人工智能发展中的数据风险及治理  

Data Risks and Governance in the Development of Artificial Intelligence

作　　者：侯东德[1] HOU Dong-de(Civil and Commercial Law School,Southwest University of Political Science and Law,Chongqing 401120)

机构地区：[1]西南政法大学民商法学院,重庆401120

出　　处：《行政法学研究》2024年第6期3-15,共13页Administrative Law Review

基　　金：2020年度教育部哲学社会科学重大攻关项目“人工智能发展中的重大风险防范体系研究”(项目编号:20JZD026)。

摘　　要：人工智能是通过对数据的学习以达到独立于人类控制而产生特定输出的技术。但其在进行数据收集、分析、储存及传输时,可能会产生数据偏差、噪声、伪相关、黑箱及安全等风险。人工智能数据风险具有关联性与隐蔽性等特征,不同数据基础上的风险在治理层面又存在矛盾性。所以人工智能数据风险的治理十分复杂。考虑到人工智能数据风险及治理层面的特殊性,在治理数据风险时,需要通过整合创新的策略化解风险治理的矛盾性,通过合理的风险分配理念缓解风险治理价值冲突,通过多元、分级的治理思维弥合数据风险的隐蔽性。最终,在宏观层面确定促进与大数据融合发展的政策,在中观层面确定风险分配的制度化设计,在微观层面实现数据分级制度与多元协同治理框架,打造多元主体共担共治风险、共建共享发展的治理机制。Artificial intelligence is a technology that learns from data to produce specific outputs independently of human control.However, in the processes of data collection, analysis, storage, and transmission, it may generate risks such as data biases, noise, spurious correlations, black box issues, and security vulnerabilities.AI data risks are characterized by their interrelatedness and secrecy, and there is a contradiction in risk governance based on different data foundations, making the governance of AI data risks complex.Given the particularities of AI data risks and governance, managing these risks requires integrating innovative strategies to resolve governance contradictions, adopting reasonable risk distribution concepts to alleviate value conflicts in risk governance, and employing diverse, tiered governance approaches to bridge the secrecy of data risks.Ultimately, at the macro level, policies that promote integration with big data need to be established;at the meso level, institutional designs for risk distribution should be determined;and at the micro level, data classification systems and diverse collaborative governance frameworks should be implemented to create a governance mechanism where multiple stakeholders share responsibilities and collaboratively build and share in development.

关 键 词：数据风险 风险特征 风险分配 新质生产力发展 

分 类 号：F49[经济管理—产业经济] TP18[自动化与计算机技术—控制理论与控制工程]