Using Natural Sentences for Understanding Biases in Language Models
This addresses the issue of biased evaluations in AI fairness research, offering a more systematic approach for researchers and practitioners, though it is incremental as it builds on existing bias evaluation methods.
The paper tackled the problem of evaluating gender-occupation biases in language models by creating a prompt dataset from real-world natural sentences in Wikipedia, finding that bias evaluations are highly sensitive to template-based prompt designs and proposing natural sentence prompts to reduce design-induced bias.
Evaluation of biases in language models is often limited to synthetically generated datasets. This dependence traces back to the need for a prompt-style dataset to trigger specific behaviors of language models. In this paper, we address this gap by creating a prompt dataset with respect to occupations collected from real-world natural sentences present in Wikipedia. We aim to understand the differences between using template-based prompts and natural sentence prompts when studying gender-occupation biases in language models. We find bias evaluations are very sensitive to the design choices of template prompts, and we propose using natural sentence prompts for systematic evaluations to step away from design choices that could introduce bias in the observations.