ISEDJ

Abstract: A key challenge in model curriculum development, especially for information systems, is its lengthy revision cycle, often spanning 5 to 10 years, and static reports that quickly become outdated. Recognizing this, recent model curricula, CC2020 and IS2020, call for a more sustained and continuous development process. In response, we present CurriculumGPT, a prototype AI tool built with a retrieval-augmented generation (RAG) architecture, to help realize that goal and address limitations of a prior tooling effort, the Computing Competencies Curricula Tool. We conducted a pilot study to evaluate the system across representative curriculum design tasks, including fact retrieval, comparative analysis, and content creation. Our training corpus included syllabi, course materials, learning objectives, and model curricula reports. Responses were assessed with a standardized human-applied rubric and the automated RAG evaluation framework, RAGAS, which evaluated quantitative metrics of answer faithfulness, relevancy, context precision, and recall. Results indicate that CurriculumGPT performed reliably, with improved performance achieved through metadata-aware chunking, a refinement to standard RAG approaches that preserves structural relationships in the data. Our findings show how AI and RAG systems can be adapted to support curriculum design while highlighting limitations and challenges that establish a foundation for future research in this emerging area.

Download this article: ISEDJ - V25 N1 Page 4.pdf

Recommended Citation: Dana, K., Babb, J., (2027). CurriculumGPT: Supporting Curriculum Design with AI-Augmented Tooling. Information Systems Education Journal 25(1) pp 4-22. https://doi.org/10.62273/CVEC6056