Unmasking Contextual Stereotypes: Measuring and Mitigating BERT's Gender Bias
This work addresses the critical problem of gender bias in large language models like BERT, which can perpetuate societal stereotypes, particularly for NLP system developers and users concerned with fairness and ethical AI.
This paper investigates gender bias in BERT by measuring associations between gendered words and professions in English and German, comparing them to real-world workforce statistics. They mitigate bias by fine-tuning BERT on the GAP corpus using Counterfactual Data Substitution, finding their measurement method suitable for English but not for morphologically rich languages like German.
Contextualized word embeddings have been replacing standard embeddings as the representational knowledge source of choice in NLP systems. Since a variety of biases have previously been found in standard word embeddings, it is crucial to assess biases encoded in their replacements as well. Focusing on BERT (Devlin et al., 2018), we measure gender bias by studying associations between gender-denoting target words and names of professions in English and German, comparing the findings with real-world workforce statistics. We mitigate bias by fine-tuning BERT on the GAP corpus (Webster et al., 2018), after applying Counterfactual Data Substitution (CDS) (Maudslay et al., 2019). We show that our method of measuring bias is appropriate for languages such as English, but not for languages with a rich morphology and gender-marking, such as German. Our results highlight the importance of investigating bias and mitigation techniques cross-linguistically, especially in view of the current emphasis on large-scale, multilingual language models.