Digital Media Literacy in the Age of AI: Exploring the Comfort–Competence Gap Among University Students in the United Arab Emirates

Abstract

Artificial intelligence increasingly mediates how citizens access information, making digital media literacy a condition for informed participation. This exploratory study conceptualizes a comfort–competence gap, whereby students report confidence and ease in using AI tools but a weaker capacity to evaluate what those tools produce, and examines it among 300 university students in the United Arab Emirates. The gap is advanced provisionally, and all measures are self-reported perceptions rather than demonstrated performance. Cross-sectional questionnaire data (twelve four-point ordinal items) were analyzed with ordinal-appropriate methods. Students generally reported greater confidence and comfort in using AI than capacity to evaluate it critically. Every operational item mean (2.51–2.55) exceeded every critical item mean (2.39–2.44), and the within-person contrast between the composite indices was small but statistically significant (difference = 0.12, p = 0.027, dz = 0.13), a pattern shown by 44.7% of individuals. No item-level contrast reached significance and 36.0% showed the reverse pattern, so the difference is read as an exploratory tendency rather than evidence of a general gap. The contrast is also sensitive to the composition of the operational index: with comfort removed, it retains its direction, but not conventional significance (p = 0.056). Digital confidence was the only correlate of comfort in AI-mediated communication (ρ = 0.131; OR = 1.29), a weak association in a model with minimal explanatory power (McFadden pseudo-R2 = 0.012); communication competence and demographic characteristics showed no detectable associations, and parallel analysis retained no factors. Interpreted cautiously, the pattern bears on responsible AI use, since comfort unmatched by evaluative capacity is a weak basis for human oversight. The study’s clearest contribution is methodological: single-item self-report batteries can profile levels and detect aggregate differences, but they cannot show whether AI literacy is one capacity or several, cannot separate real capability from what respondents are willing to claim, and cannot support strong conclusions from null results. It therefore specifies the multi-item and performance-based measurement agenda a confirmatory test would require.

The authors' abstract, as published at the source. Societies, 2026 · DOI ↗