JMIR Publications

ABSTRACT

Background:

Generative artificial intelligence is now embedded in health sciences education, but frequent use is not equivalent to maladaptive reliance. Evidence remains limited on whether self-reported artificial intelligence overdependence is associated with students' critical-thinking disposition or social well-being, particularly in Middle Eastern health professions programs.

Objective:

This study examined associations among self-reported artificial intelligence overdependence, critical-thinking disposition, and social well-being in health sciences students in the United Arab Emirates.

Methods:

A cross-sectional web-based survey was distributed by institutional email to all 222 students in the eligible sampling frame during the 2025-2026 academic year. The Google Forms questionnaire included 12 artificial intelligence overdependence items, 10 critical-thinking disposition items, and 15 social well-being items, each rated from 1 (strongly disagree) to 5 (strongly agree). All 200 submitted questionnaires were complete and unique (response rate 90.1%). Internal consistency, Pearson correlations with Fisher 95% CIs, Benjamini-Hochberg adjusted P values, Spearman sensitivity correlations, and ordinary least squares regression with HC3 heteroskedasticity-robust inference were used. Because the measures were study-developed and administered concurrently, analyses were interpreted as associations among provisional summed scores rather than causal or diagnostic effects.

Results:

Among 200 respondents, 147 (73.5%) were women, the median age was 20 (IQR 19-22) years, and 195 (97.5%) were undergraduates. Mean scores were 43.9 (SD 6.3) for artificial intelligence overdependence, 27.3 (SD 4.1) for critical-thinking disposition, and 44.2 (SD 7.2) for social well-being. Artificial intelligence overdependence had a weak inverse Pearson association with critical-thinking disposition (r=-0.15, 95% CI -0.28 to -0.01; P=.03; adjusted P=.047), with a stronger Spearman estimate (rho=-0.33; P<.001). It was not associated with social well-being (r=-0.004, 95% CI -0.14 to 0.13; P=.95). Critical-thinking disposition was positively associated with social well-being, although Pearson (r=0.73, 95% CI 0.66-0.79; P<.001) and Spearman (rho=0.24; P<.001) magnitudes differed. In the HC3-robust model, critical-thinking disposition remained positively associated with social well-being (b=1.31, 95% CI 1.07-1.56; P<.001), whereas the adjusted artificial intelligence overdependence coefficient was not statistically significant (b=0.13, 95% CI -0.01 to 0.26; P=.06).

Conclusions:

Self-reported artificial intelligence overdependence was weakly associated with lower critical-thinking disposition but was not associated with social well-being. Critical-thinking disposition was the more consistent correlate of social well-being. The divergence between Pearson and rank-based estimates, the cross-sectional design, and the provisional measures require cautious interpretation. Longitudinal, multi-institutional studies using validated measures should test whether verification-focused artificial intelligence education preserves cognitive autonomy and student well-being.