CAREER: Robust Fairness in Matching Markets

在社会中的多个关键应用中，资源分配带来了重大的决策挑战。匹配市场建立了一个有原则的框架，用于在有关各方-人类、机构、国家或自治机构-之间以稳健和高效的方式分配商品、服务和信息。在许多匹配和资源分配环境中，如择校、难民安置、食品捐赠和拼车，出于伦理、社会或法律方面的考虑，禁止或根本不可行使用货币转移支付。该项目的广泛目标是通过人工智能(AI)、经济学和计算的集成，开发一种在实际和大规模分配市场中实现稳健公平的理论基础方法。特别是，这项研究解决了由于不确定或不完全信息、市场的动态性质、复杂的约束以及影响结果的参与实体的战略行为而产生的公平问题。该项目将在联邦医疗保健和零工经济等实用领域创建新颖、公平和高效的人工智能系统。该项目还将为实施公平的制度奠定基础，以便以有原则和强有力的方式将学生与导师配对，为被辅导者指派导师，并安排会议。它将通过一个可公开访问的软件平台整合分配解决方案，以促进在日常决策中采用合理的公平解决方案，并将策划一个宝贵的学习资源，以达到学术界和教育界以外的公众和实践者。该项目通过弥合匹配理论和公平划分领域之间的差距，并能够创建新的近似算法，在设计新颖的公平人工智能系统方面做出了系统性的努力。它的目标是在以下相互关联的方向上取得进展：(I)不确定/噪声偏好下的匹配和分配，旨在设计对偏好信息中的不确定性或噪声具有健壮性的公平解决方案；(Ii)动态匹配中的公平性和多样性，其结合在线匹配和公平分割的技术来开发解决方案，以在动态市场中提供稳健的公平性和多样性保证；以及(Iii)双边市场中的激励和公平，这提供了一种系统的方法，将来自稳定匹配市场的技术与公平分工方面的最新进展相结合，在存在战略行为的情况下设计稳健的解决方案。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Resource allocation presents significant decision-making challenges across multiple critical applications in society. Matching markets establish a principled framework for distributing goods, services, and information among interested parties---humans, institutions, nations, or autonomous agents---in a robust and efficient manner. In many matching and resource allocation settings such as school choice, refugee placement, food donation, and ridesharing the use of monetary transfers is prohibited, or simply infeasible, due to ethical, societal, or legal concerns. The broad objective of this project is developing a theoretically grounded approach for robust fairness in practical and large-scale allocation markets through the integration of Artificial Intelligence (AI), economics, and computation. In particular, this research addresses fairness issues that arise due to uncertain or incomplete information, dynamic nature of markets, complex constraints, and strategic behavior of participating entities that influence the outcome. This project will lead to the creation of novel fair and efficient AI systems across practical domains such as federated healthcare and gig economies. This project will also lay the foundation for implementing fair systems for matching students to advisors, assigning mentors to mentees, and conference scheduling in a principled and robust manner. It will integrate allocation solutions through a publicly accessible software platform to foster the adoption of sound fair solutions in everyday decision-making, and will curate a valuable learning resource to reach the public and practitioners beyond academia and educators.This project makes a systematic effort in designing novel fair AI systems by bridging the gap between the fields of matching theory and fair division and enabling the creation of new approximate algorithms. It aims at making advances in the following interconnected directions: (i) Matching and allocation under uncertain/noisy preferences, that aims at devising fair solutions that are robust to uncertainty or noise in preference information, (ii) Fairness and diversity in dynamic matching, that combines techniques from online matching and fair division to develop solutions that provide robust fairness and diversity guarantees in dynamic markets, and (iii) Incentives and fairness in two-sided markets, that provides a systematic approach to blend techniques from stable matching markets with recent advances in fair division to design robust solutions in the presence of strategic behavior.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.