Using Self-supervised Learning Can Improve Model Fairness

TL;DR

This paper investigates how self-supervised learning (SSL) can enhance model fairness across demographic groups, showing that SSL can significantly improve fairness with minimal performance loss compared to supervised methods.

Contribution

The study introduces a comprehensive fairness assessment framework for SSL and demonstrates that SSL models can achieve up to 30% greater fairness across real-world datasets.

Findings

SSL improves fairness up to 30% with minimal performance loss

Representation dissimilarities correlate with demographic performance gaps

SSL models maintain comparable performance to supervised models

Abstract

Self-supervised learning (SSL) has become the de facto training paradigm of large models, where pre-training is followed by supervised fine-tuning using domain-specific data and labels. Despite demonstrating comparable performance with supervised methods, comprehensive efforts to assess SSL's impact on machine learning fairness (i.e., performing equally on different demographic breakdowns) are lacking. Hypothesizing that SSL models would learn more generic, hence less biased representations, this study explores the impact of pre-training and fine-tuning strategies on fairness. We introduce a fairness assessment framework for SSL, comprising five stages: defining dataset requirements, pre-training, fine-tuning with gradual unfreezing, assessing representation similarity conditioned on demographics, and establishing domain-specific evaluation processes. We evaluate our method's generalizability on three real-world human-centric datasets (i.e., MIMIC, MESA, and GLOBEM) by systematically comparing hundreds of SSL and fine-tuned models on various dimensions spanning from the intermediate representations to appropriate evaluation metrics. Our findings demonstrate that SSL can significantly improve model fairness, while maintaining performance on par with supervised methods-exhibiting up to a 30% increase in fairness with minimal loss in performance through self-supervision. We posit that such differences can be attributed to representation dissimilarities found between the best- and the worst-performing demographics across models-up to x13 greater for protected attributes with larger performance discrepancies between segments.