On the Interplay of Human-AI Alignment,Fairness, and Performance Trade-offs in Medical Imaging
This addresses fairness and robustness issues in medical AI systems, offering a calibrated approach to balance expert guidance with automated efficiency, but it is incremental as it builds on existing alignment concepts in a specific domain.
The paper tackles biases and fairness gaps in deep neural networks for medical imaging by exploring Human-AI alignment, showing that incorporating human insights reduces fairness gaps and improves out-of-domain generalization, though excessive alignment can cause performance trade-offs.
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.