CAREER: Design of Two-sided Markets In Dynamic and Uncertain Environments: Theoretical and Operational Issues

该学院早期职业发展（Career）项目奖的研究目标是研究动态和不确定环境下双边市场的设计。由于市场双方的随机到达，系统中的代理数量会随着时间的推移而变化。代理在进入系统时报告他们的偏好，并根据偏好的兼容性以及未来选项的价值接受或拒绝匹配提供。该奖项的主要目标是通过考虑到动态交互、不确定性和代理的前瞻性行为，以一种有效的方式设计这样的市场。研究方法是计算在给定政策下，市场设计者给出的效用承诺下，每个agent保持不匹配的预期收益，这将使（动态）个体理性匹配出价的算法成为可能。该算法将推广到agent在比赛前可能不知道自己对比赛的真实偏好，并且双方都产生了多个比赛的环境中。进一步的分析将赋予代理人诚实行为的直接激励，并考虑市场设计者通过观察代理人对匹配报价的反应来了解代理人偏好的可能性，而代理人通过观察他们得到的报价来了解可用合作伙伴的分布。如果成功，本研究的结果将导致市场布局的改善，这些市场需要匹配劳动力市场、器官移植、儿童收养和寄养等中介的双边人口。该奖项可以提供建模和分析工具，以最大限度地提高匹配系统的效率和参与者的福利。这项研究是通过结合经济学中的匹配/市场设计和作业中常用的工具而实现的。所开发的数学还将有助于培养研究生和本科生，并有助于卡耐基梅隆大学和宾夕法尼亚收养交换中心（PAE）之间正在进行的公私伙伴关系。

The research objective of this Faculty Early Career Development (CAREER) Program award is to study the design of two-sided markets in dynamic and uncertain settings. The population of agents in the system evolves over time due to stochastic arrivals to both sides of the market. Agents report their preferences upon entering the system, and either accept or reject match offers based on the compatibility of preferences as well as value of future options. The primary goal of this award is to design such marketplaces in an efficient way by taking into account dynamic interaction, uncertainty and forward looking behavior of agents. The research approach is to compute each agent's expected benefit from remaining unmatched as a utility promise from the market designer under a given policy, which will enable an algorithm for (dynamically) individually-rational match offers. The algorithm will be generalized to the environment where agents may not know their true preferences over matches before the match and there are multiple matches generated on both sides. Further analysis will endow agents with straightforward incentives for truthful behavior and consider the possibility that a market designer learns about agent preferences by observing their responses to match offers while the agents learn about the distribution of available partners by observing the offers they get.If successful, the results of this research will lead to improvements in the layout of markets where there is a need to match two-sided populations of agents such as labor markets, organ transplant, child adoption, and foster care. The award can provide modeling and analysis tools to maximize the efficiency of the matching system and the welfare of its participants. The research is made possible by combining tools developed both in matching/market design in economics and those commonly used in operations. The mathematics developed will also help train graduate and undergraduate students as well as contribute to an ongoing private public partnership between Carnegie Mellon and Pennsylvania Adoption Exchange (PAE).