Urban planners test AI for floods, policy and public input

AI is moving into urban planning as cities explore tools for flood prevention, neighborhood visualization and clearer communication with residents. Northeastern professors Esteban Moro and Ryan Wang analyzed more than 100 studies across urban science, computational social science and geospatial AI to assess how emerging models are being used and where the risks are most significant.

Their review focused on two forms of AI gaining traction in the field: distribution-fitting generative models and foundation models. Distribution generative models can create images and renderings from large datasets, including simulated flood scenarios based on satellite imagery and other visual records. Foundation models, including large language model-based chatbots, are being used to make dense policy documents easier for the public to understand.

One study examined AI agents representing residents from two neighborhoods in Beijing during planning meetings, where the agents raised community concerns and influenced future land development plans. Moro said the approach improved measures such as participant satisfaction and inclusion, showing AI can support community participation rather than simply automate planning decisions.

Researchers cautioned that generative systems can inherit and amplify bias, produce misleading information and create governance risks in high-stakes settings such as disaster response and infrastructure planning. They proposed an ethical framework built around five themes: fairness, privacy, explainability, controllability and community participation, with validation and human review shared by AI providers and urban planners.