Mirror Agents: A Dynamic Metacognitive Multi-Agent Framework for Real-Time Metacognitive Development

This research introduces a novel multi-agent framework, "Mirror Agents," to support K-12 students' metacognitive development through dynamic, real-time scaffolding.Building on the MAPS model, the system deploys four specialized agents-Guide, Progress, Reflection, and Strategy-that collaboratively monitor and respond to learners' cognitive and metacognitive activities.Unlike traditional systems offering static feedback, Mirror Agents continuously synthesize cognitive mirroring data (learning behaviors and outcomes) with metacognitive mirroring data (strategy use and self-regulation).Through multimodal inputs-including behavioral indicators, interface interactions, and physiological signals-the framework creates a comprehensive mirror of students' learning processes.This study explores patterns of metacognitive monitoring emerging from agent interactions and examines how learners adapt strategies in response to multi-agent feedback.The framework advances current educational technologies by offering a dynamic, interconnected agent model that enhances real-time metacognitive growth and provides design principles for intelligent learning environments.