Algorithmic Fairness, Loss Minimization and Outcome Indistinguishability

Explore a 51-minute lecture on algorithmic fairness, loss minimization, and outcome indistinguishability presented by Omer Reingold from Stanford University at IPAM's EnCORE Workshop. Delve into the limitations of traditional loss minimization in machine learning when addressing algorithmic fairness concerns. Examine multi-group fairness notions like multicalibration and their connection to computational indistinguishability. Discover how outcome indistinguishability offers an alternative paradigm for training predictors that can be applied to various loss functions, capacity constraints, fairness requirements, and instance distributions. Gain insights into this growing field with applications beyond fairness in machine learning.

Omer Reingold - Algorithmic Fairness, Loss Minimization and Outcome Indistinguishability