报告题目：GenAI-Enabled Causal Study of Unstructured Data: Application in LLMs and Emotion Analysis

时间：2026年6月21日16:00-17:30

地点：中关村主楼317室

报告人：薛闻道

报告人国籍：中国

报告人政治面貌：群众

报告人职务：助理教授，Decision Sciences 编委会成员

报告人职称：助理教授

报告人工作单位：香港中文大学商学院

报告人简介：

薛闻道，香港中文大学商学院决策、运营与技术系助理教授。她曾在美国得克萨斯大学奥斯汀分校McCombs商学院信息、风险与运营管理系担任博士后研究员，并于华盛顿大学获得博士学位。她的研究主要聚焦于因果人工智能、数据分析和医疗健康领域。具体而言，她从三个方面开展因果人工智能研究：（1）探索利用人工智能和机器学习改进因果推断方法的新机会；（2）识别当前人工智能和机器学习应用于因果推断过程中面临的挑战；（3）研究人工智能的创新应用场景。相关论文被Management Science、Information System Research等期刊发表或接收。她致力于运用创新的数据分析方法，识别并解决具有挑战性的现实问题，以促进企业发展、改善医疗健康服务并创造社会价值。此外，她正在持续开展关于生成式人工智能对企业和组织影响的研究。

报告内容简介：

Information systems (IS) researchers often use machine learning algorithms to extract features in massive and unstructured online content and then study the causal effect of these features on various business outcomes. Generative AI (GenAI), as autom-atic unstructured data analyzers, offer new opportunities for IS researchers to advance the current causal study of unstructured data. Nevertheless, we show that directly plugging GenAI-generated variables into econometric models can induce bias in causa-l estimation. We propose a novel algorithm to correct such bias. A key feature is that the algorithm leverages predictions gener-ated by different GenAIs to form fuzzy interval data. The algorithm meets three design requirements. First, it is unsupervised, me-aning it does not need additional labeled data to correct the causal estimation. Second, it is flexible in incorporating the predict-ions of different GenAIs (leveraging the "wisdom of the AI crowd") and can be easily adapted to various causal models. Third, th-e causal estimators are theoretically guaranteed to achieve consistency. We apply our algorithm in the context of using LLMs as- emotion analyzers. This work provides important implications for causal inference with GenAIs, the causal study of unstructure-d data, and prescriptive analytics for fuzzy interval data.

（承办：管理工程系、科研与学术交流中心）