Using chatGPT to efficiently create large numbers of isomorphic assessment problems

Rapid advancements in generative artificial intelligence (GenAI) present new opportunities to enhance STEM education. This study explores the use of GenAI, specifically ChatGPT, for efficiently creating large sets of isomorphic physics problems at the introductory physics level. Isomorphic problems share the same underlying structure but differ in context and specific details such as the direction of forces or motion, as determined by the educator. We outline a multi-step process that guides GenAI to strictly adhere to the instructor’s detailed criteria, significantly improving problem quality over simpler prompt-based approaches. Our methodology not only creates diverse and contextually appropriate problem scenarios, but also generates and validates consistent and context appropriate random numerical values. This approach demonstrates the potential of GenAI to scale high-quality, varied practice material, enriching student engagement and understanding in introductory physics.