The fairness‐accuracy Pareto front
Authors:
- 27 October 2021
- journal article
- research article
- Published by Wiley in Statistical Analysis and Data Mining: An ASA Data Science Journal
- Vol. 15 (3) , 287-302
Abstract
Algorithmic fairness seeks to identify and correct sources of bias in machine learning algorithms. Confoundingly, ensuring fairness often comes at the cost of accuracy. We provide formal tools in this work for reconciling this fundamental tension in algorithm fairness. Specifically, we put to use the concept of Pareto optimality from multiobjective optimization and seek the fairness‐accuracy Pareto front of a neural network classifier. We demonstrate that many existing algorithmic fairness methods are performing the so‐called linear scalarization scheme, which has severe limitations in recovering Pareto optimal solutions. We instead apply the Chebyshev scalarization scheme which is provably superior theoretically and no more computationally burdensome at recovering Pareto optimal solutions compared to the linear scheme.Keywords
All Related Versions
Funding Information
- Australian Research Council (DE200101253)
This publication has 17 references indexed in Scilit:
- Balancing Covariates via Propensity Score WeightingJournal of the American Statistical Association, 2017
- Fairness Beyond Disparate Treatment & Disparate Impact2017
- Certifying and Removing Disparate Impact2015
- Model-Free Feature Screening for Ultrahigh Dimensional Discriminant AnalysisJournal of the American Statistical Association, 2015
- Methods for multi-objective optimization: An analysisInformation Sciences, 2015
- Pareto-Path Multitask Multiple Kernel LearningIEEE Transactions on Neural Networks and Learning Systems, 2014
- Hypervolume-based multiobjective optimization: Theoretical foundations and practical implicationsTheoretical Computer Science, 2012
- Normal-Boundary Intersection: A New Method for Generating the Pareto Surface in Nonlinear Multicriteria Optimization ProblemsSIAM Journal on Optimization, 1998
- A closer look at drawbacks of minimizing weighted sums of objectives for Pareto set generation in multicriteria optimization problemsStructural and Multidisciplinary Optimization, 1997
- Sex Bias in Graduate Admissions: Data from BerkeleyScience, 1975