Understanding Students’ Generative AI Use in Higher Education: A SEM-ANN Analysis of AI Attributes, Behavioral Reasons, and Perceived Learning Performance

This is the authors' abstract. We don't add key points for this paper.

Generative Artificial Intelligence (AI) tools can support personalized educational experiences. However, limited understanding remains of how AI-specific human-like attributes and behavioral reasoning mechanisms jointly relate to students’ attitudes, self-reported Generative AI use, and perceived learning performance. To address this gap, this research integrates Behavioral Reasoning Theory (BRT) with key AI characteristics (perceived anthropomorphism, animacy, and intelligence) to examine students’ attitudes towards Generative AI and their association with perceived learning performance. Data were collected from 395 students who reported using Generative AI tools for academic activities. The study examined Generative AI broadly rather than a specific platform. Data were analyzed using a hybrid Structural Equation Modeling-Artificial Neural Network (SEM-ANN) approach. The results indicated that perceived anthropomorphism, animacy, intelligence, and reasons for adoption were positively associated with students’ attitudes, whereas reasons against adoption were not significantly associated with attitude. Self-reported Generative AI use was positively associated with students’ perceived learning performance. The ANN analysis ranked reasons for adoption as the most important predictor of students’ attitudes, with a normalized importance of 97.78%. The model explained 64.7% of the variance in students’ attitudes, 48% in self-reported Generative AI use, and 24.9% in perceived learning performance. This study contributes theoretically by extending conventional technology adoption research through integrating behavioral reasoning and AI-specific perception factors within the educational context.