发表机构
University of Bonn（波恩大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究委托人未知状态分布时的稳健委托问题，提出最多三区域的随机机制，并证明多维可加偏好下跨维度分别委托为稳健最优。
AI 中文摘要
我们考虑一个稳健委托问题，其中委托人不知道潜在状态所服从的分布。委托人可以选择一个一般的随机机制，并在所有状态分布上最大化其最坏情况下的期望收益。我们的主要结果刻画了稳健最优机制。该机制最多包含三个区域：（i）容纳区，其中采纳代理人的理想行动；（ii）校准随机化区，其中每种类型获得不同的彩票；（iii）合并区，其中所有类型获得相同的彩票。随后，我们将模型扩展到多维设定，其中偏好是可加可分的，并满足跨维度对称条件，并证明跨维度分别委托是稳健最优的。
英文摘要
We consider a robust delegation problem in which the principal does not know the distribution from which the underlying state is drawn. The principal can choose a general randomized mechanism and maximizes her worst-case expected payoff over all state distributions. Our main result characterizes the robustly optimal mechanism. The mechanism has up to three regions: (i) accommodation, where the agent's ideal action is taken, (ii) calibrated randomization, where each type receives a distinct lottery, and (iii) pooling, where all types receive the same lottery. We then extend our model to a multidimensional setting in which preferences are additively separable and satisfy a cross-dimensional symmetry condition, and show that delegating separately across dimensions is robustly optimal.
