A Systematic Study of Bias Amplification
This work addresses bias amplification in ML models, which is a critical issue for fairness and ethics in AI, but it is incremental as it builds on prior research to provide empirical insights.
The paper tackles the problem of bias amplification in machine learning models by conducting a systematic, controlled study using a synthetic image-classification task, revealing that bias amplification correlates with factors like model accuracy, capacity, and training data, and occurs primarily when recognizing group membership is easier than class membership.
Recent research suggests that predictions made by machine-learning models can amplify biases present in the training data. When a model amplifies bias, it makes certain predictions at a higher rate for some groups than expected based on training-data statistics. Mitigating such bias amplification requires a deep understanding of the mechanics in modern machine learning that give rise to that amplification. We perform the first systematic, controlled study into when and how bias amplification occurs. To enable this study, we design a simple image-classification problem in which we can tightly control (synthetic) biases. Our study of this problem reveals that the strength of bias amplification is correlated to measures such as model accuracy, model capacity, model overconfidence, and amount of training data. We also find that bias amplification can vary greatly during training. Finally, we find that bias amplification may depend on the difficulty of the classification task relative to the difficulty of recognizing group membership: bias amplification appears to occur primarily when it is easier to recognize group membership than class membership. Our results suggest best practices for training machine-learning models that we hope will help pave the way for the development of better mitigation strategies. Code can be found at https://github.com/facebookresearch/cv_bias_amplification.