Anthropomorphic AI Chatbots Reinforce the Credibility of a Sexist Stereotype Among Conservative Women

AI chatbots are increasingly anthropomorphic, and anthropomorphism is known to increase trust among its users. This increased trust raises the concern that anthropomorphic chatbots can reinforce stereotypical beliefs among individuals from predisposed groups. Consistent with this hypothesis, we report results from two preregistered experiments on U.S. adults (N = 2,065) showing that politically conservative women believe an unsupported sexist stereotype in a chatbot's response to be more accurate when the chatbot has lifelike features. Neither liberal women nor conservative men show this effect. This effect is mediated by perceived anthropomorphism and trustworthiness. We speculate that conservatives’ known propensity toward sexist beliefs and social identity threat may converge in making conservative women more susceptible to sexist stereotypes from anthropomorphic AI chatbots. Future research should test whether this effect of AI’s anthropomorphism occurs for other identity groups with stereotypes which they are predisposed to believe.