Artificial intelligence literacy and technology acceptance model-related constructs among first-year nursing students: a multi-institutional cross-sectional network analysis

Abstract

Artificial intelligence literacy is increasingly important in nursing education, but it remains unclear how its dimensions relate to three constructs related to the Technology Acceptance Model (TAM): perceived usefulness (PU), perceived ease of use (PEU) and attitudes toward artificial intelligence (ATU). This study aimed to examine these relationships and to identify factors associated with artificial intelligence literacy. A multi-institutional cross-sectional study surveyed first-year nursing students at six undergraduate institutions in one province of China from 20 April to 1 May 2026. Of 375 questionnaire submissions, 362 met the quality-control criteria and were analysed (analytic retention rate: 96.5%). Separate item-level confirmatory factor analyses evaluated the four-factor literacy structure and the distinctiveness of PU, PEU and ATU. Group comparisons and multivariable regression with heteroscedasticity-robust inference examined correlates of overall artificial intelligence literacy; Pearson correlations described pairwise associations. A regularised partial correlation network mapped conditional relationships, centrality and bridge connectivity among seven dimension-mean nodes: knowledge, skills, values, ethics, PU, PEU and ATU. Between-institution differences, clustering and robustness of pooled findings were assessed. In the fully adjusted multivariable model, more frequent artificial intelligence use, prior artificial intelligence education and interest in artificial intelligence were positively associated with the mean artificial intelligence literacy score. The four-factor model showed good fit and strong item loadings; skills and values were highly correlated. Measurement analyses supported PU, PEU and ATU as correlated but distinguishable constructs. Of 21 potential network edges, 17 were non-zero. Skills had the highest expected influence and PU the highest strength. One-step bridge rankings were partition-dependent: values ranked first under the primary two-community partition and PU under the alternative four-community partition; PEU ranked second in both. Bootstrap diagnostics supported strong stability of strength and expected influence under case removal (CS(cor = 0.7) = 0.751 for both). Intraclass correlation estimates were near zero; institution-adjusted analyses preserved the main regression and network findings. Artificial intelligence use frequency, prior artificial intelligence education and interest were positively associated with artificial intelligence literacy. Skills, perceived usefulness, values and perceived ease of use may inform hypotheses for prospective evaluation of artificial intelligence education in undergraduate nursing programmes.