Demographic-Agnostic Fairness without Harm

2025-09-28

AI Summary

Machine learning fairness in high-stakes domains (e.g., healthcare) faces a fundamental trade-off: group-based fairness criteria often degrade accuracy, while preference-based fairness relies on complete and explicit demographic information—frequently unavailable or privacy-sensitive. To address this, we propose DAFH (Demographic-Agnostic Fairness via Hierarchical learning), a framework that optimizes fairness without requiring explicit demographic attributes. DAFH jointly learns a latent group classifier and a decoupled predictor, implicitly discovering unknown subgroup structures and enforcing “harmless fairness”—i.e., maximizing subgroup-specific utility under fairness-aware constraints. Theoretical analysis establishes its statistical and optimization advantages. Experiments on synthetic and real-world datasets demonstrate that DAFH achieves significant fairness improvements while preserving high predictive accuracy. This work introduces a novel paradigm for fair machine learning in settings with missing demographic data or strict privacy constraints.

Technology Category

• Machine Learning: Ethics, Bias, and Fairness
• Computer Vision: Bias, Fairness & Privacy
• Humans and AI: Learning Human Values and Preferences

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

As machine learning (ML) algorithms are increasingly used in social domains to make predictions about humans, there is a growing concern that these algorithms may exhibit biases against certain social groups. Numerous notions of fairness have been proposed in the literature to measure the unfairness of ML. Among them, one class that receives the most attention is extit{parity-based}, i.e., achieving fairness by equalizing treatment or outcomes for different social groups. However, achieving parity-based fairness often comes at the cost of lowering model accuracy and is undesirable for many high-stakes domains like healthcare. To avoid inferior accuracy, a line of research focuses on extit{preference-based} fairness, under which any group of individuals would experience the highest accuracy and collectively prefer the ML outcomes assigned to them if they were given the choice between various sets of outcomes. However, these works assume individual demographic information is known and fully accessible during training. In this paper, we relax this requirement and propose a novel extit{demographic-agnostic fairness without harm (DAFH)} optimization algorithm, which jointly learns a group classifier that partitions the population into multiple groups and a set of decoupled classifiers associated with these groups. Theoretically, we conduct sample complexity analysis and show that our method can outperform the baselines when demographic information is known and used to train decoupled classifiers. Experiments on both synthetic and real data validate the proposed method.

Problem

• Achieving fairness without demographic information access
• Maintaining model accuracy while ensuring preference-based fairness
• Learning group classifiers and decoupled classifiers jointly

Innovation

• Demographic-agnostic fairness optimization without accuracy loss
• Jointly learns group classifier and decoupled classifiers
• Partitions population into groups for preference-based fairness