Skip to content
This is NOT an official site of the City of New York. Click here for official AI reports.

NYC Office of Technology and Innovation

NYC - DTPR AI Register

Back to your search

AI-Assisted Summarization of Community Planning Input

What it is for

Planning & Decision-making
  • Supports neighborhood planning by aggregating, tagging, and summarizing community input so that residents can see how their contributions shaped planning and policy decisions for the Jamaica and White Plains Road Neighborhood Plans.
Inform
  • Publishes processed community comments online, enabling members of the public to explore how their input was categorized, geolocated, and incorporated into neighborhood plans.
Run by
Department of City Planning (DCP)
Where
No fixed location
Kept
Retained Not specified in the register
Shared with
Accountable organization, Download
Your copy
You cannot see the data it holds about you. What you can do

What it is for

Planning & Decision-makingInform

Voice to Vision uses large language models to tag, geolocate, and summarize public comments collected during neighborhood planning sessions in New York City. The processed input is published online so community members can see how their feedback shaped planning decisions. No personally identifying information is collected or disclosed. All AI-generated tags, summaries, and citations are reviewed by city staff before publication.

What it collects and what happens to it

Data taken in

About behaviour
Anonymized data
  • Public comments submitted during neighborhood planning engagement sessions, including spoken or written statements about community needs, concerns, and priorities related to specific plans such as Jamaica and White Plains Road.

Processing

Language Models
  • GPT-4.1-mini and GPT-5-mini tag themes and geolocate mentions in community comments; Claude 3.5 Sonnet drafts topic-area summaries with citations. Training data is proprietary to the respective vendors.

What it does

Deciding (Analytical AI)
Human decides
  • GPT-4.1-mini and GPT-5-mini classify community comments by theme and tag mentioned locations; Claude 3.5 Sonnet scores and surfaces common topic patterns. All outputs are reviewed by staff before publication.
Creating (Generative AI)
Human decides
  • Claude 3.5 Sonnet drafts summaries of common topic areas with citations back to original community comments. Summaries are reviewed and approved by staff before being published.
Understanding (Semantic AI)
Human decides
  • The system extracts meaning from community comments to identify locations and thematic tags, and retrieves relevant source comments to support generated summaries with citations.

Outputs

About a place
Anonymized data
  • Potential locations mentioned in community comments are identified and geotagged, producing a spatial visualization of where residents raised issues in the neighborhood planning area.
Generated content
Anonymized data
  • AI-generated summaries of common topic areas drawn from community comments, each with citations back to original public statements. Published online after staff review.
About behaviour
Anonymized data
  • Themes and goals tagged from community comments, representing patterns in what residents said during planning engagement sessions, published as thematic categories on the public-facing visualization.

Run by

Department of City Planning (DCP)
  • The NYC Department of City Planning deploys Voice to Vision to tag, summarize, and publicly visualize community input gathered during neighborhood planning processes.

NYC AI Use Cases — Voice to Vision

Built by

Massachusetts Institute of Technology (MIT)
  • MIT's Center for Constructive Communication developed Voice to Vision as a research project in coordination with DCP, provided at no cost.

NYC AI Use Cases — Voice to Vision

Kept for

Retained Not specified in the register
  • The register does not specify a retention period for community comments or AI outputs. Processed data appears to be retained at least for the duration of the active neighborhood plan publication.
  • Duration: Not specified in the register

Shared with

Available to the accountable organization
  • Processed tags, geotags, and summaries are available to DCP staff, who review all AI outputs before publication.
Available to download
  • Processed community input is published and viewable online at public-facing plan websites (e.g., v2v-jamaicaplan.ccc-mit.org), accessible by any member of the public.
Not available to me
  • No identifying information is collected or disclosed, so individual community members cannot retrieve a personal record — only aggregated and de-identified thematic outputs are published.

Stored

Stored on 3rd Party Cloud
  • Outputs are hosted on MIT Center for Constructive Communication servers and publicly accessible via the plan websites (e.g., v2v-jamaicaplan.ccc-mit.org). Processing relies on third-party LLM APIs (OpenAI, Anthropic).
  • Duration: Not specified in the register
How to read the colours

Can it identify you?

Anonymized data
Data about people with the link to who is broken. Stripped of identifiers, blurred, aggregated, or noised so this system can’t reasonably tie a record back to an individual.
Pseudonymous data
Each person’s data is tied to a token (hash, ID, template) that lets this system recognise the same person across events, but the token itself doesn’t reveal a name. Reidentification is possible with extra information.
Identifiable data
The data either contains a direct identifier (name, address, account name, recognisable face or voice, plate number) or carries a token this system uses to look up legal identity during processing.

Who completes the loop?

Human decides
This mode suggests; a person decides what to do next. The AI is always advisory — a human is in the loop on every decision. Example: a triage tool ranks cases for a clinician who chooses which to see first.
Human executes
This mode decides; a person carries out the result. Example: an optimizer plans the day’s trash-collection routes, and drivers run them.
Autonomous
This mode decides and acts on its own. No person reviews each decision or carries out the resulting action.

Definitions from the DTPR standard. Amber is about your data, violet about who decides. The fuller the shape and the deeper the colour, the more identifying the data or the less a person is involved.

What you can do

Ask about this system

Questions go to the NYC Office of Technology and Innovation, not the vendor.

Your rights

  • Right to Algorithmic TransparencyMembers of the public can learn that GPT-4.1-mini, GPT-5-mini (OpenAI), and Claude 3.5 Sonnet (Anthropic) are used to tag themes, geolocate mentions, and draft summaries. The system description is published in the NYC AI Use Cases register and in academic documentation at https://dl.acm.org/doi/10.1145/3715070.3757228.
  • Right to Be Informed of AI UseCommunity members who participate in neighborhood planning sessions should be informed that AI tools are used to process and summarize their comments before publication. The public-facing websites and the NYC AI register disclose this use.

Risks and safeguards

  • Societal & cultural harmAI summarization and thematic tagging may misrepresent community voices, flatten nuanced input, or introduce cultural or linguistic bias when processing comments from diverse neighborhoods.Safeguard: All AI-generated tags, summaries, and citations are reviewed by DCP staff prior to publication to catch errors and misrepresentations.
  • Reputational harmIncorrect geolocation or thematic tagging could publicly misattribute positions or priorities to communities or locations.Safeguard: Staff review all AI outputs before publication; citation links back to original comments allow verification.