何以实现生成式人工智能技术的安全治理?——基于信息生态系统的博弈分析
DOI:
CSTR:
作者:
作者单位:

北京邮电大学 经济管理学院

作者简介:

通讯作者:

中图分类号:

TP18;G350

基金项目:

北京市社会科学基金项目“北京市人工智能大模型安全风险评估与治理长效机制研究”(编号:25JCC128)


How to Achieve the Secure Governance of Generative Artificial Intelligence Technology?: A Game-Theoretic Analysis Based on the Information Ecosystem
Author:
Affiliation:

School of Economics and Management,Beijing University of Posts and Telecommunications

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    生成式人工智能的飞速发展在加快形成新质生产力等方面成效显著,但其面临的数据、信息、内容等方面的治理挑战亟待解决,急需在技术发展与安全之间寻找平衡点。本文基于信息生态系统理论,构建生成式人工智能服务管理者、提供者、使用者三方主体协同治理博弈模型,进行数值仿真分析。研究发现,(1)生成式人工智能安全治理是一个多主体协同参与的过程,该过程受生态因素影响,加大各主体的共治收益及负面治理行为带来的损失都有利于促进信息生态系统稳定于{严格监管,积极治理,合规使用}的理想状态。(2)形象因素方面,提高生成式人工智能服务管理者的监管形象收益和提供者的声誉形象收益都可促进信息生态系统向理想状态进化。(3)奖惩因素方面,提高管理者对提供者的奖励以及提供者对使用者的奖励虽然会使实施奖励一方自身利益稍有损失,但可促进整个信息生态系统向理想状态进化,其中管理者给予的奖励存在有效阈值;加大管理者和提供者实施惩罚的力度可促进三方协同共治。(4)超额收益因素方面,压缩提供者及使用者的超额收益空间可促进三方协同共治。

    Abstract:

    The rapid development of generative artificial intelligence(GAI) has been effective in accelerating the formation of new quality productivity and other aspects, but the governance challenges it faces in terms of data, information and content need to be solved urgently, and there is an urgent need to find a balance point between technological development and security. Based on information ecosystem theory, this paper constructs a collaborative governance game model involving three stakeholders—managers, providers, and users of GAI services—and conducts numerical simulation analysis. It is found that (1) GAI safety governance is a process of multi-object collaborative participation, which is affected by ecological factors, and increasing the co-governance gains of each subject and the losses brought by negative governance behaviors are conducive to promoting the information ecosystem to be stable in the ideal state of {strict regulation, active governance, and compliant use}. (2) Regarding image factors, increasing the regulatory image gain of managers and the reputation image gain of providers can promote the information ecosystem to evolve to the ideal state. (3) Regarding rewards and penalties, increasing the rewards from managers to providers and from providers to users can promote the evolution of the entire information ecosystem to a desirable state, even though it may cause a slight loss in the interests of the party who imposes the rewards, where the rewards given by the managers are within an effective threshold; and increasing the penalties imposed by the managers and providers can promote the synergistic governance of the three parties. (4) With regard to the factor of excess earnings, compressing the space for excess earnings of providers and users can promote the synergistic governance of the three parties.

    参考文献
    相似文献
    引证文献
引用本文

马晓飞,王佳(通信作者).何以实现生成式人工智能技术的安全治理?——基于信息生态系统的博弈分析[J].技术经济,,():.

复制
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2025-10-17
  • 最后修改日期:2026-02-05
  • 录用日期:2026-01-29
  • 在线发布日期:
  • 出版日期:
文章二维码
您是第 位访问者
电话:010-65055536, 18515632865  Email:jishujingji@cste.org.cn
地址:北京东城区南竹竿胡同2号1幢(银河soho的C座)31001室  邮政编码:80-584
ICP:京ICP备05035734号-5
技术经济 ® 2026 版权所有
技术支持:北京勤云科技发展有限公司
×
《技术经济》
“扩大内需与全国统一大市场建设”专题征稿启事
《技术经济》2026年重点选题
“反内卷与经济高质量发展”专题征稿启事