Understanding university students’ generative AI use intentions in digital learning: An AIDUA-informed dual-path risk–value model

Generative artificial intelligence (GenAI) combines salient functional benefits with uncertain, opaque, and sometimes uncomfortable human-like interaction. This study develops an AIDUA-informed dual-path model to examine how value assessment and risk assessment are associated with university students’ intentions to use GenAI for learning. The model integrates digital literacy, social influence, time benefit, perceived learning effectiveness, anthropomorphic discomfort, algorithm aversion, perceived risk, and behavioral intention. Survey data from 336 university students in China were analyzed using partial least squares structural equation modeling. In the value pathway, time benefit had the strongest direct association with behavioral intention, while digital literacy and social influence shaped time benefit and perceived learning effectiveness. In the risk pathway, anthropomorphic discomfort was positively related to algorithm aversion, which strongly predicted perceived risk. Perceived risk had a statistically significant but small positive association with behavioral intention, whereas the direct path from algorithm aversion to intention was not significant. Bootstrap analysis supported a positive indirect effect of algorithm aversion through perceived risk, but its total effect on intention was not significant. The findings therefore indicate coexistence of risk awareness and use intention rather than evidence that risk causes adoption. The study contributes a context-specific refinement of AI acceptance research and identifies design, literacy, and governance implications for responsible GenAI-supported learning.