Role play is a high-impact mode of training that has demonstrated its effectiveness in improving learning outcomes. However, it is difficult to scale to teacher instruction due to its inherent dependency on providing personnel who are both trained and available to facilitate this learning environment. This poses a challenge, especially to massive online courses which may employ and aid hundreds to thousands of novice teachers. In this work, we present EducaSim: a novel framework that uses generative agents to simulate a small-group section for teachers-in-training to practice instruction. EducaSim works by implementing diverse pedagogical-based personas, actual course material, and agent-based architectures constructed for instructional practice to provide a pedagogically rich environment for teachers-in-training to engage in role play learning -- without the costly overhead that comes with it. We share our experiences with constructing and making the tool available for experimental training and preparation in a six-week CS1 course supporting 20,000 students. We found that teachers who engaged generally saw it as a positive experience. We believe that EducaSim is an important step to providing experiential teaching practice at scale for closely-defined settings and has great potential for future applications.

翻译：角色扮演是一种高影响力的训练模式，其有效性已在提升学习成果方面得到验证。然而，由于该模式本质上依赖于配备经过培训且能协助营造此学习环境的专业人员，因此难以在教师教学中规模化应用。这对大规模在线课程尤其构成挑战，此类课程可能聘用并辅助数百至数千名新手教师。本研究提出EducaSim：一种利用生成式智能体模拟小型讨论组，供实习教师进行教学实践的新型框架。EducaSim通过实现多元化的教学型人格角色、真实课程材料以及专为教学实践构建的智能体架构，为实习教师提供沉浸式角色扮演学习的教学丰富环境——同时规避其伴随的高昂成本。我们分享了在一门为期六周、服务20,000名学生的CS1课程中，构建该工具并将其用于实验性培训与备课的经验。研究发现，参与使用的教师普遍视其为积极体验。我们相信，EducaSim是为精确定义场景提供规模化体验式教学实践的重要一步，具有广阔的未来应用前景。