Enhancing scientific argumentation with visual scaffolding in generative AI-supported environments

In recent years, the educational impact of generative AI (GAI) has attracted growing attention, with student dependency emerging as a key concern. Given that GAI primarily relies on text-based interaction, this study designed an integrated visual scaffolding approach combining the Toulmin Argument Pattern (TAP) and the fishbone diagram to support students’ scientific argumentation learning. A quasi-experimental study involving 211 students compared three experimental conditions and one control group, manipulating two factors: (1) visual scaffolding (integrated vs. minimal) and (2) GAI support (present vs. removed after Topic 3). Students participated in find-the-difference argumentation tasks, constructing claims, warrants, rebuttals, and qualifiers based on images of natural phenomena. Results suggest that integrated visual scaffolding may complement text-based GAI support by sustaining learning motivation, facilitating knowledge application, and promoting diverse argument construction. After GAI withdrawal, students who continued to use visual scaffolding demonstrated more stable performance through peer interaction, particularly in constructing evidence-based warrants and rebuttals. These findings highlight the complementary role of visual scaffolds in GAI-supported learning and offer implications for learner-centred instructional design.