Validating a Generative AI competency self-efficacy scale for higher education faculty: A PLS-SEM study

The integration of Generative Artificial Intelligence (GenAI) into higher education requires sound instruments for identifying faculty development needs. This study reports the Portuguese adaptation and measurement-model evaluation of the Generative AI Competency Self-Efficacy Scale for Teachers and Researchers (GAICS-TR), a 35-item self-report instrument designed to assess academics’ perceived self-efficacy in relation to GenAI competencies. The scale comprises nine dimensions conceptually organised into four domains: Basic GenAI Understanding, Teaching with GenAI, Research with GenAI, and Professional Engagement with GenAI. The Portuguese version was developed through a pragmatic, team-based translation and adaptation procedure informed by established cross-cultural adaptation guidance. The procedure included translation of the items, review of their wording and contextual suitability, discussion of discrepancies, and consensus within the PRODIGIA research team. The items were administered on a five-point agreement scale to faculty and researchers from three partner universities, yielding 105 valid responses. Partial Least Squares Structural Equation Modelling (PLSSEM) was used to assess indicator reliability, internal consistency, convergent validity, and discriminant validity. The retained measurement model met the reported reliability and validity criteria, with the exception of the discriminant validity between GPD and GKG, which yielded a correlation of 0.923 (exceeding the standard 0.90 threshold). Item GPD2 was removed after presenting a loading of 0.482; after its removal, the professional-development dimension reached an Average Variance Extracted of 0.708 and a Cronbach’s alpha of 0.790. This item-level result should be interpreted cautiously because it may reflect wording, translation, or sample-specific effects. The adapted scale provides evidence for assessing self-reported GenAI competency self-efficacy in the participating institutions and for informing the diagnostic phase of PRODIGIA. Replication in larger, independent samples is needed.