• 中国科学学与科技政策研究会
  • 中国科学院科技战略咨询研究院
  • 清华大学科学技术与社会研究中心
ISSN 1003-2053 CN 11-1805/G3

科学学研究 ›› 2025, Vol. 43 ›› Issue (1): 79-90.

• 理论与方法 • 上一篇    下一篇

中国语境下人工智能伦理治理的路径塑造———基于价值- 工具二维理性的融合框架

吴逸菲1,樊春良2   

  1. 1. 武汉大学 政治与公共管理学院
    2. 中国科学院科技战略咨询研究院
  • 收稿日期:2023-10-25 修回日期:2024-01-29 出版日期:2025-01-15 发布日期:2025-01-15
  • 通讯作者: 樊春良
  • 基金资助:
    国家社科基金重大项目——深入推进科技体制改革与完善国家科技治理体系研究

Shaping the Ethical Governance Path of Artificial Intelligence in the Chinese Context—Based on Value-Instrument Rationality

  • Received:2023-10-25 Revised:2024-01-29 Online:2025-01-15 Published:2025-01-15

摘要: 人工智能技术因其通用性和渗透性的特征,对中国的经济、社会、政治等领域带来了严峻的挑战,如传统科层制监管失效、算法和数据的公共责任如何分配等问题,影响到了政府、企业及个人等多方主体。纯技术方案或成文的立法监管都无法有效应对人工智能伦理问题及其治理困境。人工智能伦理治理作为一条新的道路,可以重塑资源配置方式和组织制度创新,能更好地应对人工智能伦理风险的深层性和复杂性。基于韦伯的理性二分法的分析框架,解构出人工智能伦理治理内含的价值理性和工具理性,进而明晰了塑造人工智能伦理治理路径的方向:人工智能伦理治理路径的价值向路和人工智能伦理治理路径的实践进路。双重维度进一步可以纵向划分为从治理逻辑、治理理念、治理主体和治理工具四个层面,构建人工智能伦理治理路径的运行机制。在治理逻辑上,压缩组织结构和重组权力结构,提升治理结构的协作性和灵活性;在治理理念上,通过普及人工智能伦理价值共识和明确人工智能伦理治理目标,提升社会整体的伦理治理意识;在治理主体上,加强国内人工智能伦理治理主体融合,建立人工智能国际伦理治理伙伴关系;在治理工具上,保持对人工智能技术迭代的跟踪,丰富人工智能伦理治理手段,构建人工智能伦理治理工具体系。

Abstract: Due to its versatility and permeability, artificial intelligence technology will also bring serious challenges to China's economic, social, political and other fields, such as the failure of traditional bureaucratic supervision and how to distribute public responsibilities for algorithms and data. It has affected many parties including governments, enterprises and individuals. Neither purely technical solutions nor written legislative supervision can effectively deal with artificial intelligence ethical issues and governance dilemmas. As a new path, artificial intelligence ethical governance can reshape resource allocation methods and organizational system innovation to better deal with the depth and complexity of artificial intelligence ethical risks. Based on Weber's rational dichotomy analysis framework, we deconstruct the value rationality and instrumental rationality contained in artificial intelligence ethical governance, and analyze the basic concepts of artificial intelligence ethical governance. The value rationality dominated by artificial intelligence ethics enables us to carefully examine the future of technological governance, laws and policies, and understand human needs for new artificial intelligence innovations from a unique perspective; the instrumental rationality dominated by artificial intelligence governance enables us to advance more pragmatically Thinking and solving ethical issues, thereby improving the overall benefits of the development of artificial intelligence technology. On the basis of the two-dimensional rational distinction, artificial intelligence ethical governance absorbs the idea of public interest and advocates a governance path from value to tool and then to value. From the two-dimensional way of thinking of value rationality and instrumental rationality, the concept of artificial intelligence ethical governance includes five core elements: human-centered value orientation. Adhere to the basic principles of scientific and technological ethics. Adhere to the principles of fairness, responsibility and transparency. The objects of artificial intelligence ethical governance include not only the systems produced by artificial intelligence technology, involving artificial intelligence computing power, algorithms, data, etc., but also the scenarios of artificial intelligence application. Multi-agent participation. Diversification of governance tools. The two-dimensional rationality of artificial intelligence ethical governance radiates into the corresponding governance behavior, resulting in the dual dimensions of value orientation and practical orientation of ethical governance. The realization of the value direction of artificial intelligence ethical governance can be approached from two levels: governance concept and governance logic. This is a layered result based on the dual rational fusion of artificial intelligence ethical governance. The value rationality of artificial intelligence ethical governance not only requires the guidance of traditional governance concepts, but also relies on the support of the organizational structure and power structure in the governance logic. In terms of governance concepts, by popularizing the consensus on artificial intelligence ethical values and clarifying the goals of artificial intelligence ethical governance, the overall ethical governance awareness of society is improved; in terms of governance logic, the organizational structure is compressed and the power structure is reorganized to enhance the collaboration and flexibility of the governance structure. . The practical approach to artificial intelligence ethical governance should focus on the two dimensions of governance tools and governance subjects. This is the result of artificial intelligence ethical governance absorbing institutional rationality. The establishment of an appropriate institutional environment relies on the exchange of information and close interaction among multiple governance subjects. In terms of governance subjects, strengthen the integration of domestic artificial intelligence ethical governance subjects and establish an international artificial intelligence ethical governance partnership; in terms of governance tools, maintain tracking of iterations of artificial intelligence technology, enrich artificial intelligence ethical governance methods, and build artificial intelligence ethical governance tools system.