Scholarlink

<p>AI ethics often treats deployed AI systems as products, risk sources, technical systems, or components of sociotechnical arrangements. These descriptions remain necessary, but they do not fully capture what an AI system may become in operation. This article argues that a deployed AI system may acquire a role in practice that gives its output practical authority without making the system a moral subject. A hospital triage tool formally described as decision support may become the ordinary premise from which clinical judgement begins. The system need not malfunction, replace human decision-making, or acquire agency. Its output may nevertheless become action-guiding: a default, route, record, or premise. The article names this object an AI-driven operative artefact: a deployed AI system understood through the acquired role its output takes on in use. Such artefacts require operational answerability: continuing human and institutional responsibility to keep the acquired role identifiable, justifiable, contestable, and alterable. AI does not become responsible. Institutions remain answerable for the acquired role a non-agentive artefact’s output is allowed to occupy in practice.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From product to operative artefact: practical authority and operational answerability in AI ethics

  • 期刊论文

  • 发表日期：

    2026年7月24日

查看全文
  • Giovanni Velotto

摘要

AI ethics often treats deployed AI systems as products, risk sources, technical systems, or components of sociotechnical arrangements. These descriptions remain necessary, but they do not fully capture what an AI system may become in operation. This article argues that a deployed AI system may acquire a role in practice that gives its output practical authority without making the system a moral subject. A hospital triage tool formally described as decision support may become the ordinary premise from which clinical judgement begins. The system need not malfunction, replace human decision-making, or acquire agency. Its output may nevertheless become action-guiding: a default, route, record, or premise. The article names this object an AI-driven operative artefact: a deployed AI system understood through the acquired role its output takes on in use. Such artefacts require operational answerability: continuing human and institutional responsibility to keep the acquired role identifiable, justifiable, contestable, and alterable. AI does not become responsible. Institutions remain answerable for the acquired role a non-agentive artefact’s output is allowed to occupy in practice.