Evaluating generative AI teaching assistants in simulated learning environments: how instructor type and support type affect students’ perceptions

Generative AI Teaching Assistants (GAITAs) offer transformative potential for education, yet their effective integration remains under-explored. This study employed a Wizard-of-Oz-based online between-subjects experiment (N = 136) to investigate how instructor type (human vs. AI), support type (instrumental vs. both instrumental and emotional), and course type (social vs. natural sciences) influence student perceptions. Results from a three-factor analysis of variance revealed that GAITAs elicited significantly higher perceived adaptability and more positive attitudes when assisting human instructors rather than AI counterparts. In social science contexts, human-GAITA collaboration further enhanced trust and hedonic motivation. Additionally, emotional support significantly boosted hedonic motivation, perceived adaptability, and attitudes towards GAITAs in social science courses, whereas no such effect occurred in natural science courses. Overall, this study highlights the importance of collaboration between human instructors and GAITAs, and provides useful insights into the application of emotional support provided by GAITAs.