用生成式AI和数字孪生打造个性化沉浸式技能训练框架
PRISM: A Personalized, Rapid, and Immersive Skill Mastery framework for personalizing experiential learning through Generative AI
- 结合生成式AI与数字孪生，动态调整学习内容
- GPT-4在零样本情感分析中达到91%的F1分数
- 适用于本科至博士的工业4.0沉浸式培训场景
生成式AI正深刻改变教育与职业培训领域。本文提出可扩展的PRISM框架（个性化、快速、沉浸式技能掌握），融合生成式AI与数字孪生（DT），实现自适应的体验式学习。通过情感分析与检索增强生成（RAG）技术监测学习者理解程度，并动态调整内容以达成教学目标。进一步提出教育多保真度数字孪生框架（MFDT-E），将数字孪生保真度层级与布卢姆分类学及柯氏评估模型对齐，支持本科生、硕士生及博士生的培养。实验表明，GPT-4在教师-学生对话的情感分析中实现91%的零样本F1值，而GPT-3.5在非正式语言场景中表现稳健。通过四个虚拟现实模块（居家场景、工厂巡检、灌装站数字孪生、个人防护装备检测）验证了系统在工业4.0沉浸式培训中的有效性与可扩展性。结果表明，生成式AI与数字孪生的结合，能实现个性化、高效且可扩展的教育。
原文摘要 · Abstract (English)
The rise of generative AI (gen-AI) is transforming industries, particularly in education and workforce training. This chapter introduces PRISM (Personalized, Rapid, and Immersive Skill Mastery), a scalable framework leveraging gen-AI and Digital Twins (DTs) to deliver adaptive, experiential learning. PRISM integrates sentiment analysis and Retrieval-Augmented Generation (RAG) to monitor learner comprehension and dynamically adjust content to meet course objectives. We further present the Multi-Fidelity Digital Twin for Education (MFDT-E) framework, aligning DT fidelity levels with Bloom's Taxonomy and the Kirkpatrick evaluation model to support undergraduate, master's, and doctoral training. Experimental validation shows that GPT-4 achieves 91 percent F1 in zero-shot sentiment analysis of teacher-student dialogues, while GPT-3.5 performs robustly in informal language contexts. Additionally, the system's effectiveness and scalability for immersive Industry 4.0 training are demonstrated through four VR modules: Home Scene, Factory Floor Tour, Capping Station DT, and PPE Inspection Training. These results highlight the potential of integrating generative AI with digital twins to enable personalized, efficient, and scalable education.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。