Towards Understanding Gender-Seniority Compound Bias in Natural Language Generation
This addresses a societal bias problem for women in professional capacities, highlighting a specific, incremental gap in NLP bias evaluation.
The study tackled the problem of gender-seniority compound bias in pretrained neural generation models, finding that GPT-2 amplifies bias by considering women as junior and men as senior more often than ground truth in U.S. senatorship and professorship domains.
Women are often perceived as junior to their male counterparts, even within the same job titles. While there has been significant progress in the evaluation of gender bias in natural language processing (NLP), existing studies seldom investigate how biases toward gender groups change when compounded with other societal biases. In this work, we investigate how seniority impacts the degree of gender bias exhibited in pretrained neural generation models by introducing a novel framework for probing compound bias. We contribute a benchmark robustness-testing dataset spanning two domains, U.S. senatorship and professorship, created using a distant-supervision method. Our dataset includes human-written text with underlying ground truth and paired counterfactuals. We then examine GPT-2 perplexity and the frequency of gendered language in generated text. Our results show that GPT-2 amplifies bias by considering women as junior and men as senior more often than the ground truth in both domains. These results suggest that NLP applications built using GPT-2 may harm women in professional capacities.