Towards Fair and Explainable Supervised Learning
Published in ICML Workshop on Socially Responsible Machine Learning, 2021
Recommended citation: Aarshee Mishra*, Nicholas Perello*, and Przemyslaw A. Grabowicz. Towards Fair and Explainable Supervised Learning. In ICML Workshop on Socially Responsible Machine Learning (SRML), 2021.
Workshop paper presented at the ICML 2021 Workshop on Socially Responsible Machine Learning. It formalizes direct discrimination as a direct causal effect of protected attributes on decisions and induced discrimination as a change in the influence of associated non-protected features, and proposes post-processing methods that nullify the protected attribute’s influence while preserving the influence of the remaining features. The extended version appeared at ACM FAccT 2022 as Marrying Fairness and Explainability in Supervised Learning.