AI教育智能体赋能高校课程建设应用研究——以《软件构造》为例

摘 要：大语言模型（LLM）和AI智能体技术的发展，为高校课程教学提供了新的内容生成与教学支持工具。但在实际教学中，统一的课程知识结构与学生差异化的理解方式之间仍存在矛盾，教师积累的教学经验也难以通过传统方式持续复用。针对这些问题，本研究以强调抽象建模和代码推理能力的《软件构造》课程为例，依托云犀AI平台构建了由教学设计、代码设计和习题设计三个智能体组成的多智能体协作框架。教学设计智能体负责生成课程目标与教学方案。代码设计智能体负责构造和检查示例代码。习题设计智能体根据教学目标生成并校验练习题。该框架同时引入课程资料检索（如：RAG）和工具调用（Tool Use）机制，使三个环节能够共享课程知识和中间结果。初步实践表明，该框架能够提高不同教学材料之间的一致性，并为教师复用课程经验提供支持，进而帮助教师能够更好的完成对应的课程建设任务。不过，其教学效果仍需结合学生表现、教师工作量和生成内容质量等指标进一步验证。

关键词：大语言模型；AI智能体；《软件构造》；多智能体协同；课程建设

正文 

可下载并阅读全文PDF，请按照本文版权许可使用。
Download the full text PDF for viewing and using it according to the license of this paper.

Research on the Application of AI Education Intelligent Agents in Empowering University Curriculum Construction: A Case Study of “Software Construction”

Abstract: Large language models (LLMs) and AI agents are currently in a developmental phase. This trend is driving higher education to re-examine existing teaching paradigms. However, in actual teaching scenarios, there remains a significant gap between the inherent rigorous logic of the curriculum system and the fragmented state of students' cognition. Meanwhile, the challenge of adapting to students' personalized needs and the sustainability bottleneck of teachers' personal experience together constitute deep-seated obstacles in the current teaching reform. This study analyzes the aforementioned issues. Given the extremely high demand for system abstraction capability in the "Software Construction" course, this study selects it as a practical case. Leveraging the visual construction capability of the CloudRhino AI platform, this study constructs a multi-agent collaborative framework, composed of instructional design, code design, and exercise design agents. This framework abandons the point-based and fragmented tool mindset, not only using large language models (LLMs) as the core engine but also deeply integrating retrieval-augmented generation (RAG) mechanisms and tool-calling mechanisms. Its core logic can be summarized as transforming tasks such as lesson plan generation, code correction, and compliance evaluation into industrializable "teaching workflows" through refined role division and task decomposition. In the specific teaching practice of the "factory model," this framework successfully closed a complete chain from top-level logical planning, through bottom-level example construction, to final end-point exercise validation. Course practice has confirmed that the multi-agent-driven teaching system not only demonstrates strong compatibility with university curriculum development (taking software engineering as an example) but also, to some extent, promotes a paradigm shift in AI education research from "external tools" to "curricular elements.".

Keywords: Large Language Model; AI Agents; Software Construction; Multi-Agent Collaboration; Curriculum Developmen

参考文献

1. 孙立会;周亮.生成式人工智能赋能教育变革的逻辑——基于新质生产力的视角[J].教育研究, 2024(10).
2. Wang Hui.Proceeding of the 2024 International Conference on Diversified Education and Social DevelopmentC]//Advances in Social Sc-ience, Education and Humanities Research.P-roceedings of DESD 2024.Paris:Atlantis Press,2024.
3. 傅慧.人工智能技术在教学设计制作中的应用与实践[J].软件,2024,45(5):136-138.
4. 郑庆华.人工智能赋能创建未来教育新格局[J].中国高教研究,2024(3):1-7.
5. 黄荣怀.人工智能正加速教育变革：现实挑战与应对举措[J].中国教育学刊，2023（6）：26-33.
6. Holmes W,Bialik M,Fadel C.Artificial intellig-ence in education: promises and implications for teaching and learning[M].Boston:Center fo-r Curriculum Redesign,2023.
7. Zhang Z,Zhang Li D,Yu J,et al.Simulating classroom education with LLMempowered agents[J]. arXiv:2406.19226,2024.
8. Elnaffar S,Rashidi F,Abualkishik A Z.Teachin-g with AI:a systematic revie-w of chatbots,ge-nerative tools,and tutoring systems in progra-mming education[J]. arXiv:2510.03884,2025.
9. 赵丽琴,陈晓丹,范文杰.智能体辅助项目引导式教学在非计算机专业Python课程中的探索与实践[J].教育进展,2025,15(8):561–567.
10. 李建龙,牛振东.“技术-教育”互构下人工智能赋能高校教学评价研究[J].中国高教研究,2025,41(11): 15–23.
11. 李跃勇.生成式人工智能技术在高校教学中的应用研究[J].教育理论与实践研究,2025,3(24):53–60.
12. Kasneci E,Sessler K,Küchemann S,et al.ChatGPT for good?On opportunities and challenges of large language models for education[J].Nature Human Behaviour,2023,7(12):1833–1846.
13. Jiang Y H,Li R,Zhou Y,et al.AI agent for ed-ucation:von Neumann multi-a-gent system fra-mework[J].arXiv preprint arXiv:2501.00083,2024.
14. Vaccaro M Jr., Friday M, Zaghi A. Multi-Agentic LLMs for Personalizing STEM Texts. Applied Sciences. 2025; 15(13):7579.