The rapid development of artificial intelligence (AI), particularly ChatGPT, has sparked interest in its application in education. This study aims to investigate the potential of ChatGPT in helping students understand and solve physics problems, focusing on the Test of Understanding Graphs in Kinematics and the Determining and Interpreting Resistive Electric Circuit Concepts Test. The study involved 25 physics education students who completed these tests independently and with ChatGPT's assistance. The results revealed that students with a strong foundational understanding and reflective abilities interacted more effectively with ChatGPT, leading to improved answers and deeper conceptual understanding. In contrast, students with weaker prior knowledge tended to accept ChatGPT’s answers without critical reflection, perpetuating errors. Furthermore, ChatGPT showed limitations in interpreting image-based questions, reading scales, and providing consistent responses to concept-specific queries. These findings suggest that while ChatGPT has the potential to enhance learning, it requires thoughtful integration, particularly in helping students develop critical thinking and problem-solving skills. Teachers are encouraged to use ChatGPT’s limitations to design assessments that minimize the risk of cheating and foster deeper understanding. In conclusion, this study underscores the importance of combining AI tools with strong conceptual foundations and active reflection to optimize learning outcomes in physics education. Future research should focus on refining strategies for using AI in education to address its current limitations and enhance its effectiveness in complex learning scenarios.
Mafudi et al. (Mon,) studied this question.