Explaining ChatGPT Adoption in Higher Education: Insights for AI Literacy, Educational Practice, and Responsible AI
TL;DR
This study proposes an explainable artificial intelligence (XAI) framework that integrates the Unified Theory of Acceptance and Use of Technology (UTAUT2), machine learning, and explainability techniques to examine students’ intentions to use ChatGPT in academic contexts, revealing that Effort Expectancy acts as a necessary condition for high adoption levels despite its limited direct effect.
Abstract
The rapid adoption of generative artificial intelligence tools, particularly ChatGPT, is transforming teaching and learning in higher education. This study proposes an explainable artificial intelligence (XAI) framework that integrates the Unified Theory of Acceptance and Use of Technology (UTAUT2), machine learning, and explainability techniques to examine students’ intentions to use ChatGPT in academic contexts. Survey data were analyzed using Ordinary Least Squares regression, Random Forest, SHAP, Necessary Condition Analysis (NCA), Importance–Performance Map Analysis (IPMA), and K-Means clustering. The results indicate that Habit, Performance Expectancy, Hedonic Motivation, Social Influence, and Facilitating Conditions significantly influence behavioral intention, explaining 67.6% of the variance. Habit emerged as the strongest predictor, whereas Price Value had negligible influence. XAI analyses revealed that Effort Expectancy acts as a necessary condition for high adoption levels despite its limited direct effect. Four distinct student profiles were identified, highlighting heterogeneous patterns of AI integration and informing strategies for responsible and effective educational adoption. These findings provide evidence-based guidance for integrating AI literacy, responsible AI practices, and pedagogically meaningful ChatGPT use in higher education curricula.