Answers are grounded in this paper and cite their sources.
Quasi-experimental study shows generative AI scaffolding promotes learning initiation but triggers illusions of understanding, indicating that self-regulation is vital to protect deep engagement.
For learners with low prior knowledge, learning difficulties often arise not from limited ability, but from not knowing how to begin, interpret task requirements, or establish an initial learning direction. This study examined whether multimodal GenAI, when added as an initial scaffolding resource within a shared flipped learning structure, could support learning initiation while also introducing the risk of an illusion of understanding. Specifically, it investigated a dual-pathway mechanism involving reduced uncertainty, perceived readiness, self-efficacy, student engagement, illusion of understanding, deep cognitive engagement, and self-regulation. A mixed-methods quasi-experimental design with embedded qualitative analysis was adopted. Quantitative data from pre- and post-test surveys, prior knowledge assessments, and deep-task measures were analyzed using ANCOVA, structural equation modeling, and multi-group comparisons. Qualitative data from reflective journals, multimodal GenAI usage reports, and interviews were used to clarify students’ learning processes. The results suggested that the additional provision of multimodal GenAI as initial scaffolding was associated with reduced uncertainty, perceived readiness, self-efficacy, and student engagement. However, perceived readiness was also associated with the illusion of understanding, which was negatively related to deep cognitive engagement. Self-regulation moderated these pathways: highly self-regulated learners benefited more from the facilitative pathway, whereas low self-regulated learners were more vulnerable to the risk pathway. These findings suggest that multimodal GenAI can help activate an early-stage psychological pathway for learning initiation, but flipped learning designs incorporating GenAI should also sustain verification, reflection, and productive uncertainty. The study offers a dual-pathway perspective for facilitating learning initiation while protecting deep cognitive engagement.
No takes yet. Share an insight, caveat, or question.
Gu et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: