Special Seminar in CMS and HSS

The vast collection of detailed personal data has enabled machine learning to have a tremendous impact on society. Algorithms now provide predictions and insights that are used to make or inform consequential decisions on people. Concerns have been raised that our heavy reliance on personal data and machine learning might compromise people's privacy, produce new forms of discrimination, and violate other kinds of social norms. My research seeks to address this emerging tension between machine learning and society by focusing on two interconnected questions: 1) how to make machine learning better aligned with societal values, especially privacy and fairness, and 2) how to make machine learning methods more reliable and robust in social and economic dynamics. In this talk, I will provide an overview of my research and highlight some of my recent work on fairness in machine learning and differentially private synthetic data generation.