Path-Specific Counterfactual Fairness

2018/02/22 by Chiappa, Silvia, Gillam, Thomas P. S. · 10 citations
#FOS: Computer and information sciences #Machine Learning (stat.ML)

paper · doi:10.48550/arxiv.1802.08139

Abstract

We consider the problem of learning fair decision systems in complex scenarios in which a sensitive attribute might affect the decision along both fair and unfair pathways. We introduce a causal approach to disregard effects along unfair pathways that simplifies and generalizes previous literature. Our method corrects observations adversely affected by the sensitive attribute, and uses these to form a decision. This avoids disregarding fair information, and does not require an often intractable computation of the path-specific effect. We leverage recent developments in deep learning and approximate inference to achieve a solution that is widely applicable to complex, non-linear scenarios.

Cited by

• Desirable Effort Fairness and Optimality Trade-offs in Strategic Learning
• Algorithmic Fairness in AI Surrogates for End-of-Life Decision-Making
• Evidence Without Injustice: A New Counterfactual Test for Fair Algorithms
• Counterfactually Fair Conformal Prediction
• Partial Identification Approach to Counterfactual Fairness Assessment
• Calibrated Counterfactual Conformal Fairness (C³F): Post-hoc, Shift-Aware Coverage Parity via Conformal Prediction and Counterfactual Regularization
• Causal Fuzzing for Verifying Machine Unlearning
• CausalPre: Scalable and Effective Data Pre-Processing for Causal Fairness
• Fairness-Aware and Interpretable Policy Learning
• Algorithmic Fairness amid Social Determinants: Reflection, Characterization, and Approach

Related

• PC-Fairness: A Unified Framework for Measuring Causality-based Fairness
• Fairness-Accuracy Trade-Offs: A Causal Perspective
• Learning Individually Fair Classifier with Path-Specific Causal-Effect Constraint
• On Learning and Testing of Counterfactual Fairness through Data Preprocessing
• Towards Realistic Individual Recourse and Actionable Explanations in Black-Box Decision Making Systems