智慧图书馆生成式AI大模型风险治理机制研究  

Research on Risk Management Mechanism of Generative AI Large Model in Smart Library

作　　者：王静[1] 王鹏 Wang Jing;Wang Peng(Library of China University of Mining and Technology,Xuzhou 221116)

机构地区：[1]中国矿业大学图书馆,徐州221116

出　　处：《情报杂志》2024年第8期190-197,共8页Journal of Intelligence

基　　金：国家社会科学基金项目“基于数字孪生的高校智慧图书馆数据治理模式及机理研究”(编号:22BTQ023)研究成果。

摘　　要：[研究目的]探讨智慧图书馆中生成式AI大模型的风险治理机制,以期为有效治理提供模式、策略和框架,助力图书馆的数字化转型和服务创新。[研究方法]对生成式AI大模型在智慧图书馆中相关风险隐患和来源进行全面分析,提出相应的治理策略。结合敏捷治理和韧性治理理论,构建适用于智慧图书馆生成式AI大模型风险治理的多元协同参与的敏捷-韧性模式。针对生成式AI大模型在模型开发与训练、服务部署与维护、内容创建与审核、内容共享与传播、用户管理与教育和安全评估与监控不同生命周期治理阶段,提出相应的治理框架。[研究结论]研究表明,提出的多元协同参与的敏捷-韧性治理模式和生成式AI大模型全生命周期风险治理框架帮助图书馆在不同阶段采取相应的治理措施,通过建立适用于智慧图书馆的生成式AI大模型风险治理机制,旨在保障图书馆生成式AI大模型的安全和高效,进而促进智慧图书馆的持续发展。[Research purpose]This article explores the risk governance mechanism of generative AI models in smart libraries to provide models,strategies,and frameworks for effective governance and assist libraries in digital transformation and service innovation.[Research method]This paper comprehensively analyzes the related risks and sources of generative AI large models in the wisdom library and puts forward corresponding governance strategies.Combining agile governance and resilient governance theory,this paper constructs an agile and resilient model suitable for generating AI large model risk governance in a smart library.The corresponding governance framework is proposed for different lifecycle governance stages of generative AI large models in model development and training,service deployment and maintenance,content creation and review,content sharing and dissemination,user management and education,and security assessment and monitoring.[Research conclusion]Research have shown that the proposed agile resilience governance model with diverse collaborative participation and the generative AI big model full lifecycle risk governance framework help libraries take corresponding governance measures at different stages.By establishing a generative AI big model risk governance mechanism suitable for smart libraries,the aim is to ensure the safety and efficiency of the generative AI big model in libraries and thereby promote the sustainable development of smart libraries.

关 键 词：智慧图书馆 生成式AI 大模型 风险治理 数据治理 敏捷治理 韧性治理 

分 类 号：G251[文化科学—图书馆学]