When trust becomes symbolic: standards and the governance of AI
基于卢曼系统理论，重新概念化标准为生成性治理安排，通过英国AI标准案例，分析标准如何组织期望、信任和决策，将信任视为组织产出而非AI系统固有属性。
Purpose This paper aims to reconceptualise standardisation as a generative governance arrangement that organises expectations, trust and decision-making under persistent uncertainty. Drawing on Niklas Luhmann's systems theory and extending Kim's (2025a) concept of standards as thematic identities and expectational structures, it seeks to explain how standards operate beyond technical coordination. Focusing on the UK's AI standards strategy, the paper examines how standards and strategy function as programmes at semantic and structural levels. In doing so, it explores how AI assurance practices are shaped and trust recast as an organisationally produced outcome rather than an inherent property of AI systems. Design/methodology/approach The study adopts a qualitative, theory-driven analytical approach grounded in systems theory. It combines conceptual reconstruction with an in-depth case analysis of the UK AI standards and assurance ecosystem. Primary materials include UK policy documents, standards strategies, assurance roadmaps and regulatory guidance, complemented by secondary academic literature on standardisation and AI governance. These materials are analysed through Luhmannian concepts such as second-order observation, decision premises and programmes, enabling a semantic and structural analysis of how standards are institutionalised and operationalised within contemporary AI governance. Findings The paper finds that AI standards in the UK operate as multi-layered observational procedures that institutionalise trust by stabilising expectations rather than guaranteeing technological reliability. Standards and strategy function as interlinked programmes: semantically, they construct thematic identities such as “trusted AI”; structurally, they embed decision premises across assurance practices. The analysis shows that standards oscillate between dissemination media and symbolically generalised communication media, enabling coordination through simplified symbols like certification. This oscillation produces ambivalent effects, simultaneously facilitating governance while fostering strategic ignorance and the ritualisation of AI ethics. Originality/value This paper makes an original contribution by systematically linking Niklas Luhmann's core concepts – second-order observation, decision premises and programmes – to the standards and standardisation literature. Extending Kim's (2025a) concept of standards as thematic identities and expectational structures, it reconceptualises AI assurance as a multi-layered observational procedure in which trust is produced through recursive processes of decision and validation rather than embedded in AI systems. By theorising standards as mediating forms that oscillate between dissemination media and symbolically generalised communication media, the paper offers a novel Luhmannian explanation of how standards stabilise expectations while generating ambivalence, strategic ignorance and the ritualisation of ethics in contemporary AI governance.