Abstract
Effective participatory decision-making in urban land use planning requires sustained multi-stakeholder collaboration, yet current practices fragment stakeholders across planning phases, resulting in inequitable participation and decisions disconnected from diverse needs. To address communication opacity, unstructured conflict management, and power imbalances, we introduce PlanningConnect, a web-based multi-agent framework with role-specific interfaces facilitating AI-mediated participatory planning. The system implements six specialized agents—dialogue structuring, conflict detection, feasibility validation, and perspective synthesis—operating across customized interfaces for planning experts, decision-makers, and public representatives, while preserving human authority throughout. We validated PlanningConnect through formative study (N=12) and controlled evaluation (N=10, within-subject). Results demonstrate public speaking time doubled (15% → 32%), idea adoption tripled (20% → 60%), conflict resolution improved five-fold (17% → 88%, d=1.8, p<.01). This work contributes a lifecycle-integrated framework and multi-stakeholder interface toolkit with coordinated multi-agent backend, demonstrating multi-agent systems can transform inequitable participatory planning into inclusive collaboration.