Prototyping A Multi-Agent System to Enhance AI-Human Collaboration in Individualized Education Program Development

Abstract In this study, we report on the design, development, and evaluation of a prototype of CoIEP, a multi-agent system powered by large language models (LLMs) to co-create Individualized Education Programs (IEPs) with special education teachers. IEPs are legally binding documents that outline interconnected components, such as present levels of performance, annual educational goals, and specially designed instruction, all essential for ensuring access to the general education curriculum and high-quality learning experiences for students with disabilities. CoIEP is designed to streamline the complex process of IEP development by breaking down the step-by-step process of creating each IEP component and demonstrating the interconnectedness among these core components. In particular, CoIEP incorporates a “human-in-the-loop” function that provides ample opportunities to reflect on, evaluate, and improve LLM-generated content. Preliminary evaluations suggest that CoIEP has the potential to support educators in creating high-quality IEP core components. Implications for future research on developing a fully functional CoIEP and its potential use as a professional learning tool are discussed.