Abstract How should AI-generated speech balance epistemic aims, such as precision and accuracy, with ethical and social considerations? This paper examines a subtle yet consequential aspect of LLM-driven communication: the use of generic generalizations that convey information about social groups (e.g., “immigrants work low-wage jobs”). While central to human epistemic and pedagogical practices, generics are theorized to reinforce stereotypes, essentialism, and injustice. Using ChatGPT-3.5 as a case study, I uncover tendencies for AI chatbots to inconsistently hedge and refuse generics, including those that reflect well-documented social structural patterns, such as “women are more likely to get attacked while walking alone at night.” These tendencies are not only ethically dubious but also epistemically troubling, since they may reduce the accuracy and informativeness of outputs and impede the communication of important social truths. In response, I highlight the complexity of improving chatbots’ social generic use by evaluating four approaches and show that each approach carries significant tradeoffs. I suggest chatbots should be permitted to use social generics when such generics are paired with clarifications regarding their scope and fungibility. I then offer three complementary strategies to operationalize an Interdisciplinary Expert-Driven Counterfactual Dialogue proposal to improve the accuracy and social responsibility of descriptions of groups and patterns.
Tiffany A Zhu (Fri,) studied this question.