Enhancing blended learning discussions with a Scaffolded Knowledge Integration–Based ChatGPT mobile instant messaging system

Abstract

Recent expansions in the use of ChatGPT within mobile instant messaging (MIM) platforms have garnered attention for their potential to enrich blended learning discussions. However, existing implementations often prioritize quick answers rather than pedagogically structured scaffolds, potentially limiting deeper learning. In this study, we introduce SKIMIM (Scaffolded Knowledge Integration–Based ChatGPT Mobile Instant Messaging), a system designed to systematically incorporate Scaffolded Knowledge Integration framework into AI-supported discussions. SKIMIM prompts learners to elicit their initial ideas, add new concepts, distinguish among different points of view, and reflect to refine their understanding. A 15-week randomized controlled trial (RCT) was conducted with 87 master's students assigned to three groups: SKIMIM, standard ChatGPT-MIM, and traditional MIM. Data were collected through engagement questionnaires, discussion logs, and semi-structured interviews, and analyzed via a mixed-methods approach covering behavioral, cognitive, and emotional dimensions of engagement, as well as user perceptions grounded in an extended Technology Acceptance Model. The results revealed that while standard ChatGPT-MIM provided higher behavioral participation and emotional comfort through rapid AI assistance, SKIMIM significantly enhanced cognitive engagement—particularly by fostering sense-making and innovation-level thinking. Although students experienced an initial adjustment period with SKIMIM's structured prompts, they ultimately reported comparable behavioral intention to use, along with notably higher perceived learning effectiveness and discussion quality. These findings underscore the importance of integrating AI with deliberate scaffolding strategies to achieve both active engagement and deeper cognitive outcomes in blended learning discussions.