A Faculty of Informatics at Eötvös Loránd University (ELTE) has published a study examining which factors influence university students to continue using ChatGPT for study purposes. The paper appeared in the Elsevier journal Computers and Education: Artificial Intelligence (Q1/D1).
Research lead and method
The study was conducted by Chaman Verma, Assistant Professor at the Department of Media and Educational Informatics, Faculty of Informatics, ELTE, within the EKÖP-24 University Excellence Scholarship Program. It analysed questionnaire responses from 166 university students and used interpretable machine learning techniques to identify the factors most strongly associated with continued use of ChatGPT in higher education.
Key findings
The analysis found that the following factors were the strongest predictors of whether students would keep using ChatGPT:
- perceived learning support (helping to understand complex concepts and complete tasks),
- accessibility,
- study engagement,
- usability,
- perceived reliability of the tool.
Students were more likely to integrate ChatGPT into their learning processes when they perceived it as genuinely useful — for example, when it aided comprehension, supported assignment completion, and contributed to more efficient studying.
Trust and concerns
The study also highlights that acceptance of generative AI in higher education depends heavily on answer accuracy and the tool's reliability. Concerns about academic integrity and the responsible use of AI in the university context continue to shape students' attitudes toward ChatGPT.
Why this matters
The research adds empirical evidence to the body of work on AI in education and offers evidence-based considerations for educators, higher education institutions, and policymakers. Its findings can inform efforts to ensure generative AI tools are used responsibly, consciously, and effectively to support teaching and learning in higher education.
Additional note
The original study is published in English in the Computers and Education: Artificial Intelligence journal.
