Language-Agnostic Bias Detection in Language Models with Bias Probing

TL;DR

This paper introduces LABDet, a robust, language-agnostic bias probing method for pretrained language models, revealing consistent nationality biases across languages and linking these biases to pretraining data.

Contribution

The paper presents LABDet, a novel bias detection technique that is robust, language-agnostic, and capable of directly linking biases to pretraining data in language models.

Findings

Nationality bias patterns align with historical context across six languages.

Bias in English BERT correlates with pretraining data biases.

LABDet is reliable across different templates and languages.

Abstract

Pretrained language models (PLMs) are key components in NLP, but they contain strong social biases. Quantifying these biases is challenging because current methods focusing on fill-the-mask objectives are sensitive to slight changes in input. To address this, we propose a bias probing technique called LABDet, for evaluating social bias in PLMs with a robust and language-agnostic method. For nationality as a case study, we show that LABDet `surfaces' nationality bias by training a classifier on top of a frozen PLM on non-nationality sentiment detection. We find consistent patterns of nationality bias across monolingual PLMs in six languages that align with historical and political context. We also show for English BERT that bias surfaced by LABDet correlates well with bias in the pretraining data; thus, our work is one of the few studies that directly links pretraining data to PLM behavior. Finally, we verify LABDet's reliability and applicability to different templates and languages through an extensive set of robustness checks. We publicly share our code and dataset in https://github.com/akoksal/LABDet.