信息资源管理学报 ›› 2025, Vol. 15 ›› Issue (1): 54-68.doi: 10.13365/j.jirm.2025.01.054

• 研究论文 • 上一篇    下一篇

“工具-结构”视角下国内外生成式人工智能监管政策比较研究

邓胜利 丁威威 汪璠 王浩伟   

  1. 武汉大学信息管理学院,武汉,430072
  • 出版日期:2025-01-26 发布日期:2025-02-19
  • 作者简介:邓胜利,博士,教授,博士生导师,研究方向为大数据安全与信息服务;丁威威,硕士生,研究方向为信息行为与用户体验;汪璠(通讯作者),博士生,研究方向为虚假健康信息与生成式人工智能,Email: 1161252028@qq.com;王浩伟,硕士生,研究方向为生成式人工智能与数据治理。
  • 基金资助:
    本文系国家社会科学基金重大项目“信息资源管理学科研究方法知识库构建及其应用研究”(23&ZD229)的研究成果之一。

A Multinational Comparative Study of Regulatory Policies for Generative AI from the Perspective of “Tools-Structure”

Deng Shengli Ding Weiwei Wang Fan Wang Haowei   

  1. School of Information Management, Wuhan University, Wuhan, 430072
  • Online:2025-01-26 Published:2025-02-19
  • About author:Deng Shengli, Ph.D., professor, doctoral advisor, research interests include big data security and information service; Ding Weiwei, master candidate, research interests include information behavior and user experience; Wang Fan(corresponding author), Ph.D. candidate, research interests include false health information and generative artificial intelligence, Email: 1161252028@qq.com; Wang Haowei, master candidate, research interests include generative artificial intelligence and data governance.
  • Supported by:
    This paper is supported by the Major Program of the National Social Science Foundation of China "Research on the Construction and Application of Knowledge Base for Research Methods in Information Resource Management Discipline"(23&ZD229).

摘要: 本研究基于政策工具-结构特征双重视角,以不同国家生成式人工智能监管政策为研究对象,旨在挖掘生成式人工智能政策要素的结构特征和内在关联,以促进生成式人工智能健康发展。通过采集有关有效政策文本共计14份,运用文献计量法、内容分析法和BERTopic对327条相关文本单元进行编码和解读,从政策发文时间、政策发文主体和政策主题特征三个维度剖析其结构特征,分为环境型、需求型和供给型三种政策工具探讨其作用路径。研究发现,生成式人工智能监管仍处于起步阶段,不同国家政策总体特征的差距较为明显,在监管层面存在显著差异;从整体来看,政策工具作用路径由发挥环境型政策工具间接作用为主,呈现为结构不均衡、作用路径存在偏向性;进一步对比不同国家监管政策的异同点,针对性提出构建协同治理机制、优化政策工具结构、平衡政策工具作用路径和评估政策实施效果等相应对策建议,以促进生成式人工智能监管体系的协调发展。

关键词: 生成式人工智能, 监管, 政策工具, 文本分析, 主题识别, 协同治理

Abstract: Based on the dual perspective of policy tools and structural characteristics, this study takes the regulatory policies of generative AI in different countries as the research object, aiming to explore the structural characteristics and internal connections of the policy elements of generative AI, in order to promote the healthy development of generative AI. A total of 14 effective policy texts were collected, and 327 relevant text units were coded and interpreted using bibliometrics, content analysis, and BERTopic. The structural characteristics were analyzed from three dimensions: policy issuance time, policy issuance subject, and policy theme characteristics. The role paths were discussed by dividing into three types of policy tools: environmental, demand-driven, and supply-driven. The findings show that the regulation of generative AI is still in its infancy, and there are significant differences in the overall characteristics of policies among different countries, with significant differences in the regulatory level. Overall, the role path of policy tools is mainly dominated by the indirect role of environmental policy tools, showing a structural imbalance and bias in the role path. Further comparison of the similarities and differences in regulatory policies among different countries is conducted, and corresponding countermeasures and suggestions are put forward to optimize the policy tool structure, balance the role path of policy tools, assessing the effectiveness of policy implementation, and promote the coordinated development of the generative AI regulatory system.

Key words: Generative artificial intelligence, Regulation, Policy tools, Text analysis, Topic identification, Collaborative governance

中图分类号: