Social Bias Probing: Fairness Benchmarking for Language Models

TL;DR

This paper introduces SoFa, a comprehensive fairness benchmark for language models that evaluates nuanced social biases across diverse identities, revealing more complex biases than previously understood.

Contribution

It presents a novel framework for bias probing based on disparate treatment and curates the large-scale SoFa benchmark to improve bias analysis in language models.

Findings

Biases are more nuanced than binary stereotypes.

Religious identities show the most pronounced disparate treatment.

Models reflect real-life adversities faced by marginalized groups.

Abstract

While the impact of social biases in language models has been recognized, prior methods for bias evaluation have been limited to binary association tests on small datasets, limiting our understanding of bias complexities. This paper proposes a novel framework for probing language models for social biases by assessing disparate treatment, which involves treating individuals differently according to their affiliation with a sensitive demographic group. We curate SoFa, a large-scale benchmark designed to address the limitations of existing fairness collections. SoFa expands the analysis beyond the binary comparison of stereotypical versus anti-stereotypical identities to include a diverse range of identities and stereotypes. Comparing our methodology with existing benchmarks, we reveal that biases within language models are more nuanced than acknowledged, indicating a broader scope of encoded biases than previously recognized. Benchmarking LMs on SoFa, we expose how identities expressing different religions lead to the most pronounced disparate treatments across all models. Finally, our findings indicate that real-life adversities faced by various groups such as women and people with disabilities are mirrored in the behavior of these models.