- The paper develops a custom GPT that aligns with inquiry-based learning principles to support physics teachers in designing interactive lessons.
- It uses a structured approach across IBL phases with iterative refinement, role-playing, and context suitability to guide educators.
- Pilot study results, analyzed via the Wilcoxon signed-rank test, show significant improvements in teachers' attitudes towards AI-assisted personalized education.
Custom GPT for Inquiry-Based Learning in Physics Education
The paper presented by Dimitrios Gousopoulos from Panteion University of Social and Political Sciences describes the development and evaluation of a custom Generative Pre-trained Transformer (GPT) designed to enhance inquiry-based learning (IBL) in physics education. This custom GPT, referred to as IBL Educator GPT, aims to provide physics educators with a robust framework that aids in the interactive design of educational strategies by leveraging the capabilities of ChatGPT.
Overview
Generative AI, especially LLMs such as ChatGPT, has found significant utility across various domains, including education. This paper focuses on applying these advances to physics teaching through a custom GPT tailored to the principles of IBL. In the educational context, IBL encourages students to engage hands-on with scientific concepts, fostering a deeper conceptual understanding through exploration and inquiry.
The developed IBL Educator GPT facilitates educators in executing the IBL approach effectively. It aids teachers by providing a structured interaction model that spans essential IBL phases: orientation, conceptualization, investigation, conclusion, and discussion. For each phase, prompts are crafted on three foundational principles: iterative refinement, role-playing, and context suitability. This structure is intended to enable physics educators to make informed decisions by leveraging multiple perspectives generated by the GPT.
Evaluation and Results
To assess the impact of IBL Educator GPT, the authors conducted a pilot paper involving fourteen science educators. A pre- and post-intervention questionnaire, selected for its educational validity and reliability, gauged changes in teachers' perceptions of AI tools in personalized teaching environments. Importantly, statistical analysis using the non-parametric Wilcoxon signed-rank test revealed a significant improvement in educators’ perspectives post-intervention (Ζ = -3.296, p < .001). This outcome suggests that the custom GPT can effectively assist educators in lesson planning, task preparation, and professional development.
Implications
The findings suggest that integrating IBL-focused generative AI tools within educational frameworks can advance teaching practices by enriching the resources available to educators. The ability of the IBL Educator GPT to suggest varied strategies and perspectives potentially enhances pedagogical effectiveness and saves educators time on routine tasks. This time can then be reinvested in developing innovative teaching approaches that could lead to more engaging learning experiences.
Future Directions
The research opens avenues for further exploration into the application of custom GPTs for other scientific disciplines beyond physics. Additionally, the implications for scaling the tool across different educational levels could be profound, particularly in addressing the noted challenges that LLMs face in managing complex and advanced content. Future work could explore the integration of visual aids and multimodal outputs to mitigate some of the limitations noted with current LLM capabilities.
In conclusion, the paper contributes to the understanding of how custom GPTs can be aligned with educational methodologies to enhance teacher support and instructional design. It may stimulate continued efforts to develop AI-based educational tools that foster both teacher autonomy and student engagement through personalized learning journeys.