FAIR-SIGHT: Fairness Assurance in Image Recognition via Simultaneous Conformal Thresholding and Dynamic Output Repair

2025-04-10

AI Summary

This work addresses group- and individual-level fairness disparities in image recognition systems. We propose a post-hoc fairness assurance framework that requires no model retraining and is parameter-agnostic. Methodologically, we jointly model prediction error and fairness bias, introducing a fairness-aware nonconformity scoring mechanism. To our knowledge, this is the first approach to integrate conformal prediction with dynamic output repair, enabling distribution-free fairness error control under limited samples. Furthermore, we incorporate logit-space bias correction and confidence-based dynamic recalibration, supported by theoretical convergence analysis. Experiments across multiple benchmark datasets demonstrate substantial reductions in fairness gaps—e.g., a 38% decrease in Equalized Odds—while preserving over 99% of the original model’s accuracy.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Calibration & Uncertainty Quantification
• Philosophy and Ethics of AI: Bias, Fairness & Equity

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
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

We introduce FAIR-SIGHT, an innovative post-hoc framework designed to ensure fairness in computer vision systems by combining conformal prediction with a dynamic output repair mechanism. Our approach calculates a fairness-aware non-conformity score that simultaneously assesses prediction errors and fairness violations. Using conformal prediction, we establish an adaptive threshold that provides rigorous finite-sample, distribution-free guarantees. When the non-conformity score for a new image exceeds the calibrated threshold, FAIR-SIGHT implements targeted corrective adjustments, such as logit shifts for classification and confidence recalibration for detection, to reduce both group and individual fairness disparities, all without the need for retraining or having access to internal model parameters. Comprehensive theoretical analysis validates our method's error control and convergence properties. At the same time, extensive empirical evaluations on benchmark datasets show that FAIR-SIGHT significantly reduces fairness disparities while preserving high predictive performance.

Problem

• Ensures fairness in computer vision systems post-hoc
• Combines conformal prediction with dynamic output repair
• Reduces group and individual fairness disparities without retraining

Innovation

• Combines conformal prediction with dynamic output repair
• Calculates fairness-aware non-conformity score for errors
• Implements corrective adjustments without retraining model