Accountability as Infrastructure: A Design Theory of AI-Enabled Planning Authority

Problem, research strategy, and findings Urban planning agencies are adopting artificial intelligence (AI) faster than they are developing procedures for governing its continued use. In this study I developed and diagnostically applied a design theory for keeping AI-supported planning authority publicly inspectable, challengeable, and withdrawable after deployment. The research went through four stages: identifying recurrent institutional failures documented in planning and public sector AI scholarship; synthesizing six established accountability tools into an integrated institutional design; applying that design diagnostically to Gainesville’s (FL) micromobility program; and deriving six propositions with observable measures and comparative research designs. In this analysis I identified three recurrent risks: unauthorized expansion of system use, weakening oversight, and persistent distributional underperformance. Gainesville shows how compliance with an equity-zone deployment requirement coexisted with unequal use when monitoring was not connected to binding review or corrective action. I did not estimate causal effects or claim that the proposed mechanisms are necessary or sufficient across all settings. This study’s research contribution is a theoretically grounded and empirically traceable institutional design that can be tested, refined, or rejected through comparative planning research.Takeaway for practice Planning agencies can adapt the framework through risk-tiered requirements: Routine applications might require brief records and manager-level review, whereas consequential or distributionally significant uses should receive more extensive scrutiny. For consequential uses, agencies should document the evidence supporting authorization, assign responsibility for equity review, publish accessible accountability information, specify thresholds for reconsideration, and require periodic evidence-based renewal. Procurement terms and vendor contracts can support these functions by preserving access, audit, change-control, and termination rights.