈中，发送者偏好的中介在无法验证发送者报告时的沟通机制，比较中介沟通与廉价谈话和贝叶斯说服，发现中介仅在廉价谈话也能达到说服价值时才达到该价值，并刻画中介严格改进廉价谈话的条件。
摘要：
我们研究具有透明动机的发送者-接收者博弈，其中不知情但发送者偏好的中介承诺一个沟通机制，但无法验证发送者的报告。我们将中介沟通与廉价谈话和贝叶斯说服进行比较。一个信念-价值分布可通过中介实现当且仅当它满足贝叶斯合理性、接收者服从以及后验信念与发送者价值之间的零协方差。这一表述分离了承诺和可验证性的作用。我们表明，只有当廉价谈话也能达到说服价值时，中介才能达到该价值；因此，每当说服严格优于廉价谈话时，不可验证的报告会产生严格损失。我们刻画了中介严格改进廉价谈话的条件：发送者的价值必须在信念空间的某个方向上表现出抵消效应。在有限行动、二元状态的环境中，这一刻画给出了最优中介的可处理描述。应用涵盖平台、游说、接受博弈以及将模型重新解释为具有外部性的匹配，其中我们研究了效率-公平权衡。
英文摘要：
We study sender-receiver games with transparent motives, where an uninformed, sender-aligned mediator commits to a communication mechanism but cannot verify the sender's report. We compare mediated communication with cheap talk and Bayesian persuasion. A belief-value distribution is implementable by mediation exactly when it satisfies Bayes plausibility, receiver obedience, and zero covariance between posterior beliefs and sender values. This formulation separates the roles of commitment and verifiability. We show that mediation attains the persuasion value only when cheap talk also does; thus, whenever persuasion strictly outperforms cheap talk, unverifiable reports create a strict loss. We characterize when mediation strictly improves on cheap talk: the sender's value must exhibit countervailing effects along some direction in belief space. In finite-action, binary-state environments, this characterization yields a tractable description of optimal mediation. Applications cover platforms, lobbying, acceptance games, and a reinterpretation of the model as matching with externalities, where we study the efficiency-fairness tradeoff.
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。
