2026年第76期(总第1220期)

演讲主题：Social Choice Theory from Democracy to AI

主讲人：宋浩宇 普渡大学计算机科学系博士

主持人：袁世林 供应链管理与系统工程系讲师

活动时间：2026年10月21日（周三）15:00-17:00

活动地址：管院大楼107教室

主讲人简介：

宋浩宇本科毕业于美国科尔比学院（Colby College），于普渡大学计算机科学系获得博士学位。研究方向位于算法博弈论、运筹学与优化的交叉领域。

他的研究成果发表于SODA、EC、WINE、WWW以及Mathematics of Operations Research等算法与博弈论领域的国际顶级会议与期刊。近期研究聚焦于集体决策的高效算法、偏好学习，以及博弈论在大语言模型中的应用。

活动简介：

Computational social choice studies how to aggregate individual preferences into collective decisions while ensuring fairness, efficiency, and stability, with applications ranging from elections and participatory budgeting to matching markets, fair machine learning, and AI-driven collective decision-making. In this talk, we introduce Lindahl Equilibrium with Ordinal Preferences (LEO), a novel framework that extends classical Lindahl equilibrium to discrete settings where preferences are expressed through rankings rather than numerical utilities. Building on this framework, we establish several breakthroughs in fundamental problems: for participatory budgeting and proportional representation, we improve the approximation guarantee for core stability from 32 to 4.92; for collective decision-making, we prove that five candidates always suffice to form a Condorcet winning set, alongside asymptotically optimal bounds for generalized undominance; and for school choice with committee preferences, we develop an equilibrium-based approach to obtaining approximately stable matchings when schools evaluate applicants through multiple rankings.