How Do Multi-Agent Systems Shape Future Teachers? — An Experimental Study on AI-TPACK Competency Development

Abstract

Artificial intelligence (GAI) is driving the intelligent transformation of teacher education, but pre-service teachers still face challenges in integrating AI tools, selecting teaching methods, and fostering innovation in instructional design. Addressing how intelligent systems can enhance their AI-TPACK literacy has become a critical issue in teacher education. This study developed a multi-agent system (MAS)-based teaching design support tool, which includes subject knowledge agents (CK-Agent), teaching method agents (PK-Agent), and teaching tool agents (TK-Agent), working collaboratively to support the teaching design process. The experiment involved 20 pre-service teachers, randomly assigned to an experimental group (using MAS) and a control group (using traditional tools), who were tasked with completing teaching design tasks and evaluating their teaching plans based on the AI-TPACK framework. Additionally, the experimental group underwent pre- and post-intervention AI-TPACK literacy assessments. The results showed that MAS significantly improved the teaching design quality and promoted AI-TPACK literacy development among pre-service teachers. The contribution of this study lies in validating the feasibility of MAS in teacher education, expanding the application of AI-TPACK, and providing empirical evidence for the design of future intelligent teacher training systems.