Zhong-Ren Peng, professor in the UF Department of Urban and Regional Planning and director of the International Center for Adaptation Planning and Design (iAdapt), recently published the article “Symbiotic Planning Theory: The CORE Framework for Human-AI Cocreation in Urban Planning” in the Journal of the American Planning Association (JAPA) alongside a team comprised of his current and former Ph.D. students.
Left to right: Kai-Fa Lu, Yanghe Liu, Dr. Peng, Qing Hou, Qing Zhang and Khalid A. Aljuhani
The paper argues that AI should be treated not as a passive tool or an autonomous system, but as a co-creative partner in planning. While AI can generate options, reveal hidden patterns, and simulate consequences, humans must still set goals, make value judgments, and retain final authority.
The authors propose “Symbiotic Planning Theory (SPT)”, operationalized through the CORE framework—Collaboration, Options, Refinement, and Execution—to address the paper’s key question: How can planners use AI without giving up human judgment, democratic legitimacy, and institutional accountability?
As the paper puts it: AI proposes, humans authorize.
“Cities are increasingly expected to use AI to make better decisions, but the real question is who stays in charge,” said Peng. “Our framework is designed so that AI surfaces hard tradeoffs while humans—and communities—remain the decision authorities. Following rules on paper is not enough. What matters is whether outcomes are actually fair.”
By operating through this CORE framework, the paper offers not only a new planning theory, but also a practical roadmap for how public agencies can use AI responsibly. This framework is supported by the paper’s key mechanism, “Governed Friction”, which argues that conflict between technical optimization and public values should not be hidden or treated as a nuisance. Instead, planning processes should be designed so that AI-generated proposals surface those conflicts, making them visible, debatable, and actionable. The paper describes this as a three-step cycle:
• Generative Provocation: AI generates an efficiency-focused baseline that exposes hidden bias or inequity.
• Normative Recalibration: planners and communities translate those conflicts into formal, binding constraints.
• Binding Re-authorization: continued model use depends on whether outcomes meet equity and accountability standards, not merely whether procedures were followed.
The CORE framework as put forth in “Symbiotic Planning Theory: The CORE Framework for Human-AI CoCreation in Urban Planning“
Using Gainesville’s shared e-scooter program as a case study, the paper examined the gap between procedural compliance and actual equity outcomes in designated Equity Zones. The city requires e-scooter vendors to deploy at least 10% of their fleet in designated Equity Zones.
On paper, vendors were largely compliant with the city’s 10% Equity Zone deployment requirement. But the outcome data showed a clear gap between procedural compliance and substantive outcomes:
• The Equity Zone generated only ~3% of monthly trips
• The area around campus generated ~84% of trips
• Scooters in Equity Zones often sat unused for 75–100+ hours
This is what the paper calls the Compliance Trap: a situation in which a city follows the rules on paper but still produces inequitable outcomes in practice.
To investigate the problem, the research team first used a Deep Reinforcement Learning (DRL) model to generate a usage-maximizing baseline. That baseline concentrated scooters near campus and reproduced existing spatial inequities. Rather than treating that result as the answer, the paper treats it as a provocation — a way of making the hidden tradeoffs and inequities visible.
SPT in practice as depicted in the case study on Gainesville’s shared e-scooter program from “Symbiotic Planning Theory: The CORE Framework for Human-AI CoCreation in Urban Planning“
The team then conducted community engagement, including 235 survey responses, interviews, and workshops. They found that low usage in Equity Zones was linked to several structural barriers:
• Weak walking access to scooter locations
• Competition from free microtransit
• Payment exclusion for people without bank accounts
• Cost barriers relative to free alternatives
These findings were then translated into binding constraints for the model, which was retrained under revised guardrails set through planner and community input. The result shifted the plan from 103 locations / 189 scooters to 113 locations / 187 scooters, with improved Equity Zone coverage and stronger transit integration.
This case study showed how AI can be governed not only for efficiency, but also for equity, democratic accountability, and continued human oversight. It also suggests a practical pathway that may help cities like Gainesville move from monitoring problems to acting on them through more explicit governance protocols.
This paper offers both a theoretical and practical framework for human-AI collaboration, showing how AI can be integrated into planning workflows without giving up human authority, public accountability, or equity goals . Notably, the paper positions Symbiotic Planning Theory alongside rational, participatory, and co-design paradigms, arguing that those earlier theories were not developed for an era in which AI can generate alternatives at machine speed and reshape how planning problems are understood.