Research report: AI-use orientations and self-perceived AI competence among medical students: a cross-sectional study, Brazil

In a cross-sectional survey of medical students in Brazil, distinct orientations toward artificial intelligence were associated with different profiles of self-perceived competence, with practical and ethical emphases most closely aligned to different skills. Among 672 participants who could be assigned a single predominant AI-use orientation, 44.0% were classified as Ethical AI users, 36.8% as Supported AI users, 15.6% as Enabled AI users and 3.6% as Targeted AI users. Enabled AI orientation had the strongest reported correlation with technical-operational competence, whereas Ethical AI orientation had the strongest correlation with critical-ethical competence.

The study, based on electronic survey data collected from August to October 2025 in a predominantly Ceará-based sample, identifies associations rather than proving that particular ways of using AI produce better competence or educational outcomes. The findings may nevertheless offer a detailed picture of how medical students relate to AI and identify areas for further curriculum evaluation. The researchers used the Artificial Intelligence Education Use instrument, or AIEd-U, to assess four domains of AI engagement: Targeted, Supported, Enabled and Ethical AI orientations.

The study initially included 844 medical-student respondents, but 172, or 20.4% of the full sample, had tied maximum scores across domains. Because the researchers assigned participants to a predominant orientation only when one domain had a uniquely highest score, those 172 respondents were excluded from the orientation-specific analyses, leaving 672 participants. Of the analyzed group, 94.5% were from Ceará, so the results primarily characterize students in that northeastern Brazilian state and should not be generalized automatically to all Brazilian medical students or to students elsewhere.

The AIEd-U measured students’ predominant patterns of AI engagement rather than the frequency of use, while a separate study-specific questionnaire used nine items to assess self-perceived technical-operational and critical-ethical AI competence. Because competence was self-reported, the findings do not demonstrate that students could independently apply AI tools or make sound ethical judgments in practice. Correlation estimates showed that the orientation most closely associated with practical enablement was not the one most closely associated with every dimension of competence.

Enabled AI orientation had the strongest correlation with Domain 1, the technical-operational measure, at Spearman’s rho of 0.487 and a bootstrap 95% confidence interval of 0.423 to 0.545. Ethical AI orientation had the strongest correlation with Domain 2, the critical-ethical measure, at rho of 0.437, with a 95% confidence interval of 0.372 to 0.500. Targeted AI orientation was inversely associated with critical-ethical competence, with rho of −0.200 and a 95% confidence interval of −0.270 to −0.123.

These figures represent monotonic associations in the survey data: students with higher orientation scores tended to report different levels of self-perceived competence. They do not show that the orientations caused the reported competence, nor do they predict future performance, clinical safety or the quality of decisions made with AI. Competence scores also differed across the four predominant-orientation groups, but the researchers characterized the overall effects as small.

For technical-operational competence, the Kruskal–Wallis test yielded H = 36.315, P < .001, ordinal eta-squared of 0.050 and a 95% confidence interval of 0.024 to 0.091. For critical-ethical competence, the test yielded H = 26.102, P < .001, ordinal eta-squared of 0.035 and a 95% confidence interval of 0.015 to 0.063. The analysis used Dunn–Bonferroni comparisons for pairwise differences and Cliff’s delta for effect-size estimation.

Although the statistical tests indicated differences among groups, the small eta-squared values mean that predominant orientation accounted for only a modest portion of the observed variation in self-perceived competence. The reported association between Enabled AI orientation and technical-operational competence therefore cannot be interpreted as proof that one orientation generally produces greater competence across all domains, nor do the group differences establish which orientation is preferable for medical education. Overall, the study portrays AI engagement among the participating medical students as varied rather than uniform.

Ethical and Supported AI orientations together accounted for 80.8% of the analyzed sample, while Enabled and Targeted orientations represented smaller shares. The data support the conclusion that distinct patterns of AI engagement were associated with different self-perceived competence profiles, with no single orientation dominating both the technical and critical-ethical domains. The authors call for further evaluation of curricula that combine practical AI use with critical appraisal, ethical awareness and learner self-regulation, but no such curriculum was tested in this research.

The results therefore provide a baseline for future research in medical education rather than evidence of causal learning gains, clinical benefit or curriculum effectiveness.

Subject of Research: Associations between AI-use orientations and self-perceived competence among medical students in Ceará, Brazil

Article Title: AI-use orientations and self-perceived AI competence among medical students: a cross-sectional study, Brazil

Article References: Kubrusly, M., Gonzaga, M. L., Moreira, D. P., de Pádua Rebouças Crisóstomo de And, F., do Carmo, J. M., Pontes, M. C. B., de Carvalho Rocha, J. L., & Rocha, H. A. L. (2026). AI-use orientations and self-perceived AI competence among medical students: a cross-sectional study, Brazil. BMC Medical Education. https://doi.org/10.1186/s12909-026-10329-0

Image Credits: AI Generated

DOI: 10.1186/s12909-026-10329-0

Keywords: medical education, artificial intelligence, medical students, AI competence, critical ethics, cross-sectional study, Brazil, Ceará, self-assessment, curriculum research, AI-use, orientations