Understanding Trust and Willingness to Use GenAI Tools in Higher Education: A SEM-ANN Approach Based on the S-O-R Framework

Student trust plays a pivotal role in shaping the future integration of artificial intelligence (AI) in higher education. This study investigates how AI Facilitating Conditions (FCs), Performance Expectancy (PE), and task type influence students’ System-like Trust (AST) and Human-like Trust (AHT) in AI and further examines the mediating role of human-like trust in fostering the willingness to continue AI-assisted learning. Drawing on valid data collected from 466 Chinese university students, we employed partial least squares structural equation modeling (PLS-SEM) in combination with artificial neural networks (ANN) to test the hypothesized relationships, mediating mechanisms and the relative importance of influencing factors. The findings indicate that AI facilitating conditions significantly enhance both system-like trust and usage intention; performance expectancy exerts a positive effect on both forms of trust, with particularly strong effects observed in subjective tasks. Moreover, system-like trust positively promotes human-like trust, and together, these dimensions jointly strengthen students’ intention to engage in AI-assisted learning. Results from the ANN analysis further highlight that performance expectancy, system-like trust, and facilitating conditions are the primary determinants of system-like trust, human-like trust, and usage intention, respectively. This study extends the application of interpersonal trust theory to the AI domain and offers theoretical insights for fostering more positive and effective patterns of AI adoption in higher education.