Scholarlink

<p>To examine how clinical AI systems establish de facto standards of care through design and administrative mechanisms; analyze power redistribution across stakeholders in EU, US, and UK governance regimes; and propose comprehensive policy tools to mitigate unintended normative effects while enhancing equity, accountability, and public trust in healthcare AI systems. A comparative qualitative document analysis was conducted on regulatory texts (e.g., the EU AI Act, FDA guidance), hospital protocols, and vendor materials (2015–2025) from the EU, the US, and the UK. The NORM5 typology (Nudges, Override friction, Responsibility choreography, Metric coupling, Scripted workflows) was developed through systematic inductive coding, informed by political science theories of institutional drift and bureaucratic power. Data collection involved systematic sampling of 127 documents across three jurisdictions, with thematic analysis conducted using established qualitative research protocols. NORM5 mechanisms subtly shift clinical norms through five distinct pathways, redistributing power toward vendors, payers, and healthcare organizations while systematically eroding clinician autonomy. The EU’s risk-based regulatory framework formalizes compliance burdens but enables systematic oversight; the US’s fragmented incentive structures promote defensive AI adoption patterns; and the UK’s polycentric governance approach supports coordinated policy responses. Analysis reveals significant societal risks including algorithmic bias amplification, professional autonomy erosion, and reduced public trust in healthcare systems. Algorithmic normativity governs healthcare by design, necessitating ethical governance to balance innovation with human-centered values. The NORM5 typology and policy toolkit offer actionable pathways for responsible AI governance.</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

Governing by design: algorithmic normativity, clinical standards, and health policy implications of AI in healthcare

  • 期刊论文

  • 发表日期：

    2026年1月28日

查看全文
  • Ali Asadollahi

摘要

To examine how clinical AI systems establish de facto standards of care through design and administrative mechanisms; analyze power redistribution across stakeholders in EU, US, and UK governance regimes; and propose comprehensive policy tools to mitigate unintended normative effects while enhancing equity, accountability, and public trust in healthcare AI systems. A comparative qualitative document analysis was conducted on regulatory texts (e.g., the EU AI Act, FDA guidance), hospital protocols, and vendor materials (2015–2025) from the EU, the US, and the UK. The NORM5 typology (Nudges, Override friction, Responsibility choreography, Metric coupling, Scripted workflows) was developed through systematic inductive coding, informed by political science theories of institutional drift and bureaucratic power. Data collection involved systematic sampling of 127 documents across three jurisdictions, with thematic analysis conducted using established qualitative research protocols. NORM5 mechanisms subtly shift clinical norms through five distinct pathways, redistributing power toward vendors, payers, and healthcare organizations while systematically eroding clinician autonomy. The EU’s risk-based regulatory framework formalizes compliance burdens but enables systematic oversight; the US’s fragmented incentive structures promote defensive AI adoption patterns; and the UK’s polycentric governance approach supports coordinated policy responses. Analysis reveals significant societal risks including algorithmic bias amplification, professional autonomy erosion, and reduced public trust in healthcare systems. Algorithmic normativity governs healthcare by design, necessitating ethical governance to balance innovation with human-centered values. The NORM5 typology and policy toolkit offer actionable pathways for responsible AI governance.