Parametric Generation of Advanced Mathematics Assessment Items Using ChatGPT for Moodle Integration

Abstract

This paper explores how generative artificial intelligence—specifically ChatGPT—can
support the creation of assessment items in lower secondary mathematics education. The goal is to
enhance efficiency, cognitive diversity, and curriculum alignment in both formative and summative
evaluations, while addressing current challenges faced by teachers in item design. Prior Work: The
study builds upon existing research in AI-assisted learning, prompt engineering, and intelligent tutoring
systems. Prior developments in automated item generation have highlighted the promise of language
models but also the need for pedagogical oversight and contextual relevance. Approach: A designbased
research (DBR) framework was employed, involving iterative experimentation in classroom
settings and expert validation by mathematics teachers. The study focused on prompt formulation,
refinement cycles, and integration with learning management systems such as Moodle. Results:
ChatGPT was able to generate mathematically valid and curriculum-aligned items, ranging across
cognitive levels defined by Bloom’s Taxonomy. Human-in-the-loop validation was critical to ensuring
accuracy, clarity, and instructional value. Implications: The results suggest practical strategies for
teacher training in AI literacy and prompt engineering. Policymakers and educational institutions can
adapt this framework to scale AI adoption across STEM curricula. Value: The paper contributes a
replicable methodological model for AI integration in education. It provides empirical evidence of
ChatGPT’s utility in supporting authentic, personalized assessment design at the secondary level. This
aligns with global findings about the essential role of teacher-facing AI literacy and assessment training.