Gendering conversational AI in human–AI interaction: The role of interaction context and sexist ideology in shaping user evaluations
Ye Wang, Xuxu Liu, Yaling Deng, Hongjiang Xiao, Yuan Zhang · Computers in Human Behavior · 2026
As human–AI interaction becomes increasingly embedded in everyday life, users often infer social identities from AI systems, including gender. Moving beyond the conventional view of AI gender as a static, designer-assigned attribute, this study conceptualizes AI gender as a user-constructed perception shaped by contextual cues. Integrating social role theory, the stereotype content model, and ambivalent sexism theory, we conducted a 2 (AI gender cues: male vs. female) × 3 (interaction context: service-oriented, warm-authoritative, vs. serious-authoritative) between-subjects experiment with Chinese participants ( N = 833) to examine how interaction context, perceived AI gender, and sexist ideology jointly shape evaluations of AI competence, warmth, and trustworthiness. The results showed that interaction context significantly influenced gender attribution, with the serious-authoritative condition eliciting stronger gender categorization and a masculinization bias regardless of assigned gender cues. The evaluative consequences of perceived AI gender were also context-dependent: perceived femininity consistently linked with warmth, but was associated with competence and trustworthiness primarily in the serious-authoritative condition. Hostile and benevolent sexism further moderated these relationships in distinct ways. By shifting attention from designed gender to perceived gender, this study offers a more process-oriented account of AI anthropomorphism and demonstrates the value of integrating the stereotype content model with ambivalent sexism theory to explain bias in human–AI evaluation.