基于大语言模型的医学人工智能类课程AI教学助手开发与实践

Development and Practice of AI Teaching Assistants for Medical Artificial Intelligence Courses Based on Large Language Models

摘要: 在AI大语言模型赋能教育的背景下，针对通用大模型对课程知识不熟悉的问题，本研究基于DeepSeek-R1模型，结合RAG技术，以Dify低代码平台为基础，通过优化提示词(设定角色、明确格式)、构建课程知识库(导入教材等资料)等手段，构建了《智能医疗技术》课程AI答疑教学助手。该课程助手可7 × 24小时响应，支持多轮追问，82%使用者认可其解决课后即时疑问的价值，还能强化知识应用理解。研究指出其在高阶思维培养、知识更新上尚有局限，未来可向个性化学习伴侣发展。该方法显著降低开发门槛，为医学人工智能教育工具创新提供了一种有效的实践模式。

Abstract: Against the backdrop of education empowered by AI large language models (LLMs), and in response to the issue that general-purpose large models are not familiar with course-specific knowledge, this study, based on the DeepSeek-R1 model, integrates Retrieval-Augmented Generation (RAG) technology and takes the Dify low-code platform as the foundation. Through measures such as optimizing prompts (setting roles, clarifying formats) and constructing a course knowledge base (importing teaching materials and other resources), an AI Q&A teaching assistant for the “Intelligent Medical Technology” course has been built. This course assistant can provide 24/7 responses and support multi-turn follow-up questions. Eighty-two percent (82%) of users recognize its value in resolving immediate post-class doubts, and it can also enhance the understanding of knowledge application. The study points out that it still has limitations in the cultivation of higher-order thinking and knowledge updating, and can develop into a personalized learning companion in the future. This method significantly lowers the development threshold and provides a new paradigm for the innovation of educational tools in medical artificial intelligence.