Answers are grounded in this paper and cite their sources.
Background Generative artificial intelligence (GenAI) can support learning and emotion regulation, yet frequent or coping-oriented use does not necessarily constitute difficult-to-regulate dependency. It remains unclear how academic rumination and affective coping are positioned relative to dependency and whether dependency is differentially associated with perceived costs to knowledge internalization and independent reasoning. This study examined these distinctions within a single educational psychology model. Methods A cross-sectional survey included 611 current GenAI users recruited through university settings in Tianjin, China (91.7% first-year students). Dependency was a continuous, nonclinical tendency. Ordinal CFA and WLSMV SEM evaluated a prespecified partial association model. Covariate adjustment, four alternative measurement specifications, an AR-score hybrid, and observed-score regressions assessed sensitivity. Results In the reference model, academic rumination was associated with affective coping (β = 0.283), and both were associated with dependency (β = 0.187 and 0.555). Dependency was associated with perceived internalization/memory debt (β = 0.550) and independent-reasoning debt (β = 0.644); their standardized difference was 0.095 (95% Monte Carlo CI [0.032, 0.157]). The corresponding observed-score coefficients were 0.413 and 0.518, with a difference of 0.105 (95% bootstrap CI [0.035, 0.175]). Both serial coefficient products remained positive, but the negative conditional coping-internalization association was not sustained in the observed-score analysis. Reported frequency had no unique association in the combined adjusted model. Alternative measurement models retained robust RMSEA of 0.090–0.097; robust hybrid indices were unavailable. Conclusion The results provide tentative evidence of differentiated self-reported associations among coping, frequency, dependency, and cognitive concerns. Measurement misfit remains unresolved, and the coping-outcome coefficients were sensitive to construct representation. Observed-score consistency for dependency and serial products does not validate the latent structure or establish temporal or causal mediation. These predominantly first-year, Tianjin-based cross-sectional data do not establish objective cognitive decline or generality across student populations.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: