人工智能作为规范性参与者:人机互动中的规范关系分析
Artificial Intelligence as a Normative Participant:An Analysis of Normative Relations in Human-AI Interaction
摘要
随着人工智能日益深入地参与人类社会的实践活动,人类与人工智能之间形成何种实践关系,不仅取决于对人工智能内在属性的理论判断,还取决于人们在实际互动中如何对待和回应它.当人工智能以类人角色参与人机互动时,人类并不能始终对其保持纯粹的客体性立场,而会在互动中对其形成各种反应性态度,但这种参与关系的形成,并不足以证明人工智能已经具备进入相互问责关系所需的第二人称能力.借助达尔沃的第二人称理论,可以将人工智能参与规范实践与作为问责主体的情况区分开,并据此提出"规范性参与者"概念.进一步结合拉兹的规范性权力分析,可以说明人工智能作为规范性参与者并不意味着其自身获得规范性权力.在此基础上,我们可以从能动性保全、意义空间保护和责任可追溯性三个方面评估人工智能的参与方式,并通过司法人工智能和人工智能伴侣两个案例检验这些评估标准的适用性.
As artificial intelligence becomes increasingly involved in human social practices,the practical rela-tions that develop between humans and AI depend not only on theoretical assessments about the intrinsic properties of AI but also on how people treat and respond to AI in actual interactions.When AI takes on human-like roles in human-AI interactions,people may not consistently maintain a purely objective stance towards it.Instead,they may develop various reactive attitudes towards it.Such a participant relationship,however,does not suffice to establish that AI possesses the second-personal competence required for mutual accountability.Drawing on Darwall's account of second-personal normativity,a distinction can be made between AI's participation in normative practices and its status as an accountable subject.This distinction provides the basis for the concept of the"normative participant".Raz's account of normative powers further clarifies that the role of AI as a normative participant does not entail that AI itself acquires normative powers.On this basis,the ways in which AI companions can be assessed in terms of three criteria:the preservation of agency,the protection of the space of meaningful options,and the traceability of responsibility.The applicability of these criteria is tested through two cases:judicial AI and AI companionship.
年,卷(期):
2026, (5)
分类号:
B82
机标分类号:
G647
G40-011.8
D6
在线出版日期:
2026年11月6日