具有粗信念的评分拍卖
Scoring Auctions with Coarse Beliefs
AI总结:
本文在无共同先验的机制设计场景下构建含粗信息获取的拍卖模型,刻画评分拍卖存在粗信念均衡的性质,提出区分两类拍卖的规则条件,并用于现实拍卖的策略简洁性分类。
摘要:
本文研究机制设计场景中的一种简洁性概念,该场景下智能体未必共享共同先验。我构建了一个模型:智能体先参与一个无先验的(粗)信息获取博弈,随后进行拍卖。获取信息后,智能体对其所处的博弈环境以及对手的高阶信念存在不确定性。若即便存在这种不确定性,智能体仍能采取最优反应,则称该机制存在粗信念均衡。聚焦于多维评分拍卖,我完整刻画了一种能让拍卖形式存在粗信念均衡的性质。主要结果将拍卖分为两类:一类是智能体只需了解相对较少的自身场景相关信息,另一类则是智能体必须完全掌握类型分布才能形成均衡策略。随后我找到了一个关于拍卖规则的简单原生条件,用于区分这两类拍卖。我还利用该条件根据策略简洁性对现实中的评分拍卖进行了分类。
英文摘要:
This paper studies a simplicity notion in a mechanism design setting in which agents do not necessarily share a common prior. I develop a model in which agents participate in a prior-free game of (coarse) information acquisition followed by an auction. After acquiring information, the agents have uncertainty about the environment in which they play and about their opponents' higher-order beliefs. A mechanism admits a coarse beliefs equilibrium if agents can play best responses even with this uncertainty. Focusing on multidimensional scoring auctions, I fully characterize a property that allows an auction format to admit coarse beliefs equilibria. The main result classifies auctions into two sets: those in which agents learn relatively little about their setting versus those in which they must fully learn a type distribution to form equilibrium strategies. I then find a simple, primitive condition on the auction's rules to distinguish between these two classes. I then use the condition to categorize real-world scoring auctions by their strategic simplicity.
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。