Removing biased data to improve fairness and accuracy

Abstract

Machine learning systems are often trained using data collected from historical decisions. If past decisions were biased, then automated systems that learn from historical data will also be biased. We propose a black-box approach to identify and remove biased training data. Machine learning models trained on such debiased data (a subset of the original training data) have low individual discrimination, often 0%. These models also have greater accuracy and lower statistical disparity than models trained on the full historical data. We evaluated our methodology in experiments using 6 real-world datasets. Our approach outperformed seven previous approaches in terms of individual discrimination and accuracy.

16 pages, 5 Figures, 8 Tables

References in corpus (9)

• Equality of Opportunity in Supervised Learning
• Fairness Testing: Testing Software for Discrimination
• Data Decisions and Theoretical Implications when Adversarially Learning Fair Representations
• On Fairness and Calibration
• A Convex Framework for Fair Regression
• What's Sex Got To Do With Fair Machine Learning?
• A statistical framework for fair predictive algorithms
• Provably Fair Representations
• Learning unbiased features