The Glass Box Planner: From Deductive Policy Audits To a Normative Framework for Urban AI

This paper makes two contributions. First, it bridges the land use analysis gap by replacing manual methods with a scalable, open-source engine implementing a transparent 'policy-as-code' approach. We applied the Compatibility Audit Tool (CAT) to Qazvin, Iran, analyzing over 65,000 land parcels and revealing that a critical 2.04% of the urban fabric—concentrated at residential-industrial interfaces—was in direct policy conflict. The framework provides planners with a robust instrument for a systematic 'policy audit' to identify contradictions between policy and reality. Second, it proposes a normative framework for urban AI, shifting from optimization-focused models toward forensic instruments that enforce accountability by quantifying the divergence between stated policy and spatial reality. It transforms the planning audit from a bureaucratic formality into a mechanism for liability discovery.