Safety
When majority rules, minority loses: bias amplification of gradient descent
arXiv:2505.13122v3 Announce Type: replace-cross Abstract: Despite growing empirical evidence of bias amplification in machine learning, its theoretical foundations remain poorly understood. We develop
arXiv:2505.13122v3 Announce Type: replace-cross Abstract: Despite growing empirical evidence of bias amplification in machine learning, its theoretical foundations remain poorly understood. We develop a formal framework for majority-minority learning tasks, showing how standard training can favor majority groups and produce stereotypical predictors that neglect minority-specific features. Assuming population and variance imbalance, our analysis reveals three key findings: (i) the close proximity between ``full-data'' and stereotypical predictors, (ii) the dominance of a region where training the entire model tends to merely learn the majority traits, and (iii) a lower bound on the additional training required. Our results are illustrated through experiments in deep learning for tabular and image classification tasks.
Related
- The Effect of Mini-Batch Noise on the Implicit Bias of Adam
- A Woman with a Knife or A Knife with a Woman? Measuring Directional Bias Amplification in Image Captions
- When to Stop Reusing: Dynamic Gradient Gating for Sample-Efficient RLVR
- BRAIN: Bias-Mitigation Continual Learning Approach to Vision-Brain Understanding
Source: arXiv cs.AI | 2026-09-16