Abstract

There is recent evidence that people process and respond to the sex type of gendered artificial intelligence (AI). This raises the possibility that people can act sexist toward AI in task delegation. This is important to consider given that AI might be elevated from tools to teammates. Here, we contribute to the emerging discourse on this possibility by first, theorizing that humans might develop sexism toward AI using both (1) traditional gender role schemas and (2) homophily considerations (a preference for same-sex collaboration). We further posit that sexism toward AI might be triggered by both the explicit gender cues some AI tools provide, through anthropomorphism, and implicit cues, through responses that may be seen as more feminine or masculine. Last, we posit that this sexism has important outcomes in form of judgment of the AI’s abilities to execute traditionally feminine or masculine tasks, and task delegation to the AI. We test these ideas in two pilot studies, four main studies, and one follow-up study (total n = 1,879). Results revealed that humans act sexist toward gendered AI in multiple contexts. This sexism has at least two sources: homophily and gender role schemas, it spans across traditionally feminine, traditionally masculine, and gender-neutral tasks, and it cannot always be easily alleviated by eliminating typical sex type cues, such as faces, or with performance indicators, such as star ratings. These findings improve the understanding of appraisal mechanisms in human-AI task delegations and the future of working with AI.

DOI

10.17705/1jais.01013

Download

Share

COinS