Cognitive capability and behavioral exposure in generative AI use: a dual-pathway model of perceived facilitation and integrity-related risk

The rapid adoption of generative artificial intelligence (GenAI) has introduced new forms of cognitive support as well as potential behavioral risks in academic contexts. Rather than treating AI use as a purely homogeneous phenomenon, this study conceptualizes GenAI engagement through a psychological dual-pathway framework that explicitly separates capability-oriented AI familiarity from exposure-based AI use frequency. We examine how these distinct pathways map onto perceived cognitive facilitation, cognitive reliance, and integrity-related behavioral risk. Data were collected from two independent samples of university students ( N = 407 and N = 228; total N = 635) and analyzed using structural equation modeling. Measurement invariance analyses supported configural, metric, and scalar invariance across the two samples, indicating that the latent constructs were interpreted in broadly comparable ways across cohorts. Results from the main sample indicated that AI familiarity was positively associated with perceived cognitive facilitation and negatively associated with integrity-related behavioral risk, whereas AI use frequency showed weaker and less consistent relationships with the outcome variables. However, these structural relationships were substantially weaker and largely non-significant in the replication sample, suggesting that the observed associations may be context-sensitive rather than uniformly stable across learning environments. The findings highlight an important distinction between cognitive capability and behavioral exposure in understanding students’ engagement with GenAI. Overall, the results suggest that students’ educational experiences with GenAI may depend less on simple usage frequency and more on their ability to critically understand, evaluate, and regulate AI-supported learning processes. These findings have implications for AI literacy development, academic integrity governance, and assessment design in higher education.