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 Risk Scoring for Tenant Harassment Prevention Outreach

What it is for

Risk Assessment & Triage
  • The THRM scores residential buildings on tenant harassment risk and landlord-related hardship across four dimensions (Construction, Eviction, Financial Distress, Physical Distress) plus an overall score, to prioritize which buildings receive proactive outreach visits.
Eligibility & Public Benefits
  • Buildings scoring highly are selected for proactive door-knocking outreach and potential referral to the HPD Partners in Preservation tenant organizing program, connecting residents with services and support.
Run by
Department of Social Services (DSS)
Where
No fixed location
Kept
Retained Not specified
Shared with
Accountable organization
Your copy
You cannot see the data it holds about you. What you can do

What it is for

Risk Assessment & TriageEligibility & Public Benefits

This AI system scores residential buildings in New York City on their likelihood of tenant harassment or landlord-related hardship. Scores are used by outreach workers to decide which buildings to visit proactively and which to connect to tenant-organizing programs. The system ranks buildings — it does not directly determine any benefit or penalty — but the selection it drives shapes who receives proactive help.

What it collects and what happens to it

Data taken in

Operational data
Anonymized data
  • Building-level operational data including housing code violations, 311 requests and complaints, building financials, and eviction filings — sourced from NYC Open Data, agency partners, and third parties such as Property Shark. All residential buildings are included.
About a place
Anonymized data
  • Residential building identifiers and addresses used to locate and score properties. Output is filtered to buildings with at least 11 units where at least 50% of units are rent-regulated.

Processing

Classification & Prediction
  • The THRM uses multiple linear regressions to estimate component scores and an overall tenant harassment risk score (each 0–100) per building. These are scoring models that predict likelihood of tenant harassment conditions, not classifiers that assign binary labels.

What it does

Deciding (Analytical AI)
Human decides
  • The model combines multiple linear regressions to generate component scores (0–100) for Construction, Eviction, Financial Distress, and Physical Distress, plus an overall harassment risk score. Outreach workers use these scores to decide which buildings to visit; the model is informative, not determinative.

Outputs

A recommendation or prediction
Anonymized data
  • The THRM produces component scores (Construction, Eviction, Financial Distress, Physical Distress, each 0–100) and an overall Tenant Harassment Risk Score (0–100) per qualifying building. PEU uses these scores to prioritize weekly outreach visits; no binding decision is made automatically.
About a place
Anonymized data
  • Scored output is at the building level — residential properties with at least 11 units and at least 50% rent-regulated units. The location (building address) is the unit of analysis used to direct outreach.

Run by

Department of Social Services (DSS)
  • The Public Engagement Unit (PEU) operates within the Department of Social Services and uses the THRM model to identify and conduct proactive tenant outreach, including door knocking and connections to tenant organizing programs.

NYC Algorithmic Tools Compliance Report — THRM

Built by

NYC Department of Housing Preservation and Development
  • HPD developed the Tenant Harassment Risk Model and provides the scored output to PEU for outreach purposes. HPD also administers the Partners in Preservation program that buildings may be referred to.

NYC Algorithmic Tools Compliance Report — THRM

Kept for

Retained Not specified
  • The register does not specify how long model outputs or input data are retained. The model informs weekly outreach starting in October, suggesting outputs are used on an ongoing operational basis, but no retention period is disclosed.
  • Duration: Not specified

Shared with

Available to the accountable organization
  • PEU (Department of Social Services) receives the scored output from HPD. Scores for buildings with at least 11 units and at least 50% rent-regulated units are provided to PEU to inform weekly outreach decisions.
Not available to me
  • The register does not describe a mechanism by which tenants, building owners, or the general public can view, query, or access the score assigned to a specific building. No public-facing portal or individual access mechanism is mentioned.

Stored

Not stated by the NYC Office of Technology and Innovation.

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 Be Informed of AI UseTenants and building owners are not explicitly informed that their building has been scored by the THRM before an outreach worker visits. The register does not describe any notification mechanism. The NYC AI use-policy register is publicly accessible, providing indirect notice.
  • Right to Algorithmic TransparencyThe THRM methodology is partially disclosed in the public AI register: it uses multiple linear regressions combining variables such as violations, 311 complaints, building financials, and eviction filings. Full model weights, thresholds, and complete variable lists are not published. Members of the public may access the register at the source URL.
  • Right to ContestThe register does not describe any formal mechanism by which a building owner, landlord, or tenant can challenge or seek correction of a THRM score. This is a gap in the disclosed governance of the system.

Risks and safeguards

  • Civil liberties harmAlgorithmic scoring of buildings could create stigmatizing profiles of landlords or tenants, or result in targeting of communities based on correlated protected characteristics embedded in historical data (e.g., eviction filings, 311 complaints).Safeguard: The model is informative only — outreach workers make all contact decisions. PEU collaboration with HPD provides a check on model use. Proactive outreach is designed to help tenants, not penalize them or their buildings.
  • Reputational harmA building owner or landlord may be incorrectly scored as high-risk for harassment due to data errors or model limitations, leading to reputational damage from increased regulatory or outreach attention.Safeguard: The register does not specify a formal correction or challenge mechanism for scored buildings or landlords. Scores inform outreach decisions, not public-facing disclosures; however, the absence of an explicit appeals process is a gap.
  • Financial & business harmInput data including building financials and eviction filings is sourced in part from third parties (Property Shark) whose accuracy and coverage may be uneven, potentially skewing scores.Safeguard: The register acknowledges multiple data sources but does not describe data quality validation procedures. Third-party data dependency is a transparency gap.