On the Interplay of Human-AI Alignment,Fairness, and Performance Trade-offs in Medical Imaging

2025-05-15

AI Summary

AI models for medical imaging often suffer from reduced fairness across diverse populations due to data biases, and inappropriate incorporation of human expert knowledge may further amplify such biases. Method: This work systematically investigates how the strength of Human-AI alignment affects both fairness and out-of-distribution (OOD) generalization, revealing—for the first time—a fundamental trade-off among alignment strength, fairness, and OOD performance. We propose a synergistic optimization framework integrating expert annotation guidance, adversarial debiasing, domain-generalization training, and fairness constraints. Contribution/Results: Evaluated on multiple medical imaging benchmarks, our method achieves an average 37% improvement in fairness metrics and a 12% gain in OOD accuracy, significantly enhancing model robustness and generalizability. The code is publicly available.

Technology Category

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

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
• Economics, Online Markets and Human Computation: Fairness and ethical considerations in crowd work and in human-in-the-loop AI systems

Abstract

Deep neural networks excel in medical imaging but remain prone to biases, leading to fairness gaps across demographic groups. We provide the first systematic exploration of Human-AI alignment and fairness in this domain. Our results show that incorporating human insights consistently reduces fairness gaps and enhances out-of-domain generalization, though excessive alignment can introduce performance trade-offs, emphasizing the need for calibrated strategies. These findings highlight Human-AI alignment as a promising approach for developing fair, robust, and generalizable medical AI systems, striking a balance between expert guidance and automated efficiency. Our code is available at https://github.com/Roypic/Aligner.

Problem

• Addressing fairness gaps in medical AI across demographic groups
• Exploring Human-AI alignment to reduce biases in medical imaging
• Balancing performance trade-offs between expert guidance and AI efficiency

Innovation

• Incorporating human insights reduces fairness gaps
• Balancing Human-AI alignment avoids performance trade-offs
• Calibrated strategies enhance out-of-domain generalization