Feeling or thinking? User aversion to AI customer service in online education
Related Experiment Videos
1School of Economics and Management, Dongguan University of Technology, Dongguan, China.
Introduction:
AI customer service is increasingly integrated into online education, yet users may not always respond favorably to AI-assisted service interactions. Drawing on robotic role theory, this study examines how thinking and feeling AI customer service influence user aversion in online educational service encounters.
Methods:
Three experiments were conducted to examine the effect of AI customer service type on user aversion (Study 1, N = 120), the mediating role of perceived intelligence (Study 2, N = 120), and the moderating role of AI identity disclosure timing (Study 3, N = 120). Across the three between-subjects experiments, thinking AI was operationalized through analytical, information-processing, and task-oriented communication, whereas feeling AI emphasized emotional recognition and relational support.
Results:
The results show that feeling AI generates greater user aversion than thinking AI. Perceived intelligence mediated this effect, such that feeling AI is perceived as less intelligent, which in turn increases user aversion. Further, AI identity disclosure timing moderates the indirect effect through perceived intelligence. Early disclosure of AI identity attenuates the negative effect of feeling AI.
Discussion:
These findings demonstrate that the effectiveness of different AI customer service types depends on the service context and highlight the importance of aligning AI communication with users' task-oriented service needs.
