Abstract
This study aimed to visualize collaboration processes in a generative AI–integrated advanced university Japanese course using Multi-Channel Sequence Analysis (MCSA). Based on five-dimensional qualitative coding, the processes were categorized into two types. Frequency comparison, Epistemic Network Analysis (ENA), and state-transition visualization showed broadly similar overall structures, but a clear difference in the timing of AI use: Type 1 followed an “embedded” pattern with on-demand use during production, whereas Type 2 showed a “pre-processing” tendency, using AI after initial interaction to refine concepts and expressions before moving to production.
