Scaffolds or crutches? Examining the relationships between artificial intelligence dependency and academic resilience in higher education

As generative artificial intelligence (AI) becomes increasingly embedded in higher education, an important question is whether such support provides productive scaffolding or fosters maladaptive dependency. This study examines the relationships among AI dependency, academic self-efficacy, and academic resilience, and whether perceived feedback elaborativeness and latency moderate the dependency–self-efficacy relationship. A three-wave lagged survey was conducted with 892 first-year university students in China over one academic semester. Higher AI dependency was associated with lower academic self-efficacy, which in turn was positively associated with academic resilience. Academic self-efficacy accounted for a significant indirect association between dependency and resilience. The negative dependency–self-efficacy association was weaker when students perceived feedback as more elaborative and as allowing more time for independent thought before assistance. Responsible use of AI in education calls for feedback design that supports independent thinking and preserves learner agency.