Creating and evaluating instructional materials with generative artificial intelligence: visual representations in astronomy education

Abstract Recent advancements in generative artificial intelligence for visualization offer significant opportunities to develop instructional materials. These opportunities can benefit astronomy education, which frequently relies on visual aids like representations. This document analysis study investigates ChatGPT-4o’s capabilities for creating and evaluating visual representations for astronomy education. Initially, ChatGPT-4o was utilized to create representations illustrating the phases of the Moon. These visuals were then evaluated based on predefined criteria, focusing on iconic features for reading and interpreting the representation. Subsequently, existing visuals on the same topic were provided to ChatGPT-4o, and it was asked to evaluate these visuals using the same predefined criteria. Finally, the assessments were compared with ChatGPT-4o’s evaluations. Findings suggest that while ChatGPT-4o-generated visual representations often contain inaccuracies that are potentially misleading for students, ChatGPT-4o effectively evaluates the iconic features to read and interpret the representations. This underscores the potential utility of generative artificial intelligence as a critical tool in ensuring the quality of visual representations in science education. It is recommended to integrate advanced artificial intelligence tools like ChatGPT-4o into the evaluation processes for instructional materials.