Abstract:

Since the public release of Large Language Models (LLMs) such as ChatGPT, the education landscape has witnessed groundbreaking innovations in personalized learning, automated content creation, and improved accessibility. Although previous research has explored generating curricula with GPT models, the methodology in AI-based curriculum generation remains insufficiently scientific. Therefore, this study explores the efficacy of two prompting techniques—Zero-Shot Role and Zero-Shot Chain-of-Thought (CoT)—in producing educationally sound curricula by comparing the time and iterations needed to create them. To determine curriculum quality, expert reviewers provided feedback using a curriculum evaluation rubric and semi-structured interviews. The CoT technique demonstrated greater efficiency, likely due to its step-by-step reasoning process, while the Role technique yielded more creative solutions but required more time and revisions. It is crucial to understand that the usefulness of both techniques varies depending on their context of use. Nevertheless, it is recommended that the design process be accompanied by an expert to enhance the quality of AI-generated curricula.

Keywords:

Learning plans Design, Curriculum Design, ChatGPT for Education Design, educational technology.