Child welfare predictive analytics systems

Dear Mr. Feathers:

The New Jersey Department of Children and Families (NJDCF) has received your request as follows: "Pursuant to the New Jersey Open Public Records Act, I hereby request the following records:

1) Contracts with any company, nonprofit, or other entity that currently provides software to the department that includes predictive analytics models for child welfare (i.e., models designed to identify home placement risk, abuse risk, human trafficking risk, etc.). This should include contracts for CCWIS or comparable systems that include predictive models as well as project descriptions for any models created in-house.

2) Data dictionaries for all child welfare predictive models used by the department, including those provided by outside entities as well as in-house models. Data dictionaries are documents commonly created for predictive models and databases. They describe the categories of data that are inputs (age, gender, etc.) and outputs (risk level, etc.) of predictive models. They are equivalent to headers on a spreadsheet, and do not include actual data.

3) A list of all memorandums of understanding (MOUs) and/or data sharing agreements with state or local agencies for the provision of data that is incorporated into the department's child welfare predictive models. Should the department not maintain a list of these MOUs, I will accept copies of the MOUs themselves.

4) Any differential impact assessments, validation studies, or comparable documents that describe the accuracy of child welfare predictive models used by the department. This should include any documents that describe common measures of predictive accuracy, such as area under the curve (AUC).

5) All department policies, employee manuals, and guidelines describing how employees can and/or should use child welfare predictive models.

6) All governance board, committee, or outside expert reviews assessing the efficacy of the department's proposed or current child welfare predictive models.

Should you have questions regarding any aspect of this request, please don't hesitate to contact me.

I ask that all fees be waived as I am a journalist and intend to use the requested records to publish articles in the public interest. In the event you choose to impose fees, I request a detailed breakdown of the fees, including the hourly wage of each employ involved and an explanation justifying the employee hours required to fulfill the request.

Should you choose to reject this request or redact portions of it, I ask that you provide a detailed breakdown of the statutory exemptions and associated case law underlying your decision to withhold each/any portions from public review. I would prefer the request be filled electronically, by e-mail attachment if available or CD-ROM if not.

Thank you very much for your help, and I look forward to receiving your response to this request within 7 business days, as the statute requires. Sincerely, Todd Feathers"

Please be advised that our response follows:

1) Contracts with any company, nonprofit, or other entity that currently provides software to the department that includes predictive analytics models for child welfare (i.e., models designed to identify home placement risk, abuse risk, human trafficking risk, etc.). This should include contracts for CCWIS or comparable systems that include predictive models as well as project descriptions for any models created in-house.

DCF does not have a contract with a company, nonprofit or other entity that provides software to the department that includes predictive analytic models

2) Data dictionaries for all child welfare predictive models used by the department, including those provided by outside entities as well as in-house models. Data dictionaries are documents commonly created for predictive models and databases. They describe the categories of data that are inputs (age, gender, etc.) and outputs (risk level, etc.) of predictive models. They are equivalent to headers on a spreadsheet, and do not include actual data.

What do we have here re PAP?

Please see the list of the categories of data to be used in each stage of the modeling from PAP attached above.

3) A list of all memorandums of understanding (MOUs) and/or data sharing agreements with state or local agencies for the provision of data that is incorporated into the department's child welfare predictive models. Should the department not maintain a list of these MOUs, I will accept copies of the MOUs themselves.

Please see the DUA with TACC attached above.

4) Any differential impact assessments, validation studies, or comparable documents that describe the accuracy of child welfare predictive models used by the department. This should include any documents that describe common measures of predictive accuracy, such as area under the curve (AUC).
We have provided a link to the PAP website: https://www.predict-align-prevent.org/resources and have attached language in the generic PAP protocol around validation of the model, please see page 3).

5) All department policies, employee manuals, and guidelines describing how employees can and/or should use child welfare predictive models.

DCF has no responsive records.

6) All governance board, committee, or outside expert reviews assessing the efficacy of the department's proposed or current child welfare predictive models.

DCF has no responsive records.

Attached is a receipt closing your request. Thank you for your interest in NJDCF.

Sybil R. Trotta, Esq.
Legal Specialist
Department of Children and Families
Office of Legal Affairs
P.O. Box 717
50 East State Street
Trenton, New Jersey 08625
(609) 888-7814 (office)
(201) 463-1016 (cell)
(609) 292-0731 (fax)
Sybil.Trotta@dcf.nj.gov<mailto:Sybil.Trotta@dcf.nj.gov>

This E-mail, including any attachments, may be intended solely for the personal and confidential use of the sender and recipient(s) named above. This message may include advisory, consultative and/or deliberative material and, as such, would be privileged and confidential and not a public document. If you have received this e-mail in error, you must not review, transmit, convert to hard copy, copy, use or disseminate this e-mail or any attachments to it and you must delete this message. You are requested to notify the sender by return e-mail.