摘要： 完全自动化决策更易引发系统性风险,对隐私和公平造成威胁。欧盟立法以“人在环路”作为防范自动化决策系统滥用危害的基本遵循,对完全自动化决策一般禁止、例外允许。“人在环路”应被解释为信息处理者的干预义务,而非信息主体的权利,利于个体正义的实现和责任主体的明确。我国立法规定了完全自动化决策拒绝权,信息主体可行使请求权以迂回实现“人在环路”之法律效果,亦避免给处理者过重义务,对人工智能产业的发展造成不合理的阻碍。该项权利并非事后救济权,而是信息主体在决策的事前、事中及事后的“个人参与”,应满足“仅通过自动化决策的方式作出决定”和“对个人权益有重大影响”两个要件,且需与信息主体的其他权利配合行权。

关键词: 完全自动化决策, 混合决策, 个人参与, 人工干预, 人在环路

Abstract: Solely automated decision-making system has higher risks and is more likely to threaten value of privacy and fairness in different contexts.Therefore, EU legislation adopts the general clause plus enumeration mode, and takes “keep human in the loop” as the basic principle to prevent the abuse of automated decision-making system.“Human in the loop” is the intervention obligation of the information processor, rather than the right of the information subject.It is the superior path to realize individual justice and clarify the accountability subject.China's law chooses a more eclectic right-model.Right to refusing solely automated decision-making can detour to realize the “human in the loop” on the one hand, and avoid unreasonable obstacles to the development of artificial intelligence on the other hand.This right is not an ex post remedy right, which actually is to ensure “individual participation” of the whole loop of the decision-making.It must meet the two requirements of “making a decision by solely automatic decision-making” and “having a significant impact on personal rights and interests”, and it needs to be exercised in cooperation with other rights of information subject.

Key words: Solely Automated Decision-making, Hybrid Decision-making, Individual Participation, Human Intervention, Human in the Loop