通過算法審計規制自動化決策，以社會技術系統理論為視角

自動化決策並非純粹的技術工具，而是一種社會技術系統，由社會子系統和技術子系統構成，其引發的風險具備社會性，有必要進行規制。以個體賦權、信息規制、技術問責為代表的傳統規制路徑缺乏社會技術思維，容易陷入“破損謬誤”泥淖，難以因應自動化決策的風險與挑戰，無法有效實現法律及倫理規範提出的規制目標。算法審計作為一種社會技術方法，能夠很好地契合自動化決策的社會技術系統屬性，具備諸多方法論優勢和國際實踐經驗。從建構社會技術系統工程的角度出發，根據審計學的一般原理，我國可以從規範指引、實施主體、方法選擇和實施結果四個方面推動算法審計實現制度化發展，並在此基礎上探索基於場景的算法審計，實現可信賴的自動化決策。

Automated decision-making is a socio-technical system that consists of social and technical subsystems. It is not solely a technical tool and poses social risks that require regulation. The conventional regulatory approaches of individual empowerment, information regulation, and technical accountability lack socio-technical considerations and are susceptible to the“fallacy of the broken part”, which is inadequate for addressing the risks and challenges of automated decision-making. As a result, they are unable to effectively achieve the regulatory objectives outlined by legal and ethical norms. Algorithmic auditing can be a suitable socio-technical method for automated decision-making systems. It has many methodological advantages and practical experience internationally. To construct socio-technical system engineering, China can promote the institutionalized development of algorithmic auditing in four aspects: normative guidelines, implementation subjects, methodological choices and implementation results. The general principles of auditing should be followed. Additionally, context-based algorithmic auditing should be explored to achieve trustworthy automated decision-making.