Using the Reasoned Action Approach to Predict Students’ Misuse of Generative Artificial Intelligence in the Production of University Assessments

We tested whether the Reasoned Action Approach (RAA) model could predict both students’ intentions and the frequency with which they had previously misused generative artificial intelligence (GenAI) for producing assessment content (behaviour). We also tested the extent to which the additional constructs of moral norm and anticipated regret improved the predictive validity of the RAA in respect of this behaviour. The participants (N = 369 psychology undergraduates) completed an online questionnaire measuring the RAA and additional constructs. Hierarchical linear regressions showed that the RAA accounted for 74% of the variance in students’ intentions and 65% of the variance in behaviour. Moral norm and anticipated regret led to a significant (1% point) increment to explained variance in the prediction of intentions but not behaviour. The independent predictors of intentions were instrumental attitude, injunctive norm and anticipated regret. Intention was the sole predictor of behaviour. The results demonstrate, for the first time, that the RAA is a useful model for predicting university students’ misuse of GenAI in the production of assessment content, and that anticipated regret constitutes a useful addition to the model in this context. Practical implications for decreasing GenAI misuse are discussed.