Adaptive Human-Oversight Allocation in AI-Enabled Analytics
讲座通知
2026 年 10 月 15 日（星期四）；
上午 10:00 - 11:00
信息管理与工程学院308室
上海财经大学（第三教学楼西侧）
上海市杨浦区武东路100号
主题
Adaptive Human-Oversight Allocation in AI-Enabled Analytics
主讲人
Jiameng Lyu
Fudan University
Jiameng Lyu is an Assistant Professor in the Department of Management Science at the School of Management, Fudan University. He obtained his Ph.D. in Applied Mathematics from Tsinghua University and received his B.S. in Statistics from the University of Science and Technology of China. His research interests center around data-driven decision-making, including online learning, online optimization, statistical machine learning, and their applications to OR/OM. He is also interested in how to allocate human oversight resources in AI-assisted decision-making. His work has been published in journals such as Management Science, Operations Research, and Production and Operations Management.
讲座简介
Artificial intelligence provides low-cost predictions for many analytics tasks, but their reliability and the value of another human label vary across tasks. We study how to allocate a limited human-validation budget while learning, from the labels purchased, how much uncertainty remains after estimation assisted by artificial intelligence. For known task difficulties, the allocation benchmark is a cost-adjusted Neyman rule; we develop an online policy based on upper confidence bounds for unknown difficulties and establish allocation regret bounds for the stated confidence-radius policy. In synthetic experiments, the policy tracks the oracle allocation and reduces realized estimation error by about one quarter relative to uniform allocation at 2,000 labels. On a 68-question digital-twin survey pool, it achieves the lowest allocation objective among the tested implementable policies and saves up to 4.5% of labels at equal precision. Further analyses examine how task heterogeneity and adaptive sampling affect these gains.
