| 3,105 | 20 | 1746 |
| 下载次数 | 被引频次 | 阅读次数 |
研究基于检索增强生成技术(RAG)，提出将大语言模型及私有知识库本地化部署，赋能新一代智能导学系统本地构建。研究首先构建基于RAG的智能导学增强生成系统框架。框架包括外部空间与本地空间：外部空间的多模态知识源、学习者信息经过筛选、脱敏、授权后可被部署在私有知识库中以备检索，学习者则通过用户界面与本地智能导学系统交互；本地空间在本地硬件的支持下，用户界面、大语言模型、私有知识库组成双循环结构，通过内外循环接力实现导学系统增强生成。其次，研究提出“人智共生”智能导学增强共同体，在“师-智”“生-智”“师-生”场域交界处衍生出智能导学系统的应用场域，形成人智协同的新一代智能导学模式。最后，为应对语言屏障、隐私安全、数字鸿沟及技术冲击等问题，研究提出四条风险治理策略，旨在保障新一代智能导学系统可持续发展，推动我国教育数智化转型。
Abstract：Large language models(LLMs) have significantly enhanced the capabilities of intelligent tutoring systems(ITS) in terms of learner perception, content generation, and cross-domain applications, fostering a leap from basic to advanced intelligence. However, challenges such as hallucinations, lack of deep logical reasoning, and privacy risks remain unresolved, posing severe threats to the secure application of ITS. To address these issues, this study proposes the localized deployment of LLMs and private knowledge bases empowered by RetrievalAugmented Generation(RAG) technology to enable localized operation of ITS. The study first constructs a RAG-based framework for the enhanced generation of ITS. This framework comprises external and local spaces: multimodal knowledge sources and learner information in the external space are filtered, anonymized, and authorized before being deployed in the private knowledge base. Learners interact with the local ITS through a user interface. In the local space, supported by local hardware, the user interface, LLM, and private knowledge base form a dual-loop structure, achieving enhanced tutoring through internal and external loops. Secondly, the study introduces a “Human-AI Symbiosis”intelligent tutoring enhancement community, where the intersections of the “Teacher-AI,” “Learner-AI,” and “Teacher-Learner” domains give rise to application fields of the ITS, forming a new generation of ITS characterized by human-AI collaboration. Lastly, to address challenges such as language barriers, privacy security, the digital divide, and technological impacts, the study proposes four risk governance strategies aimed at ensuring the sustainable development of the next-generation ITS and promoting the digital transformation of education in China.
[1] Nwana H S.Intelligent tutoring systems:an overview[J].Artificial Intelligence Review,1990,4(4):251-277.
[2] Mousavinasab E,Zarifsanaiey N,R.Niakan Kalhori S,et al.Intelligent tutoring systems:a systematic review of characteristics,applications,and evaluation methods[J].Interactive Learning Environments,2021,29(1):142-163.
[3]屈静,刘凯等.对话式智能导学系统研究现状及趋势[J].开放教育研究,2020,26(4):112-120.
[4]夏立新,杨宗凯等.教育数字化与新时代教育变革(笔谈)[J].华中师范大学学报(人文社会科学版),2023,62(5):1-22.
[5] Steenbergen-Hu S,Cooper H.A meta-analysis of the effectiveness of intelligent tutoring systems on K–12 students’ mathematical learning[J].Journal of educational psychology,2013,105(4):970.
[6]卢宇,骈扬等.新型智能导学系统构建及其关键技术[J].中国远程教育,2023,43(7):30-38+46.
[7]徐升,佟佳睿等.下一代个性化学习:生成式人工智能增强智能辅导系统[J].开放教育研究,2024,30(2):13-22.
[8] Jeon J,Lee S.Large language models in education:A focus on the complementary relationship between human teachers and ChatGPT[J].Education and Information Technologies,2023,28(12):15873-15892.
[9]王炜,赵帅等.生成式人工智能教育创新应用的人本主义追求——对UNESCO《教育与研究领域生成式人工智能指南》的解读[J].现代远程教育研究,2024,36(1):3-11.
[10] Ma Y,Zhang C,Zhu S C.Brain in a vat:On missing pieces towards artificial general intelligence in large language models[DB/OL].https://arxiv.org/abs/2307.03762,2023-07-07.
[11] Tang X,Zheng Z,et al.Large language models are in-context semantic reasoners rather than symbolic reasoners[DB/OL].https://arxiv.org/abs/2305.14825,2023-06-08.
