Academic Integrity and Generative Artificial Intelligence in Higher Education: Psychometric Validation of a Short Questionnaire
TL;DR
The final model showed adequate fit, high internal consistency, convergent validity, acceptable but borderline discriminant validity, and preliminary evidence of measurement invariance across gender, although the latter should be interpreted with caution because the configuration model did not show optimal absolute fit.
Abstract
Generative artificial intelligence has reshaped writing, learning, and assessment practices in higher education, posing new challenges for academic integrity. In this context, the objective of this study was to analyze the psychometric properties of an academic integrity questionnaire among college students. An instrumental study with a quantitative approach and cross-sectional design was conducted with a sample of 419 higher education students. The original instrument included 17 items distributed across three dimensions: ethical use of AI, awareness of risks of misuse, and support for academic writing. Descriptive analyses, reliability estimates, exploratory factor analysis, and confirmatory factor analysis were performed using a robust approach based on DWLS, considering the ordinal nature of the items. The results supported a refined version of 15 items organized into two correlated factors: normative integrity in AI use and support for academic writing. The final model showed adequate fit, high internal consistency, convergent validity, acceptable but borderline discriminant validity, and preliminary evidence of measurement invariance across gender, although the latter should be interpreted with caution because the configural model did not show optimal absolute fit.