Application of Large Language Models-Based Pedagogical Agents in Classroom Teaching

Abstract

With The innovative potential of Large Language Models (LLMs) in classroom instruction is becoming increasingly prominent, offering a transformative path for the field of education. The paper focuses on the application of LLM-based pedagogical agents in classroom teaching, aiming to address the limitations of traditional classrooms in providing personalized support and proactive services through their capabilities in multi-modal understanding, natural language generation, and task planning. Centered around an LLM, the pedagogical agents construct a digital brain equipped with reasoning, planning, and interactive abilities, serving multiple roles throughout the entire teaching process—including as a teacher’s assistant, a learning companion, and a personal tutor. The paper elaborates on its specific applications: generating intelligent resources and supporting instructional design during lesson preparation, acting as an interactive medium to facilitate teacher-student communication and personalized guidance during class, and serving as a one-on-one tutoring tool for reinforcement and generative assessment after class. Research shows that the pedagogical agent can effectively enhance teaching efficiency, increase student engagement, and promote the practical implementation of the modern educational philosophy of student-centered learning.