[12] Kasneci E,Se?ler K,et al.ChatGPT for good?On opportunities and challenges of large language models for education[J].Learning and individual differences,2023,103:102274.
[13] Carbonell J R.AI in CAI:An artificial-intelligence approach to computer-assisted instruction[J].IEEE transactions on man-machine systems,1970,11(4):190-202.
[14]刘清堂,吴林静等.智能导师系统研究现状与发展趋势[J].中国电化教育,2016,(10):39-44.
[15] VanLehn K.The behavior of tutoring systems[J].International journal of artificial intelligence in education,2006,16(3):227-265.
[16]刘华,祝智庭.大模型支持的会话辅导系统开发:范式转换、架构设计与能力训练[J].现代远程教育研究,2024,36(3):11-19.
[17] McNamara D S,Levinstein I B,Boonthum C.iSTART:Interactive strategy training for active reading and thinking[J].Behavior Research Methods,Instruments,&Computers,2004,36(2):222-233.
[18] Lane H C,VanLehn K.Teaching the tacit knowledge of programming to novices with natural language tutoring[J].Computer Science Education,2005,15(3):183-201.
[19] Suebnukarn S,Haddawy P.A collaborative intelligent tutoring system for medical problem-based learning[C].New York:Association for Computing Machinery,2004.14-21.
[20] Aleven V,Mclaren B,et al.Toward meta-cognitive tutoring:A model of help seeking with a Cognitive Tutor[J].International Journal of Artificial Intelligence in Education,2006,16(2):101-128.
[21] Steenbergen-Hu S,Cooper H.A meta-analysis of the effectiveness of intelligent tutoring systems on K–12 students’ mathematical learning[J].Journal of educational psychology,2013,105(4):970-987.
[22] VanLehn K.The relative effectiveness of human tutoring,intelligent tutoring systems,and other tutoring systems[J].Educational psychologist,2011,46(4):197-221.
[23] Jaques N,Conati C,et al.Predicting affect from gaze data during interaction with an intelligent tutoring system[C].Cham:Springer,2014.29-38.
[24] Pardos Z A,Baker R S J D,et al.Affective states and state tests:Investigating how affect and engagement during the school year predict end-of-year learning outcomes[J].Journal of Learning Analytics,2014,1(1):107-128.
[25] Vaessen B E,Prins F J,Jeuring J.University students’ achievement goals and help-seeking strategies in an intelligent tutoring system[J].Computers&Education,2014,72:196-208.
[26]廖剑,刘选,刘革平.理感联通:人工智能赋能智慧教育新范式[J].中国电化教育,2023,(6):18-24.
[27] Chen W Y.Intelligent Tutor:Leveraging ChatGPT and Microsoft Copilot Studio to Deliver a Generative AI Student Support and Feedback System within Teams[DB/OL].https://arxiv.org/abs/2405.13024,2024-05-15.
[28] Abdelghani R,Wang Y H,et al.GPT-3-Driven Pedagogical Agents to Train Children’s Curious Question-Asking Skills[J].International Journal of Artificial Intelligence in Education,2024, 34(2):483-518.
[29] Yadav G,Tseng Y J,Ni X.Contextualizing problems to student interests at scale in intelligent tutoring system using large language models[DB/OL].https://arxiv.org/abs/2306.00190,2023-05-31.
[30] Firdaus M F,Wibawa J N,Rahman F F.Utilization of GPT-4 to Improve Education Quality Through Personalized Learning for Generation Z in Indonesia[J].IT for Society,2023,8(1):6-14.
[31]翟雪松,季爽等.基于多智能体的人机协同解决复杂学习问题实证研究[J].开放教育研究,2024,30(3):63-73.
[32] Huang L,Yu W,et al.A survey on hallucination in large language models:Principles,taxonomy,challenges,and open questions[DB/OL].https://arxiv.org/abs/2311.05232,2023-11-09.
[33]罗云鹏.大模型发展亟需高质量“教材”相伴[N].科技日报,2024-01-15(06).
[34]张家军,闫君子.“道”“器”融合:教师减负的技术旨归与实现路径[J].中国电化教育,2022,(12):68-73+83.
[35]任玉鑫,王光臣等.数字赋能背景下研究生使用ChatGPT实际效能、风险检视及治理路径[J].黑龙江高教研究,2024,42(5):73-79.
[36] Dong C.How to build an AI tutor that can adapt to any course and provide accurate answers using large language model and retrieval-augmented generation[DB/OL].https://arxiv.org/abs/2311.17696,2024-06-21.
