Learning Progression Informed Multi-Agent AI Feedback System for STEM Sensemaking (Poster 15)

• In Event: AERA Poster Session Saturday 11:45 am
  In Invited Poster Session: Excellence in Education Research: Early Career Scholars and Their Work Poster Session

Abstract

This study presents a learning progression (LP) informed, two-agent AI system supporting students’ math-science sensemaking (MSS) through interactive simulations. Grounded in a validated MSS LP, the system integrates two coordinated large language model (LLM)-based agents. The LP Classifier Agent evaluates learners’ written responses to determine their current LP level, while the Feedback Agent generates LP-aligned guidance to promote understanding. The system scaffolds learners as they explore phenomena, identify quantitative relationships, and construct mathematical representations of their observations. Expert review and iterative refinement improved the precision, clarity, and theoretical alignment of the feedback, establishing the system’s pedagogical validity prior to student deployment. Findings contribute to research on AI-supported learning and demonstrate how multi-agent systems can embody cognitive theory to enhance STEM learning.

Author

• Leonora Kaldaras, Texas Tech University