Faculté et Recherche
From Defining Privacy to Modeling It: A Machine Learning Perspective
10 avr
2026
14H00 - 15H15
Jouy-en-Josas
Anglais
Participer
Information Systems and Operations Management
Speaker: Heng Xu from University of Notre-Dame
Room T-020
Abstract
This talk explores what machine learning can teach us about the future of privacy research. I argue that privacy has resisted unified theorizing because it is inherently complex, context dependent, and often difficult for individuals themselves to articulate in rule-based terms. Drawing on the history of AI -- from symbolic systems to modern statistical learning -- I suggest that privacy research may advance by focusing less on defining privacy once and for all, and more on identifying underlying structures in the ways individuals make sense of privacy across contexts. I discuss how heterogeneity and situational uncertainty shape privacy judgments, why these dynamics complicate empirical research, and what they imply for the governance of contemporary AI systems. The talk concludes by highlighting emerging challenges such as inference-based privacy harms, and the growing tradeoffs among privacy, fairness, and substantive meaningfulness in AI.