Socio-Technical Risk Analysis of Intelligent Infrastructure Systems

Abstract

Intelligent infrastructure systems -- AI-managed power grids, autonomous transport networks, AI-integrated water systems, and smart telecommunications backbone -- present socio-technical risks that conventional technical risk analysis cannot fully capture. Socio-technical risk analysis (STRA) treats infrastructure systems as coupled technical-social systems where risks arise not only from technical failures but from the interaction between technical components, human operators, institutional structures, and social contexts that collectively determine system behaviour and failure modes. This paper proposes the Socio-Technical Risk Analysis Framework for Intelligent Infrastructure (STRAIF), a systematic methodology integrating five risk analysis approaches -- STAMP/STPA socio-technical hazard analysis, organisational accident analysis, AI-specific failure mode analysis, community impact risk assessment, and democratic legitimacy risk analysis -- applied to three intelligent infrastructure domains: AI-managed power grids, autonomous urban transport systems, and AI-integrated water management. STRAIF introduces the Socio-Technical Risk Score (STRS) and demonstrates that AI-specific and socio-technical risks are systematically missed by conventional technical risk analysis in all three domains. Key results: democratic legitimacy risk is the most consistently underanalysed risk category; AI-specific failure modes (distribution shift, adversarial attacks, model drift) are absent from current infrastructure risk registers in 84.2% of reviewed deployments; community impact risk is highest for autonomous transport systems. The framework provides STRA methodology guidance for infrastructure operators, safety regulators, and policymakers.