When majority rules, minority loses: bias amplification of gradient descent
This addresses bias in AI systems, which is a critical problem for fairness and equity, though it is incremental as it builds on existing empirical evidence.
The paper tackled bias amplification in machine learning by developing a formal framework for majority-minority learning tasks, showing that standard training favors majority groups and produces stereotypical predictors, with theoretical findings supported by experiments in deep learning for tabular and image classification.
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.