Automatically Generating Interactive Learning Experiences with an LLM-Driven Agentic Pipeline

Interactive and hands-on learning experiences encourage deeper engagement and produce better learning outcomes than passive consumption of media. We see an opportunity to improve learning by leveraging the ability of large language models (LLMs) to generate text and code for the automated production of interactive learning experiences. We present a chain of LLM agents that plan, generate, review, and refine a static interactive tutorial website for any topic. Their prompts were engineered to follow best prompting practices and employ pedagogic theory. This system is motivated by the effectiveness of interactive learning and the prospect of making education more accessible.