Abstract:The most prevalent notions of fairness in machine learning are statistical definitions: they fix a small collection of pre-defined groups, and then ask for parity of some statistic of the classifier across these groups. Constraints of this form are susceptible to intentional or inadvertent "fairness gerrymandering", in which a classifier appears to be fair on each individual group, but badly violates the fairness constraint on one or more structured subgroups defined over the protected attributes. We propose instead to demand statistical notions of fairness across exponentially (or infinitely) many subgroups, defined by a structured class of functions over the protected attributes. This interpolates between statistical definitions of fairness and recently proposed individual notions of fairness, but raises several computational challenges. It is no longer clear how to audit a fixed classifier to see if it satisfies such a strong definition of fairness. We prove that the computational problem of auditing subgroup fairness for both equality of false positive rates and statistical parity is equivalent to the problem of weak agnostic learning, which means it is computationally hard in the worst case, even for simple structured subclasses. We then derive two algorithms that provably converge to the best fair classifier, given access to oracles which can solve the agnostic learning problem. The algorithms are based on a formulation of subgroup fairness as a two-player zero-sum game between a Learner and an Auditor. Our first algorithm provably converges in a polynomial number of steps. Our second algorithm enjoys only provably asymptotic convergence, but has the merit of simplicity and faster per-step computation. We implement the simpler algorithm using linear regression as a heuristic oracle, and show that we can effectively both audit and learn fair classifiers on real datasets.

Simple and near-optimal algorithms for hidden stratification and multi-group learning

Minimax Group Fairness in Strategic Classification

Agnostic Multi-Group Active Learning

Group-wise oracle-efficient algorithms for online multi-group learning

Multi-group Learning for Hierarchical Groups

Learning by Grouping: A Multilevel Optimization Framework for Improving Fairness in Classification without Losing Accuracy

Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairness

Group conditional validity via multi-group learning

On Learning Fairness and Accuracy on Multiple Subgroups

Flexible Multi-Task Learning with Latent Task Grouping.

A theory of stratification learning

Regularization and Optimal Multiclass Learning

Optimal Multi-Distribution Learning

Supervised Multivariate Learning with Simultaneous Feature Auto-grouping and Dimension Reduction

A Mechanism Design Approach for Multi-Party Machine Learning

Algorithms and Learning for Fair Portfolio Design

Learning With Multi-Group Guarantees For Clusterable Subpopulations

Near-Optimal Resilient Aggregation Rules for Distributed Learning Using 1-Center and 1-Mean Clustering with Outliers

Mechanism Design for Multi-Party Machine Learning

Aggregating Data for Optimal and Private Learning

An Empirical Study of Rich Subgroup Fairness for Machine Learning