[37] Lewis P,Perez E,et al.Retrieval-augmented generation for knowledgeintensive nlp tasks[J].Advances in Neural Information Processing Systems,2020,33:9459-9474.
[38]王明程,王高开等.基于大模型智能体的安全风险态势感知框架构建[J].情报理论与实践,2024,47(7):190-198.
[39]卢宇,骈扬等.新型智能导学系统构建及其关键技术[J].中国远程教育,2023,43(7):30-38+46.
[40] Zhao P,Zhang H,et al.Retrieval-augmented generation for ai-generated content:A survey[DB/OL].https://arxiv.org/abs/2402.19473,2024-06-21.
[41]杨霞,范蔚.许诺与虚诺:数字化学习空间的异化与消解[J].中国远程教育,2024,44(10):46-59.
[42] Wenger B E,See F B J.Artificial intelligence and tutoring systems[M].Burlington:Morgan Kaufmann,1987.
[43]杜修平,王崟羽等.AIGC赋能“中文+职业教育”资源智能生成与质量进化——内涵、机理与模式构建[J].电化教育研究,2024,45(5):121-128.
[44]赵晓伟,祝智庭等.教育提示语工程:构建数智时代的认识论新话语[J].中国远程教育,2023,43(11):22-31.
[45]童慧,杨彦军.基于“技术道德化”理论的生成式人工智能教育应用潜能与风险研究[J].电化教育研究,2024,45(7):12-18.
[46]刘清堂,黄景修等.PST视角下智能导师系统的设计与新发展——以句酷批改网为例[J].现代教育技术,2017,27(5):81-87.
[47] Garrison D R,Anderson T,Archer W.Critical inquiry in a text-based environment:Computer conferencing in higher education[J].The internet and higher education,1999,2(2-3):87-105.
[48]谢晓雪,柳士彬.数智时代人类教师与虚拟教师融合的目标、场景与路径[J].现代远程教育研究,2024,36(2):45-51.
[49]戴岭,赵晓伟等.智慧问学:基于ChatGPT的对话式学习新模式[J].开放教育研究,2023,29(6):42-51+111.
[50]龚礼林,刘红霞等.情感导学系统(ATS)的关键技术及其导学模型研究——论智能导学系统走向情感导学系统之意蕴[J].远程教育杂志,2019,37(5):45-55.
[51]王凯,余发碧等.基于情绪识别技术的学科情感目标达成检测与教学改进[J].中国电化教育,2023,(6):83-89.
[52]王志军,滕志强等.国际生成式人工智能教育应用创新——全球人工智能学习和教育研究联盟之“人工智能赋能学习者会议”综述[J].远程教育杂志,2024,42(2):65-74.
[53]杜修平,杨小荣等.构建科技汉语“预本全链条衔接”教学新模式[J].云南师范大学学报(对外汉语教学与研究版),2024,22(2):76-84.
[54]杜修平,尹晓静.理工中文[M].北京:北京语言大学出版社,2024.
[55]冯志伟,张灯柯.语言模型与人工智能[J].外语研究,2024,41(1):1-19+112.
[56]刘邦奇,聂小林等.生成式AI赋能教育:技术框架、应用场域及价值——2024智能教育发展研究报告[J].中国电化教育,2025,(3):61-70.
[57]卢宇,余京蕾等.基于大模型的教学智能体构建与应用研究[J].中国电化教育,2024,(7):99-108.
[58]兰国帅,杜水莲等.技术何以赋能未来高等教育教学:趋势、实践和场景——《2024年EDUCAUSE地平线报告(教学版)》要点与启示[EB/OL].http://kns.cnki.net/kcms/detail/32.1843.G4.20240702.1011.002.html,2024-07-30.
[59] Gan C,Yang D,et al.Similarity is Not All You Need:Endowing Retrieval Augmented Generation with Multi Layered Thoughts[DB/OL].https://arxiv.org/abs/2405.19893,2024-05-30.
基本信息:
中图分类号:G434
引用信息:
[1]杜修平,王崟羽.检索增强生成赋能智能导学系统构建研究——基于本地大模型与私有知识库[J].中国电化教育,2025,No.460(05):117-127.
基金信息:
2022年教育部中外语言交流合作中心国际中文教育研究课题重点项目“‘中文+职业教育’融合发展机理研究”(项目编号：22YH30B)研究成果
2025-05-10
2025-05-10