AI Risk Scoring for Child Welfare Prevention Services

What it is for

Risk Assessment & Triage

Predicts the likelihood that a child in an active prevention case will be involved in a future indicated child maltreatment investigation within 24 months of the end of that prevention service. Used by Department of Preventive Services managers to identify low-risk cases and assess program utilization and capacity.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Eligibility & Public Benefits

The model informs internal resource allocation decisions about which prevention service cases may be candidates for earlier closure, indirectly affecting how long children and families remain enrolled in ACS prevention programs.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Run by: Administration for Children's Services (ACS)
Where: No fixed location
Kept: Not stated by the NYC Office of Technology and Innovation.
Shared with: Accountable organization, Not available to vendor
Your copy: You cannot see the data it holds about you. What you can do

What it is for

Risk Assessment & Triage

Eligibility & Public Benefits

This system uses a machine learning model to predict the likelihood that a child currently receiving prevention services will be involved in a future substantiated child maltreatment investigation within 24 months of leaving those services. It generates a ranked list of open prevention cases, ordered from lowest to highest risk, reviewed biweekly by managers in the NYC Administration for Children's Services. The tool is used internally to understand program capacity and service length — it is not shared with service providers or used to make individual case decisions.

What it collects and what happens to it

Data taken in

Sensitive personal information

Pseudonymous data

ACS administrative data about prior and current child welfare involvement for children in active prevention service cases, including New York State Central Register (SCR) investigation records and time spent in foster care. No identifying information is collected or disclosed externally.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Operational data

Anonymized data

ACS internal administrative case records describing program enrollment, service duration, and case status for open prevention service cases. Only ACS administrative data are used — no external datasets.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Processing

Classification & Prediction

A machine learning model trained on historical ACS administrative data (closed investigations, January 2014 – December 2021, approximately 138,712 training observations) that predicts the probability of a future indicated SCR investigation within 24 months of a prevention service ending.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

What it does

Deciding (Analytical AI)

Human decides

The model scores and ranks open prevention cases by predicted likelihood of repeat maltreatment; DPS managers review the ranked output biweekly to make program management and capacity decisions. The model output is advisory — no automated action is taken.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Outputs

A recommendation or prediction

Pseudonymous data

A rank-ordered list of open prevention cases with the lowest predicted likelihood of repeat maltreatment within 24 months, produced biweekly and reviewed by DPS managers to understand program utilization and identify cases with high service length that are low risk. Not shared with service providers.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Run by

Administration for Children's Services (ACS)

The Administration for Children's Services (ACS), Division of Policy, Planning, and Analysis, developed and deploys this model to support internal program management within the Department of Preventive Services.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Built by

Not stated by the NYC Office of Technology and Innovation.

Kept for

Not stated by the NYC Office of Technology and Innovation.

Shared with

Available to the accountable organization

Output reports are available only to ACS Department of Preventive Services managers for internal program assessment. The information is not shared with prevention service providers.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Not available to me

Individuals and families in prevention cases do not have access to the risk scores assigned to their cases. The register states no identifying information is collected or disclosed.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Not available to vendor

The model was developed in-house by ACS with no external vendor. Data is not shared with any third-party technology provider.

NYC Algorithmic Tools Compliance Report — Service Utilization Model

Stored

Stored locally

The model runs on ACS internal administrative systems using only ACS data within New York City government infrastructure. No external cloud provider is mentioned.
Duration: Not specified

NYC Algorithmic Tools Compliance Report — Service Utilization Model

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 TransparencyACS has disclosed the general logic of this tool through the NYC Algorithmic Tools Compliance Report, publicly available at the NYC Open Data portal. The report describes the purpose, training data, input and output data, and autonomy level of the model.
Right to Be Informed of AI UseThe register does not describe any mechanism by which families or children in prevention cases are informed that a predictive model is generating risk scores about their case. The public-facing disclosure exists only through the NYC AI register.

Risks and safeguards

Societal & cultural harmPredictive models in child welfare have documented disparate impacts on families of color, and scoring children for risk of future maltreatment may entrench systemic biases present in historical administrative data used for training.Safeguard: The register designates this as a low-risk, internally supervised tool used only for program-level resource allocation — not for case-level decisions about individual families. Human managers review all outputs; no automated decisions are made. The tool is not shared with providers, limiting downstream misuse.
Loss of autonomyFamilies in prevention services are unaware their cases are being scored and ranked, and the score may influence decisions about service duration without their knowledge or consent.Safeguard: The register states the tool is used only for internal program-level assessment and not for individual case decisions; no identifying information is disclosed. Supervisory human review is required before any program management action is taken.