On the Correlation between Individual Fairness and Predictive Accuracy in Probabilistic Models

2025-09-16

AI Summary

This work investigates the intrinsic relationship between individual fairness and predictive accuracy in generative probabilistic classifiers, focusing on whether posterior robustness to perturbations of sensitive attributes can serve as a unifying principle for jointly enhancing both properties. Method: We propose a “robustness–fairness–accuracy” positive-correlation hypothesis and formulate individual fairness constraints as a Most Probable Explanation (MPE) problem within an auxiliary Markov Random Field, enabling joint optimization of fairness and accuracy. Using Bayesian networks as the generative model, we integrate robustness analysis with probabilistic inference. Contribution/Results: Extensive evaluation across 14 diverse datasets demonstrates that posterior robustness exhibits significant positive correlation with both individual fairness and classification accuracy. This study is the first to reveal—through a generative modeling lens—that fairness and accuracy need not be traded off, establishing a novel paradigm and scalable technical pathway for building classifiers that are simultaneously highly accurate and individually fair.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Reasoning under Uncertainty: Relational Probabilistic Models
• Computer Vision: Adversarial Attacks & Robustness

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
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

We investigate individual fairness in generative probabilistic classifiers by analysing the robustness of posterior inferences to perturbations in private features. Building on established results in robustness analysis, we hypothesise a correlation between robustness and predictive accuracy, specifically, instances exhibiting greater robustness are more likely to be classified accurately. We empirically assess this hypothesis using a benchmark of fourteen datasets with fairness concerns, employing Bayesian networks as the underlying generative models. To address the computational complexity associated with robustness analysis over multiple private features with Bayesian networks, we reformulate the problem as a most probable explanation task in an auxiliary Markov random field. Our experiments confirm the hypothesis about the correlation, suggesting novel directions to mitigate the traditional trade-off between fairness and accuracy.

Problem

• Investigating correlation between individual fairness and predictive accuracy
• Analyzing robustness of posterior inferences to private feature perturbations
• Addressing computational complexity in fairness-accuracy trade-off analysis

Innovation

• Bayesian networks for generative probabilistic classifiers
• Reformulating robustness as Markov random field task
• Empirical correlation analysis between fairness and accuracy