Authors

Robin Kim, Jerry Zhao, Yixuan Hong

Research Question

How can AI support early-stage screening of proposed New York City projects without letting a language model make unsupported planning or regulatory decisions? The project builds an interactive AI and GIS platform that turns a proposed project into a transparent, source-linked and reproducible pre-application screening report.

Background

Planning and environmental-review work draws on regulations, technical manuals, historical documents, maps and numerous public datasets. These resources exist, but connecting them to a specific proposed project is time-consuming and requires specialized knowledge. City Environmental Quality Review (CEQR) applies when qualifying projects involve discretionary approvals, city funding or actions by city agencies, and the required level of review depends on the action and project characteristics.

The platform demonstrates a responsible architecture: AI interprets user requests, deterministic GIS tools perform calculations, version-controlled rules apply screening logic, and human professionals keep responsibility for judgment and final review. It will not issue legal conclusions, determine approval, replace an environmental professional or complete a full environmental review.

The platform can improve access to urban-planning information while creating a foundation for professional software that supports consultants, public agencies, developers, researchers, and community organizations.

Methodology

The project involves:

• Letting users enter a project location and describe a proposal in natural language (for example, a 150-unit mixed-use development near a subway station).
• Converting the description into structured project parameters.
• Retrieving relevant planning guidance and public datasets.
• Running predefined geospatial analyses with deterministic GIS functions and version-controlled rules, keeping language interpretation separate from calculation.
• Identifying review topics that may need professional attention, with each result showing its data source, date, assumptions, calculations and limitations.
• Optionally retrieving similar historical projects from public CEQR records and explaining similarities and differences.
• Focusing in year one on land use, zoning and public policy; community facilities and open space; and potential effects on disadvantaged communities.

Deliverables

• Academic paper
• Data visualizations
• GitHub repository with documentation
• Interactive data dashboard
• Project website
• Technical report
• Web application
• Policy recommendations
• A documented evaluation benchmark will be developed to assess the accuracy, reproducibility, citation quality, tool selection, and appropriate refusal behavior of the AI-assisted geospatial system. The team will also create a prototype review-readiness report that summarizes project context, data sources, spatial analyses, assumptions, and limitations.

Data Sources

• NYC planning guidance, regulations and technical manuals (including CEQR guidance)
• Public geospatial datasets for land use, zoning, community facilities, open space and disadvantaged communities
• Publicly available historical CEQR records