Towards Understanding and Mitigating Social Biases in Language Models

TL;DR

This paper identifies social biases in large language models, introduces benchmarks and metrics to measure them, and proposes mitigation strategies that balance fairness and text quality.

Contribution

It defines sources of biases, develops new benchmarks and metrics, and presents mitigation methods that improve fairness without sacrificing text fidelity.

Findings

Bias mitigation improves fairness in generated text

Metrics effectively quantify social biases

Mitigation methods maintain high-quality text generation

Abstract

As machine learning methods are deployed in real-world settings such as healthcare, legal systems, and social science, it is crucial to recognize how they shape social biases and stereotypes in these sensitive decision-making processes. Among such real-world deployments are large-scale pretrained language models (LMs) that can be potentially dangerous in manifesting undesirable representational biases - harmful biases resulting from stereotyping that propagate negative generalizations involving gender, race, religion, and other social constructs. As a step towards improving the fairness of LMs, we carefully define several sources of representational biases before proposing new benchmarks and metrics to measure them. With these tools, we propose steps towards mitigating social biases during text generation. Our empirical results and human evaluation demonstrate effectiveness in mitigating bias while retaining crucial contextual information for high-fidelity text generation, thereby pushing forward the performance-fairness Pareto frontier.