''Fifty Shades of Bias'': Normative Ratings of Gender Bias in GPT Generated English Text
This work addresses the need for nuanced bias assessment in LLMs for AI ethics and fairness, though it is incremental as it builds on prior binary classification methods.
The study tackled the problem of measuring gender bias in GPT-generated text by creating the first dataset with normative bias ratings, revealing that identity-attack themes are most closely linked to gender bias and evaluating existing automated models on this dataset.
Language serves as a powerful tool for the manifestation of societal belief systems. In doing so, it also perpetuates the prevalent biases in our society. Gender bias is one of the most pervasive biases in our society and is seen in online and offline discourses. With LLMs increasingly gaining human-like fluency in text generation, gaining a nuanced understanding of the biases these systems can generate is imperative. Prior work often treats gender bias as a binary classification task. However, acknowledging that bias must be perceived at a relative scale; we investigate the generation and consequent receptivity of manual annotators to bias of varying degrees. Specifically, we create the first dataset of GPT-generated English text with normative ratings of gender bias. Ratings were obtained using Best--Worst Scaling -- an efficient comparative annotation framework. Next, we systematically analyze the variation of themes of gender biases in the observed ranking and show that identity-attack is most closely related to gender bias. Finally, we show the performance of existing automated models trained on related concepts on our dataset.