Language Models Generate Widespread Intersectional Biases in Narratives of Learning, Labor, and Love

Abstract

The rapid deployment of generative language models (LMs) has raised concerns about social biases affecting the well-being of diverse consumers. The extant literature on generative LMs has primarily examined bias via explicit identity prompting. However, prior research on bias in language-based technology platforms has shown that discrimination can occur even when identity terms are not specified explicitly. Here, we advance studies of generative LM bias by considering a broader set of natural use cases via open-ended prompting, what we refer to as a laissez-faire environment. In this setting, we find that across 500,000 observations, generated outputs from the base models of five publicly available LMs (ChatGPT3.5, ChatGPT4, Claude2.0, Llama2, and PaLM2) are hundreds to thousands of times more likely to omit or subordinate characters with minoritized race, gender, and/or sexual orientation identities. We also document patterns of stereotyping across LM-generated outputs with the potential to disproportionately affect minoritized individuals. Our findings highlight the urgent need for regulations to ensure responsible innovation while protecting consumers from potential harms caused by language models as well as further investments in critical artificial intelligence education programs tailored towards empowering diverse consumers.

Ähnliche Arbeiten