人工智能敏捷治理实践:分类监管思路与政策工具箱构建
Agile Governance Practices in Artificial Intelligence:Categorizing Regulatory Approaches and Constructing a Policy Toolbox
薛澜 1贾开 2赵静3
作者信息
- 1. 清华大学苏世民书院;清华大学公共管理学院,北京 100084
- 2. 上海交通大学国际与公共事务学院;清华大学产业发展与环境治理研究中心
- 3. 清华大学公共管理学院,北京 100084
- 折叠
摘要
纷繁复杂的人工智能应用风险在带来监管挑战的同时,也让追求敏捷成为当前人工智能治理的共识性理念.但如何在具体路径、机制、工具上实现真正的敏捷监管尚缺乏系统的学理分析与实践指引.遵循科层体系运作逻辑的监管部门在克服制度张力和应对不确定性方面存在天然不足,实践改革因此普遍将"分类"视为一种适应复杂性与提升敏捷性的潜在思路.通过对人工智能多领域业态发展模式与技术创新规律的比较分析,以及综合人工智能各领域现有国际治理经验,本文构建了人工智能治理风险的关键分类维度及与之相适应的全谱系政策工具箱.研究试图为监管实践者提供一个合意的分类思路与政策工具组合指引,使监管实践可从产业特征标定分类治理对象,并在对应的政策工具箱中寻找政策组合.
Abstract
The diverse and complex risks associated with artificial intelligence applications present regulatory challenges,emphasizing the need for agility in current AI governance.However,there is still a lack of systematic theoretical analysis and practical guidance on achieving agile regulation in terms of specific paths,mechanisms,and tools.Regulatory agencies,often constrained by bureaucratic approaches,face difficulties in responding to uncertainties.As a result,practical reforms often consider"classification"as a potential approach to adapt to complexity and enhance agility.Through a comparative analysis of AI development across diverse domains,along with a comprehensive review of international experiences,this paper establishes the dimensions for classifying AI governance risks and corresponding policy toolboxes.The study aims to provide regulatory practitioners a concise classification framework and a guide for combining policy tools.This enables regulatory practices to identify and categorize governance targets based on industry characteristics and locate suitable policy combinations within the corresponding toolbox.
关键词
人工智能治理/分类治理/敏捷治理工具箱Key words
artificial intelligence governance/classification governance/agile governance toolbox引用本文复制引用
基金项目
国家社会科学基金(23BGL244)
清华大学自主科研项目(2021THZWJC12)
清华大学-丰田基金专项()
出版年
2024
NETL
NSTL
维普
万方数据