Scholarlink

<p>Municipal governments face mounting pressure to deliver infrastructure services with constrained resources, driving interest in large language models (LLMs) for planning and operational support. Yet empirical evidence reveals a persistent gap between analytical capability and governance readiness: while LLMs can replicate structured reasoning, they frequently produce unverifiable outputs, lack regulatory awareness, and misinterpret local operational constraints. This study presents a hybrid human-AI governance framework designed for decision-making in urban infrastructure and municipal governance settings. Building on a thematic analysis of 78 coded responses from 20 infrastructure professionals and six commercial LLMs across three infrastructure decision scenarios, the study identifies where human and machine reasoning converge and diverge, and translates those empirical patterns into an operational governance model. The framework defines a five-stage workflow supported by a RACI (Responsible, Accountable, Consulted, Informed) matrix that clarifies roles: AI systems perform data synthesis and option generation, while human professionals ensure contextual judgment, ethical oversight, and final authority. Each workflow stage and role assignment is explicitly linked to specific thematic findings from the comparative analysis, ensuring the framework follows from evidence rather than prescriptive assumption. The model addresses algorithmic accountability, data governance, and digital inclusion by embedding oversight, training, and documentation standards into municipal workflows. Grounded in international AI governance frameworks including the EU AI Act, the NIST AI Risk Management Framework, and the OECD AI Principles, the framework offers a replicable model for cities to adopt LLM-assisted decision support in ways that safeguard equity, transparency, and public trust. An 18-month phased implementation roadmap provides municipal agencies with a realistic adoption pathway from readiness assessment through evaluation and scaling.</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

Hybrid human-AI governance framework for accountable decision-making in urban infrastructure management

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月9日

查看全文
  • Alence Poudel,
  • Carla Barrios,
  • Paola De La Torre,
  • Varenya Mehta,
  • Samanata Silwal

摘要

Municipal governments face mounting pressure to deliver infrastructure services with constrained resources, driving interest in large language models (LLMs) for planning and operational support. Yet empirical evidence reveals a persistent gap between analytical capability and governance readiness: while LLMs can replicate structured reasoning, they frequently produce unverifiable outputs, lack regulatory awareness, and misinterpret local operational constraints. This study presents a hybrid human-AI governance framework designed for decision-making in urban infrastructure and municipal governance settings. Building on a thematic analysis of 78 coded responses from 20 infrastructure professionals and six commercial LLMs across three infrastructure decision scenarios, the study identifies where human and machine reasoning converge and diverge, and translates those empirical patterns into an operational governance model. The framework defines a five-stage workflow supported by a RACI (Responsible, Accountable, Consulted, Informed) matrix that clarifies roles: AI systems perform data synthesis and option generation, while human professionals ensure contextual judgment, ethical oversight, and final authority. Each workflow stage and role assignment is explicitly linked to specific thematic findings from the comparative analysis, ensuring the framework follows from evidence rather than prescriptive assumption. The model addresses algorithmic accountability, data governance, and digital inclusion by embedding oversight, training, and documentation standards into municipal workflows. Grounded in international AI governance frameworks including the EU AI Act, the NIST AI Risk Management Framework, and the OECD AI Principles, the framework offers a replicable model for cities to adopt LLM-assisted decision support in ways that safeguard equity, transparency, and public trust. An 18-month phased implementation roadmap provides municipal agencies with a realistic adoption pathway from readiness assessment through evaluation and scaling.