Generative AI in higher education: Ethical and behavioral factors influencing students' intentions to use ChatGPT

Synthesis: Rizun, Bordean, Nikiforova, Beleiu, and Revina (2026) investigate the behavioral and ethical considerations shaping students' adoption of ChatGPT for academic tasks. By integrating the Fairness, Accountability, Transparency & Ethics in AI (FATE) framework with the Technology Acceptance Model (TAM), the Theory of Planned Behavior (TPB), and UTAUT, they developed and tested an integrated model with 17 constructs and 23 hypotheses using Partial Least Squares Structural Equation Modeling (PLS-SEM) on survey data from 344 students across four European countries (Estonia, Germany, Poland, Romania). The findings reveal that explainability and Privacy are the strongest predictors of Trust in ChatGPT, yet trust does not directly influence intention to use it — instead trust operates indirectly through perceived performance. Meanwhile, social influence from university professors and perceived behavioral control emerge as the strongest behavioral drivers of adoption, highlighting that ChatGPT must be both user-friendly and ethically sound in Higher Education.

Key Findings

Integrated adoption model. The study combines FATE, TAM, TPB, and UTAUT into a single framework, demonstrating that ChatGPT adoption in academia is shaped by technological, behavioral, and ethical factors simultaneously. Model validation used indicator loadings, Cronbach's alpha and Composite Reliability, Average Variance Extracted (AVE) for convergent validity, and Fornell–Larcker, cross-loading, and HTMT criteria for discriminant validity, with all VIF values below the multicollinearity threshold of 5.0.

Cross-cultural sample. Data were collected from 344 students across Estonia, Germany, Poland, and Romania (selected for their Northern, Western, Central, and Eastern European educational and digital contexts). Of an initial 528 responses, 177 were removed for not reporting university-related ChatGPT use and 8 for being over 40 years old. The sample was predominantly bachelor's-level (60.47%), with females more heavily represented in Romania (79.26%) and master's-level students predominating in Estonia and Romania (62.26%).

Ethics and trust interplay. Ethical concerns shape trust and perceived risk in using ChatGPT, with explainability and privacy the strongest predictors of trust. Crucially, trust does not directly predict intention to use; instead, trust significantly impacts perceived performance, suggesting students evaluate ChatGPT's reliability through its effectiveness rather than as a direct factor in adoption — a nuanced result that refines standard TAM/UTAUT theorizing about Trust and the knowledge base's Ethics concept.

Behavioral drivers of adoption. Social influence from university professors and perceived behavioral control emerged as the key behavioral drivers of ChatGPT adoption, with professor-endorsed norms more influential than peer influence alone. Perceived usefulness and attitude toward use also contribute to intention, and intention to use in turn predicts intention to purchase a subscription.

Multifaceted adoption. The findings emphasize the multifaceted nature of AI-enabled adoption, arguing that ChatGPT must be not only user-friendly but also ethically sound, with ethical risk perception moderating the usefulness–intention pathway. The authors call for institutions to design ChatGPT integration that addresses usability, behavior, and ethics together.

Implication. Institutions should design ChatGPT integration that addresses usability, behavior, and ethics together to support responsible Student Experience in Higher Education. The study provides practical insights for educators and policymakers on promoting adoption by addressing opportunities, associated risks, and typical student usage patterns, while noting limitations of self-reported cross-sectional survey data and the diversity of national digitalization contexts (Estonia's and Germany's high digitalization vs. Poland's and Romania's more gradual adoption).

What this means for practice

• Learners. Check a tool's explainability and data-handling terms before you rely on it, because those two factors—more than any other construct in this study—shaped students' Trust in ChatGPT.
• Learners. Do not read institutional or professorial endorsement as evidence that a tool works: trust influenced intention only indirectly through perceived performance, so judge the tool on how it performs on your own tasks.
• Instructors. State in the syllabus which AI uses you expect and where the limits are, since social influence from university professors was the strongest behavioral driver of adoption—outweighing peer influence.
• Administrators. Provide hands-on training that raises students' perceived behavioral control, one of the two strongest behavioral predictors of intention to use, and design integration around usability, behavior, and Ethics together rather than adoption alone.
• Administrators. Treat ethical risk perception as a live moderator of the usefulness–intention pathway rather than a compliance checkbox, and account for differences in national digitalization when standardizing tool policy across campuses.

Limitations

• The evidence is a self-reported, cross-sectional survey of 344 students measured at a single time point, so the path model establishes associations among intentions rather than causal effects or observed use.
• Of the initial 528 responses, 185 were removed—177 for not reporting university-related ChatGPT use and 8 for being over 40 years old—so about a third of the sample was excluded and the retained students are users by construction.
• The outcome is intention to use (and, downstream, intention to purchase a subscription), not actual adoption behavior.
• The four countries differ in digitalization—Estonia and Germany adopted AI in education early while Poland and Romania are progressing more gradually—so cross-national equivalence of the 17-construct model is not fully established.

Citation

Rizun, N., Bordean, O. N., Nikiforova, A., Beleiu, I. N., & Revina, A. (2026). Generative AI in higher education: Ethical and behavioral factors influencing students' intentions to use ChatGPT. Computers and Education Open, 100336. https://doi.org/10.1016/j.caeo.2026.100336