Using Large Language Models to Scale Up Teacher Guides

Many of the most effective interventions for improving literacy and numeracy skills center on structured pedagogy. These programs provide teachers with teacher guides that contain easy-to-use, high-quality lesson plans, and materials such as handouts to allow them to implement the lesson plans. A key constraint to scaling up these interventions is adapting materials to different contexts. Schools in the developing world vary significantly in terms of the language of instruction, the curriculum, and students’ initial skill levels. Scaling up an education program across Uganda alone, for example, would require adapting materials into dozens of languages of instruction. This project will address that constraint by using Large Language Models (LLMs) like ChatGPT to adapt existing high-quality teacher guides and other materials between contexts. In the pilot phase, we will test the feasibility of using LLMs to adapt existing materials for reading instruction from English into Rukiga.