Simple and Effective Specialized Representations for Fair Classifiers

2025-05-16

AI Summary

Fair classification is critical in high-stakes decision-making, yet existing adversarial learning or distribution-matching approaches suffer from training instability and high computational overhead. This paper proposes a characteristic function distance (CFD)-based representation learning framework that explicitly minimizes leakage of sensitive information by measuring statistical dependence between sensitive attributes and learned representations, while preserving task-discriminative power. We introduce a novel convex relaxation objective embedding provably guaranteed group fairness constraints—such as demographic parity (DP) and equalized odds (EO)—enabling strict fairness enforcement without accuracy degradation. By eliminating adversarial training and implicit distribution alignment, our method significantly improves training stability and computational efficiency. Extensive experiments on multiple benchmark datasets demonstrate that our approach consistently outperforms state-of-the-art methods in both fairness metrics (e.g., ΔDP, ΔEO) and classification accuracy, exhibiting strong potential for industrial deployment.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Fair classification is a critical challenge that has gained increasing importance due to international regulations and its growing use in high-stakes decision-making settings. Existing methods often rely on adversarial learning or distribution matching across sensitive groups; however, adversarial learning can be unstable, and distribution matching can be computationally intensive. To address these limitations, we propose a novel approach based on the characteristic function distance. Our method ensures that the learned representation contains minimal sensitive information while maintaining high effectiveness for downstream tasks. By utilizing characteristic functions, we achieve a more stable and efficient solution compared to traditional methods. Additionally, we introduce a simple relaxation of the objective function that guarantees fairness in common classification models with no performance degradation. Experimental results on benchmark datasets demonstrate that our approach consistently matches or achieves better fairness and predictive accuracy than existing methods. Moreover, our method maintains robustness and computational efficiency, making it a practical solution for real-world applications.

Problem

• Ensuring fair classification with minimal sensitive information
• Overcoming instability and computational intensity in existing methods
• Achieving fairness without degrading predictive accuracy

Innovation

• Uses characteristic function distance for fairness
• Ensures minimal sensitive information in representations
• Simple relaxation guarantees fairness without performance loss