Conceptions of GenAI-assisted learning among university students with different levels of self-perceived AI literacy: a drawing-based content analysis and epistemic network analysis

Abstract

As generative artificial intelligence (GenAI) becomes increasingly prevalent in higher education, understanding how students conceptualize its role in learning has become an important issue in AI education research. Existing studies have examined learning outcomes, usage intentions, and technology acceptance, but less is known about how students visually represent GenAI-supported learning and how these representations differ across levels of self-perceived AI literacy. This study combined drawing-based content analysis and epistemic network analysis (ENA) to examine drawings and self-report AI literacy data from 125 undergraduates at a university in Southwest China. Participants were divided into higher- and lower-scoring groups relative to the sample mean for exploratory network comparison. The higher-scoring group more often represented GenAI as a peer, depicted virtual-class settings, and included questioning, whereas the lower-scoring group more often represented GenAI as a tool and included information-search activities. ENA further showed different patterns of co-occurrence among the coded elements. These findings indicate an association between self-perceived AI literacy and selected visual representations of GenAI-supported learning, without implying differences in objectively demonstrated AI competence or actual learning behavior.