The gap the scale fills is more than academic. Nursing education shapes the workforce that delivers the majority of direct patient care worldwide, and as AI-driven decision support, predictive analytics, and automated documentation become embedded in clinical practice, graduates must arrive fluent in these systems. That fluency depends on educators who can model, critique, and teach AI-related skills. Without a validated measure of faculty competency, curriculum committees have been designing AI integration strategies essentially blind, unable to benchmark where their educators stand or evaluate whether professional development programs actually work. The AI-CSNE was built to give institutions that benchmark, and its development followed the kind of methodological discipline that measurement scientists insist on but that few new instruments actually receive.

The researchers grounded the scale in the UNESCO Artificial Intelligence Competency Framework for Teachers, an international blueprint that outlines what educators across disciplines need to know about AI, supplemented by a review of the relevant literature. Content validity was established through expert evaluation and pilot testing before the instrument faced its real trial: a large cross-sectional survey of 470 nurse educators drawn from multiple nursing faculties in Egypt, conducted between mid-October 2025 and February 2026. Ethical approval came from the Scientific Research Ethics Committee of Tanta University, and participation was voluntary, anonymous, and electronically consented, with no personal identifiers collected at any point.

What makes the study methodologically interesting is how the authors handled validation. Rather than analyzing all 470 responses in a single pass, they randomly split the sample into two independent subsamples of 235 educators each. The first half was used for exploratory factor analysis, a statistical technique that lets the underlying structure of a questionnaire emerge from the data rather than being imposed by the researchers. The second, untouched half was reserved for confirmatory factor analysis, which tests whether the structure discovered in the first sample holds up in fresh data. This two-step design is considered the gold standard in scale development because it guards against overfitting, the statistical equivalent of memorizing answers rather than learning the subject.

The exploratory analysis revealed a clean three-factor structure. The first domain, Foundational AI Knowledge and Instructional Application, captures the technical and pedagogical core: understanding what AI is, how it works, and how to weave it into teaching. The second, Ethical and Human-Centered AI Use, addresses the values dimension, including responsible deployment, awareness of bias, and keeping human judgment at the center of care education. The third, AI-Supported Professional Growth, reflects educators’ use of AI for their own continuing development. Factor loadings, which indicate how strongly each item relates to its domain, ranged from 0.661 to 0.822, and together the three factors explained 56.10 percent of the total variance in responses, a respectable share for a multidimensional instrument.

Reliability, the consistency with which the scale measures what it measures, was assessed using Cronbach’s alpha. The Foundational AI Knowledge and Instructional Application domain scored 0.842, and Ethical and Human-Centered AI Use scored 0.810, both comfortably within the range conventionally regarded as satisfactory. The AI-Supported Professional Growth domain came in at 0.697, slightly below the traditional 0.70 threshold but close enough that the authors judged it acceptable for a newly developed scale, with comparable estimates appearing in the confirmatory subsample. These figures suggest the scale produces stable, internally coherent measurements across its domains, though the third subscale may benefit from refinement in future iterations.

The confirmatory factor analysis delivered the study’s most striking numbers. The three-factor model achieved a Comparative Fit Index of 0.989 and a Tucker-Lewis Index of 0.986, both far above the 0.95 benchmark typically considered evidence of excellent fit. The Root Mean Square Error of Approximation was 0.023 and the Standardized Root Mean Square Residual was 0.046, both well under the cutoffs that signal good model fit. In plain terms, the structure discovered in the first half of the sample reproduced almost perfectly in the second half, giving strong initial evidence that the AI-CSNE genuinely measures three coherent dimensions of perceived AI competency rather than statistical noise. Within the confirmatory sample, the Foundational Knowledge domain showed the highest standardized loadings, between 0.626 and 0.785, while the Ethics domain loaded somewhat lower, from 0.570 to 0.734, still within acceptable bounds.

One finding carries particular weight for anyone planning to use the instrument: the correlations between the three domains were near zero. This means the subscales are empirically distinct, tapping genuinely separate facets of competency rather than slightly different expressions of a single underlying ability. The practical consequence is that scores should be interpreted domain by domain rather than summed into one overall AI competency score. An educator might be technically knowledgeable yet ethically cautious, or professionally curious yet lacking instructional skills, and the scale’s structure is designed to expose exactly those profiles. The authors also assessed convergent, criterion, and known-groups validity, finding no significant differences across demographic subgroups, which they interpret cautiously as a sign that further validation work remains necessary.

Perhaps the most important caveat concerns what the scale does not measure. The AI-CSNE captures perceived competency, what educators believe they can do, not objectively demonstrated skill. Self-perception and actual ability can diverge, sometimes dramatically, in any domain of expertise. The authors are explicit that the instrument is therefore best suited for exploratory group-level needs assessment, research comparisons, program evaluation, and self-reflective faculty development, rather than for individual appraisal, certification, or high-stakes evaluation of any kind. Used appropriately, however, it offers nursing schools a practical starting point: a way to map where their faculty stand on AI knowledge, ethics, and professional growth, and to design training that targets real weaknesses. As AI continues its advance into clinical education, instruments like this one, built on transparent psychometric foundations and tested against independent data, will be essential for ensuring that the humans teaching the caregivers are not left behind by the machines now entering the classroom.

Subject of Research: Development and psychometric validation of a three-domain scale measuring perceived artificial intelligence competency among nurse educators

Article Title: Development and structural validation of the Artificial Intelligence Competency Scale for Nurse Educators (AI-CSNE): a three-domain measure of perceived AI competency

Article References: Mohamed, M. G., Mohamud, Z. A., & Sorour, M. S. (2026). Development and structural validation of the Artificial Intelligence Competency Scale for Nurse Educators (AI-CSNE): a three-domain measure of perceived AI competency. BMC Nursing. https://doi.org/10.1186/s12912-026-05453-7

Image Credits: AI Generated

DOI: 10.1186/s12912-026-05453-7

Keywords: artificial intelligence, nurse educators, psychometric validation, scale development, nursing education, factor analysis, competency assessment, UNESCO AI Competency Framework, confirmatory factor analysis, faculty development, health professions education, self-perceived competency