Curriculum as Code: An AI-Assisted Architecture for Instructional Design in STEM Education

A recent arXiv paper proposes an AI-assisted instructional design architecture, termed 'Curriculum as Code', to streamline the creation of STEM education materials. This framework integrates Generative AI with LaTeX and Python to produce reproducible, visually consistent, and technically precise content, aiming to reduce faculty workload and enhance mathematical accuracy in active learning environments. It addresses current limitations of AI in formalizing instructional authoring and preventing 'hallucinations' in technical content.

Why it matters

This development is significant for sectors involved in technical education and training, as it offers a methodology to scale the production of high-quality instructional materials. By reducing the burden on subject matter experts and enhancing accuracy through automation, it can accelerate curriculum development and standardisation.

Key insights

• A six-phase AI-assisted instructional design architecture, 'Curriculum as Code', has been developed.
• The architecture combines Generative AI with LaTeX and Python for material creation.
• It targets the automation of reproducible, visually consistent, and technically precise STEM educational content.
• The primary goal is to alleviate the heavy workload on faculty in developing customized instructional materials for active learning.
• The framework seeks to overcome challenges such as the inadequacy of standard presentation tools for technical content and the propensity of current AI to 'hallucinate' or lack formalization in academic design.
• Intended outcomes include reduced preparation time and ensured mathematical accuracy.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.07364