报告题目：Algorithmic Collusion or Competition: The Role of Platforms'  Recommender Systems

时间：2026年6月21日13:00-14:30

地点：中关村主楼317室

报告人：Yong Tan（谭勇）教授

报告人国籍：美国

报告人职务：信息系统和运营管理系主任

报告人职称：教授

报告人工作单位：华盛顿大学Michael G. Foster商学院

报告人简介：

Yong Tan（谭勇），华盛顿大学Michael G. Foster商学院信息系统讲席教授，担任信息系统和运营管理系主任，清华大学经济管理学院长江学者讲座教授，INFORMS信息系统协会杰出会士，以及华盛顿州科学院院士。他师从2016年诺贝尔奖得主David J. Thouless教授获得华盛顿大学物理学博士学位，并在华盛顿大学获得工商管理博士学位。他的研究兴趣包括电子商务，移动和社交商务，大数据，Al，信息系统经济学，社会和经济网络，共享经济以及健康IT。他的成果发表于Management Science, Information Systems Research, Operations Research, Management Information Systems Quarterly, Journal of Management Information Systems, Production and Operations Management, INFORMS Journal on Computing, IEEE/ACM Transactions on Networking, IEEE Transactions on Software Engineering, IEEE Transactions on Knowledge and Data Engineering, IEEE Transactions, European Journal on Operations Research, Decision Support Systems。他曾担任Information Systems Research高级主编和和副主编，Management Science的副主编,现在是Production and Operations Management高级编辑和Journal of Management Information Systems编委会成员。他是2010年CIST的联合主席，2012年INFORMS的分会主席，2013、2021年ICIS的联合主席，2014年WITS的联合主席，2019年INFORMS数据科学研讨会的联合主席。他获得2017年Management Science的最佳论文, AIS2012年最佳论文，2012年ISR最佳论文第二名。他指导的学生在Carnegie Mellon University, Purdue University, Indiana University, University of Notre Dame, University of Texas at Dallas,Georgia State University, University of Florida, Arizona State University,和University of California,Irvine等世界知名大学任职。

报告内容简介：

Recent research and regulations have increasingly focused on algorithmic collusion driven by AI-enabled pricing algorithms. H-owever, on online platforms where such pricing algorithms are most prevalent, recommender systems also operate alongside the-m. To examine this unique interaction between two types of AI, we propose a novel repeated game framework that integrates seve-ral key components. First, we develop and estimate a structural search model to characterize consumers' decision-making process-es in response to varying recommendation sets. Building on this consumer model, we formulate personalized recommendation al-gorithms designed to maximize platform revenue and/or consumer utility. We then introduce pricing algorithms for sellers and in-tegrate all these elements to conduct numerical experiments. Our results reveal that a revenue-maximizing recommender system intensifies algorithmic collusion, whereas a utility-maximizing recommender system promotes greater competition. Moreover, wei-ghted objective recommendations that balance both sides yield two new fairness notions, one targeting price equivalence and the other targeting utility equivalence relative to the no-recommendation baseline. Further, increasing the size of recommendation set-s does not consistently enhance the platform's objectives. This unexpected “more is less” effect arises because expanding the r-ecommendation set weakens the recommender system's ability to effectively influence pricing algorithms through dynamic reward-s. Horizontal differentiation further moderates this effect by altering the relative difficulty of collusion. These findings provide valu-able insights for regulators, platform designers, and market participants, and open up promising avenues for future research.

（承办：管理工程系、科研与学术交流中心）