Adaptive Scaffolding Through LLM-Based Immediate Elaborative Formative Feedback: Advancing Self-regulated Learning Phases in Serious Games

Serious games in investment education often lack real-time support for self-regulated learning (SRL). This study evaluates adaptive scaffolding using LLM-based immediate elaborative formative feedback (LLM-IEFF) across three SRL phases: forethought, performance, and self-reflection. A quasi-experimental design compared a standard no-feedback serious game, tutor-based formative feedback (TBFF), and LLM-IEFF. A one-way ANCOVA showed significant differences (p < 0.001) among groups for all variables, with LLM-IEFF achieving the highest mean score gains, surpassing both no-feedback and TBFF methods. The findings suggest that LLM-driven feedback’s immediacy and depth offer superior cognitive support, reducing learner frustration and enhancing mastery. This research shows that generative AI can effectively scale personalized scaffolding in complex simulations, providing a better alternative to traditional human tutoring for developing self-regulated learners in investment education.