Evidence-Decision-Feedback: Theory-Driven Adaptive Scaffolding for LLM Agents
This work addresses the need for personalized adaptive scaffolding in educational LLM agents for students, though it appears incremental as it builds on existing tutoring systems and agentic behavior.
The paper tackled the problem of limited personalization in multi-agent LLM architectures for educational support by introducing the Evidence-Decision-Feedback (EDF) framework, which was tested in a high school classroom study and shown to align feedback with student understanding, promote scaffold fading, and support interpretable explanations without overreliance.
Multi-agent LLM architectures offer opportunities for pedagogical agents to help students construct domain knowledge and develop critical-thinking skills, yet many operate on a "one-size-fits-all" basis, limiting their ability to provide personalized support. To address this, we introduce Evidence-Decision-Feedback (EDF), a theoretical framework for adaptive scaffolding using LLMs. EDF integrates elements of intelligent tutoring systems and agentic behavior by organizing interactions around evidentiary inference, pedagogical decision-making, and adaptive feedback. We instantiate EDF through Copa, an agentic collaborative peer agent for STEM+C problem-solving. In an authentic high school classroom study, we show that EDF-guided interactions align feedback with students' demonstrated understanding and task mastery; promote gradual scaffold fading; and support interpretable, evidence-grounded explanations without fostering overreliance.