A Multi-Group Approach to Algorithmic Fairness - IPAM at UCLA

Explore a comprehensive lecture on algorithmic fairness presented by Guy Rothblum of Apple Inc. and Omer Reingold of Stanford University at IPAM's "Who Counts? Sex and Gender Bias in Data" workshop. Delve into the multi-group approach to addressing discrimination in predictive algorithms, examining key concepts such as risk prediction setup, group notions of fairness, and multi-calibration. Gain insights into the challenges of defining fairness in algorithms and learn about post-processing techniques for achieving multi-calibration. The 37-minute talk covers a range of topics, including the prevalence of predictive algorithms, concerns about discrimination, and the complexities of fairness definitions. Understand the importance of addressing bias in data and algorithms through this informative presentation, which was recorded on July 19, 2022, at the Institute for Pure & Applied Mathematics (IPAM) at UCLA.

Predictive Algorithms Everywhere
Concern: Discrimination
THE* Definition of Fairness?
Risk Prediction: Setup and Goal
Group Notions of Fairness
Multicalibration: Flavor of Results
Density Plot: Group (mis)Calibration
Post-Procesing for Multi-Calibration
Beyond Multi-Calibration