Role Incongruity and Information Accuracy in AI-Mediated Communication: An Integrated Approach to AI Evaluation and Perception of Gender Stereotype

Jiwon Kim (Ph.D. student) and Glenna Read, “Role Incongruity and Information Accuracy in AI-Mediated Communication: An Integrated Approach to AI Evaluation and Perception of Gender Stereotype,” paper accepted to the International Communication Association Conference, 4-8 June 2026, Cape Town, South Africa.  Abstract: This study examines how gender stereotypes and information accuracy in AI services influence users’ evaluations of AI systems and their perceptions of gender-stereotypical traits, by integrating Role Congruity Theory (RCT), Expectancy Violation Theory (EVT), and the Limited Capacity Model of Motivated Mediated Message Processing (LC4MP). Although AI-provided information is often perceived as reliable, users may evaluate the same information differently depending on the social roles of the AI developers. In particular, whether the information provided is accurate can affect expectancy violations, which in turn lead to emotional responses (valence and arousal). The study posits a main effect of information accuracy on valence and an interaction between role congruity and information accuracy on arousal. These emotional responses can act as catalysts for cognitive evaluations of the AI service and perceptions of the agent’s gender-stereotypical characteristics. A 2 (Gender: Male vs. Female) × 2 (Topic: Masculine vs. Feminine) × 2 (Accuracy: Accurate vs. Inaccurate) between-subjects experimental design (N = 400) will test these mechanisms. The study underscores the importance of informational accuracy in the management of AI systems.