Learning Fair Scoring Functions: Bipartite Ranking under ROC-based Fairness Constraints

Abstract

Many applications of AI involve scoring individuals using a learned function of their attributes. These predictive risk scores are then used to take decisions based on whether the score exceeds a certain threshold, which may vary depending on the context. The level of delegation granted to such systems in critical applications like credit lending and medical diagnosis will heavily depend on how questions of fairness can be answered. In this paper, we study fairness for the problem of learning scoring functions from binary labeled data, a classic learning task known as bipartite ranking. We argue that the functional nature of the ROC curve, the gold standard measure of ranking accuracy in this context, leads to several ways of formulating fairness constraints. We introduce general families of fairness definitions based on the AUC and on ROC curves, and show that our ROC-based constraints can be instantiated such that classifiers obtained by thresholding the scoring function satisfy classification fairness for a desired range of thresholds. We establish generalization bounds for scoring functions learned under such constraints, design practical learning algorithms and show the relevance our approach with numerical experiments on real and synthetic data.

arXiv comment35 pages, 13 figures, 6 tables

References in corpus (13)

• Scikit-learn: Machine Learning in Python
• TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
• An overview of gradient descent optimization algorithms
• Equality of Opportunity in Supervised Learning
• Fairness of Exposure in Rankings
• FA*IR: A Fair Top-k Ranking Algorithm
• On Fairness and Calibration
• Learning Non-Discriminatory Predictors
• Ranking the best instances
• Policy Learning for Fairness in Ranking
• The Fairness of Risk Scores Beyond Classification: Bipartite Ranking and the xAUC Metric
• Fair lending needs explainable models for responsible recommendation
• A Probabilistic Theory of Supervised Similarity Learning for Pointwise ROC Curve Optimization

Cited by in corpus (5)

• Whither Bias Goes, I Will Go: An Integrative, Systematic Review of Algorithmic Bias Mitigation
• Optimizing Two-way Partial AUC with an End-to-end Framework
• Fairness in Forecasting of Observations of Linear Dynamical Systems
• Fairness Through Regularization for Learning to Rank
• Towards Model-Agnostic Post-Hoc Adjustment for Balancing Ranking Fairness and Algorithm Utility