发表机构
MIT（麻省理工学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究证明多阶段机制在序贯均衡下超越贝叶斯单调性的实施能力是脆弱的，通过构造任意小的信息扰动，使非贝叶斯单调的社会选择对应产生不可接受决策的概率远离零，且该结果在条件独立性或已知收益类型下失效。
AI 中文摘要
贝叶斯单调性是贝叶斯-纳什均衡中完全实施的必要条件。然而，在序贯均衡精炼下，多阶段机制可以扩大实施的范围。我们表明这种额外的实施能力是脆弱的。每当一个多阶段机制序贯实施了一个非贝叶斯单调的社会选择对应关系时，我们构造任意小的“信息扰动”，并证明在这些扰动下存在序贯均衡，这些均衡以远离零的概率产生不可接受的决策。在这些扰动下，参与者的类型在给定状态下不是条件独立的，并且某些参与者有时不确定其收益类型。我们通过例子表明，如果要求扰动满足条件独立性或“已知收益类型”，则一般结果不成立。
英文摘要
Bayesian monotonicity is a necessary condition for full implementation in Bayes-Nash equilibrium. Under the sequential equilibrium refinement, however, multi-stage mechanisms can expand the scope for implementation. We show that this additional implementation power is fragile. Whenever a multi-stage mechanism sequentially implements a social choice correspondence that is not Bayesian monotone, we construct arbitrarily small "information perturbations'' and show that there are sequential equilibria under these perturbations that yield inadmissible decisions with probability bounded away from zero. Under these perturbations, players' types are not independent conditional on the state and some player is sometimes uncertain of their payoff type. We show by example that the general result fails if perturbations are required to satisfy conditional independence or "known payoff types.''
