Research on the Design and Application of Pair Programming Scripts for Primary School Students: Based on the Perspective of Productive Failure

Abstract

With the rapid development of generative large language model technology, multi-agent systems driven by large models have been widely applied in the field of education. Applying multi-agent systems to the evaluation of scientific problem-solving abilities helps to process multi-modal data, output efficient evaluation results, and provide personalized learning paths. This study, targeting primary and secondary school science education scenarios, adapted and modified a general intelligent agent framework based on large models, designing an intelligent agent framework for evaluating scientific problem-solving abilities and a multi-agent self-organizing collaborative framework. Through example demonstrations, the roles of intelligent agents are defined, the collaborative workflow of multiple intelligent agents is showcased, and the work responsibilities and processes of each agent are described in detail.