Evaluating an AI-Powered Moodle Plugin for Enhancing Conceptual Understanding in Secondary Physics

This pilot study explores the feasibility of integrating large language model (LLM) assistants into physics education through a Moodle plugin designed to address conceptual understanding in Newtonian mechanics. Using OpenAI's GPT for real-time Socratic dialogue, the plugin guides students throu...

Descripción completa

Detalles Bibliográficos
Autores: Kahaleh, Rabih|||0009-0000-2418-2222, López Simó, Víctor|||0000-0002-2161-9211, Imad, Rodrigue, Maneva, Elitza Nikolaeva|||0000-0002-8638-1013
Tipo de recurso: artículo
Fecha de publicación:2025
País:España
Institución:Universitat Autònoma de Barcelona
Repositorio:Dipòsit Digital de Documents de la UAB
Idioma:inglés
OAI Identifier:oai:dnet:uabarcelona_::b7edb3139fef2fe75718abb9194fbc7e
Acceso en línea:https://ddd.uab.cat/record/327974
https://dx.doi.org/urn:doi:10.3991/ijet.v20i04.57879
Access Level:acceso abierto
Palabra clave:Large Language Models
Force Concept Inventory
Learning Management System
Conceptual change
Physics misconceptions
Socratic dialogue
AI Moodle Plugin
Artificial Intelligence in STEM education
Constructivist learning
Descripción
Sumario:This pilot study explores the feasibility of integrating large language model (LLM) assistants into physics education through a Moodle plugin designed to address conceptual understanding in Newtonian mechanics. Using OpenAI's GPT for real-time Socratic dialogue, the plugin guides students through misconception-targeted questions adapted from the Force Concept Inventory (FCI). Aligned with principles of inquiry-based learning, the intervention compares AI-guided feedback with instructor-guided materials over a one-week, three-session classroom study. Results suggest that students receiving AI-guided Socratic dialogue showed greater conceptual gains on certain targeted items (e.g., Newton's Third Law), whereas instructor guidance proved more effective for other concepts (e.g., mass and free-fall independence). Survey feedback highlights the immediacy and interactive nature of the AI while also noting a preference for the clarity provided by instructors. Qualitative analysis of open-ended question responses suggests that AI-driven dialogue promotes deeper reasoning when restating student ideas and scaffolding reflection. These preliminary findings underscore the potential of LLMs to support conceptual change in physics education when thoughtfully embedded within learning management systems, highlighting the complexity and value of personalized, interactive feedback for addressing student misconceptions.