Answers are grounded in this paper and cite their sources.
Generative artificial intelligence (GenAI) is increasingly playing a role in learning, information work, and decision-making. The present study examines whether how people use GenAI matters more than how often they encounter it. In a quantitative cross-sectional survey of 350 GenAI users, usage intensity (years of use, weekly frequency, and daily duration) was distinguished from functional AI reliance (using AI instead of completing a task independently). Cognitive dependence, self-efficacy, critical thinking, and decision-making were analysed as separate outcomes, with cognitive dependence also tested as an indirect-path variable and response checking as a moderator. Usage intensity was not a significant predictor in the final models. Adding response checking and functional reliance jointly increased explained variance, with the largest unique contribution coming from reliance for cognitive dependence (ΔR²=.366) and self-efficacy (ΔR²=.072), and from response checking for critical thinking (ΔR²=.241) and decision-making (ΔR²=.026). Functional reliance was positively associated with cognitive dependence (β=.611, p<.001) and self-efficacy (β=.271, p<.001), negatively associated with critical thinking (β=−.117, p=.007), and not significantly associated with decision-making (β=.027, p=.626). Indirect effects through cognitive dependence were observed for critical thinking (ab=−.1316, 95% CI [−.1971, −.0668]) and self-efficacy (ab=.0829, 95% CI [.0247, .1468]); for decision-making, the negative indirect effect (ab=−.0917, 95% CI [−.1630, −.0192]) opposed the positive direct path and the total association was non-significant. The linear-ordinal moderation model showed a significant AI reliance × response checking interaction for decision-making (B=.0978, p=.012), but the small increment (f²=.017) and sensitivity to how response checking was specified warrant cautious interpretation. Because the design is cross-sectional and the principal outcomes are self-reported, the findings indicate statistical associations rather than causal effects or demonstrated changes in cognitive ability.
No takes yet. Share an insight, caveat, or question.
Awadelkarim et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: