统计与数据科学系系列学术报告之五百零六期

时间：2026年4月1日（星期三）09:30-10:30

主持人：复旦大学管理学院统计与数据科学系黎德元教授

地点：史带楼403室

报告人：苗旺副教授北京大学

题目：Identifying the Desert Decision Rule to Assess and Achieve Fairness

摘要： The fairness of statistical and machine learning models has become a prominent concern. When data encode historical discrimination against certain demographic groups, such as race or gender, models trained on such data may inherit and reproduce these biases, leading to unfair or unwarranted predictions. In this paper, we propose a novel framework for characterising and addressing fairness issues by introducing the notion of desert decision, a latent variable representing the decision an individual rightfully deserves based on their actions, efforts, or abilities. We advocate making decisions by predicting the desert decision, in contrast to existing methods that typically focus on predicting the observed decision subject to fairness constraints. We propose to assess the degree of unfairness in the data by measuring the discrepancy between desert and observed decisions. We establish identification results under causally interpretable assumptions on the fairness property of the desert decision and the unfairness mechanism of the observed decision. For estimation, we develop a sieve maximum likelihood estimator for the target decision rule and an influence-function-based estimator for the degree of unfairness. Sensitivity analysis procedures are further proposed to assess the robustness of our methods to violations of identifying assumptions.

个人简介：苗旺现为北京大学概率统计系和统计科学中心副教授，2008-2017年在北京大学数学科学学院读本科和博士，2017-2018年在哈佛大学生物统计系做博士后研究，2018年入职北京大学。苗旺的研究兴趣包括因果推断，缺失数据，半参数统计及其应用，与合作者提出混杂分析的代理推断理论，发展非随机缺失数据的识别性和双稳健估计理论，以及数据融合的半参数理论，获得自然科学基金原创探索项目和国家重点研发计划青年科学家项目资助。担任中国现场统计研究会因果推断分会常务副理事长。