Seminars topic: In an era where organizations increasingly rely on data-driven systems to guide decisions, operations, and customer interactions, the trustworthiness of these socio-technical systems depends not only on analytical design but also on the quality and integrity of the underlying data. This talk presents research at the intersection of data quality, human-AI interaction, algorithmic fairness, and interpretable machine learning to advance trustworthy AI. The speaker focuses on a critical data integrity challenge - label bias - systematic errors in labeling that vary across social groups. To address this issue, the talk introduces Decoupled Confident Learning (DeCoLe), a principled machine learning framework designed to detect mislabeled data and improve data quality under label bias. Theoretical analysis and empirical evidence, including applications in hate speech detection, demonstrate that DeCoLe outperforms existing approaches in bias-aware mislabeling detection and provides actionable guidance for integrating trustworthy AI methods into organizational data management practices.

Scholars Background: Ms. LI Yunyi is a PhD candidate in Information Systems (Data Science track) at the McCombs School of Business, University of Texas at Austin. Her research interests include algorithmic fairness, trustworthy AI, human-AI interaction, and interpretable machine learning. Her work has been published in leading outlets such as INFORMS Journal on Computing, Notices of the American Mathematical Society, and Stat, and presented at major conferences including CIST, INFORMS, ICML, NeurIPS, and Big XII MIS Research Symposium. She has received the UT Austin Machine Learning Research Award, the Good Systems Award for Ethical AI Research, and NSF Big Data project funding.

Time and Location: January 6, 2026, 14:00–15:00, Room A523, School of Management

Language: EN

Host: Prof. CHEN Xi, School of Management, Zhejiang University