Who Understands Instructional Design Better?——A Comparative Study on Text Evaluation Modes of Large Language Models and Multi-Agents

With the rapid development of artificial intelligence technology, the automation and intelligence of instructional design have become an important developmental trend in the field of education. However, the core bottleneck restricting the automated iteration of instructional design lies in the lack of efficient and stable evaluation methods for instructional design texts that are compatible with large language models. Based on evaluation experiments under different modes, this paper compared the performance of three evaluation methods in assessing information technology instructional design texts, namely direct evaluation by large language models, multi-agent collaborative evaluation, and multi-agent debate evaluation..The results revealed that the multi-agent collaborative evaluation mode not only significantly improved evaluation stability and human-machine consistency, but also generated more actionable improvement suggestions through role division and viewpoint debates. This provided a solid methodological foundation for the automated iteration of instructional design and offered new ideas and directions for the future intelligent development of instructional design.