Beyond the Townhall: Spatial Anchoring and LLM Agents for Scalable Participatory Urban Planning

A team from ETH Zurich has developed a novel AI-powered platform designed to revolutionize public participation in urban planning. The system moves beyond static PDFs and text surveys by placing citizens inside a navigable digital twin of their city. During a guided virtual walkthrough of proposed sustainability projects, the platform employs spatial anchoring—a technique linking information to specific locations—to enhance memory and understanding. This immersive experience is powered by two specialized large language model (LLM) agents: a factual assistant that provides source-grounded clarifications about the plans, and a discussion facilitator that prompts reflective conversation.

In a randomized controlled online experiment with 195 participants, the platform was tested against conventional industry methods like static visualizations. The results were striking. The spatially anchored, AI-augmented presentation led to a significant improvement in information recall among participants. More importantly, it fundamentally shifted the nature of their feedback. Instead of focusing on individual inconveniences (like parking or noise), participants' attention was redirected toward collective, community-oriented sustainability benefits. Consequently, they provided significantly more constructive and solution-focused suggestions to the simulated municipality. This research, published on arXiv, establishes a practical, scalable tool for cities to foster more inclusive and informed democratic participation, especially in complex sustainability transitions.