Google unemployment fraud docs

Todd Feathers filed this request with the Wisconsin Department Of Workforce Development of Wisconsin.
Status
Completed

Communications

From: Todd Feathers

To Whom It May Concern:

Pursuant to the Wisconsin Open Records Act, I hereby request the following records:

1) All contracts and statements of work (including addendums) between the department and Google (and/or Google's re-seller) regarding Google's provision of an unemployment fraud detection system. This partnership has been publicly announced: https://dwd.wisconsin.gov/press/2020/201019-cloud-collaboration.htm.

2) Documentation of the predictive model(s) built by Google for unemployment fraud detection. This should include, where possible, a list of the variables the model considers, how those variables are weighed by the model, explanations of the machine learning techniques employed, the results of accuracy validation tests, and the results of disparate impact tests.

3) A de-identified data set containing all unemployment claims flagged as potentially improper/fraudulent by the Google predictive model(s) from the time the system was implemented until the present date. Please note, that I am requesting this dataset not contain data fields that are personally identifiable (e.g., name, street address, SSN).

4) A de-identified data set containing all unemployment claims NOT flagged as potentially improper/fraudulent by the Google predictive model(s) from the time the system was implemented until the present date. Please note, that I am requesting this dataset not contain data fields that are personally identifiable (e.g., name, street address, SSN).

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 employee 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.

The requested documents will be made available to the general public, and this request is not being made for commercial purposes.

In the event that there are fees, I would be grateful if you would inform me of the total charges in advance of fulfilling my request. I would prefer the request filled electronically, by e-mail attachment if available or CD-ROM if not.

Thank you in advance for your anticipated cooperation in this matter. I look forward to receiving your response to this request within 10 business days.

Sincerely,

Todd Feathers

From:

We have received your open records request and have begun processing your request. We will respond to your request as soon as practicable. Certain fees may apply to your request.

Details about the Department's open records process, including fee information, is available at this website - https://dwd.wisconsin.gov/dwd/publications/dwd/sec-6895-p.htm.

Thank you for contacting the Wisconsin Department of Workforce Development.

From: Wisconsin Department Of Workforce Development

Hello Jeff -

Is there a phone number where you could be reached? I would like to discuss your request and it would be best to do so verbally so we can move forward as quickly as possible.

Thanks!

Tyler L. Tichenor
Communications Specialist, Adv.
DWD Communications Office
Office: 608-267-4780
Cell: 608-630-6391
201 E. Washington Avenue
Madison, WI 53703

Twitter: @WIWorkforce<https://twitter.com/WIWorkforce>
Facebook: /WIWorkforce<https://www.facebook.com/WIWorkforce>
YouTube: /wiworkforce<https://www.youtube.com/user/wiworkforce>

From: Todd Feathers

Hello Tyler,

Just wanted to check if this message was sent in relation to the right request. It's addressed to Jeff, but my name is Todd. These mixups happen all the time with MuckRock addresses, so just wanted to see if you indeed wanted to talk about my request for documents related to Google's unemployment fraud detection contract, etc.

Thanks!
Todd

From: Wisconsin Department Of Workforce Development

Yes, that is the right request. Sorry, the email information was correct, i confused names from 2 requests.

Let me know a number and time when we can connect. Im deer hunting next week and not in work status but would be happy to connect between 11am and 1pm central time any day.

Appreciate your response

From: Todd Feathers

Hi Tyler,

Excellent. Sent you an email from my work account with my cell.

Todd

From: Wisconsin Department Of Workforce Development

Hi Todd -

Let me know a good time for a phone call. I am available from 10am -3pm.

TT

From: Todd Feathers

Hi I'm available any time after 11 a.m. ET today.

Thanks!

From: Wisconsin Department Of Workforce Development

Sorry, Todd. I was out for a few weeks around the holidays and then had a COVID situation to deal with. Are you available at 1pm Central Time Monday? IF so please send that phone number through again or you can call me at 608-333-6591

From: Todd Feathers

Unfortunately I have a meeting at 1 CT Monday. Any other times that work for you?

Todd

From: Wisconsin Department Of Workforce Development

Hey Todd -

Tried calling you yesterday at 12:30pm CT and didn't get an answer. I am pretty booked up today but could do a call tomorrow. Just let me know a time that works for you.

From: Todd Feathers

9-11 CT tomorrow or noon to 1 CT tomorrow work for me!

From: Wisconsin Department Of Workforce Development

Mr. Feathers:

This email is in response to your November 3, 2021 request for the following:

1. All contracts and statements of work (including addendums) between the department and Google (and/or Google's re-seller) regarding Google's provision of an unemployment fraud detection system. This partnership has been publicly announced: https://dwd.wisconsin.gov/press/2020/201019-cloud-collaboration.htm.

1. Documentation of the predictive model(s) built by Google for unemployment fraud detection. This should include, where possible, a list of the variables the model considers, how those variables are weighed by the model, explanations of the machine learning techniques employed, the results of accuracy validation tests, and the results of disparate impact tests.

1. A de-identified data set containing all unemployment claims flagged as potentially improper/fraudulent by the Google predictive model(s) from the time the system was implemented until the present date. Please note, that I am requesting this dataset not contain data fields that are personally identifiable (e.g., name, street address, SSN).

1. A de-identified data set containing all unemployment claims NOT flagged as potentially improper/fraudulent by the Google predictive model(s) from the time the system was implemented until the present date. Please note, that I am requesting this dataset not contain data fields that are personally identifiable (e.g., name, street address, SSN).

The Department's response to each of your requests is below.

Contracts and Statements of Work (Request 1)

The records responsive to your request are available for download at the following link. The link will be active until March 26, 2022.

https://www.dropbox.com/s/ddhspeix11ljvx5/Feathers%2C%20Todd%2011.03.2021.pdf?dl=0

Remaining Requests (Requests 2-4)

The Department has identified some records responsive to your remaining requests. However, pursuant to Wis. Stat. § 19.35(1)(a), and applying the balancing test, the Department has determined that the public interest in nondisclosure outweighs the general presumption favoring release. See Woznicki v. Erickson, 202 Wis. 2d 178, 183-84, 549 N.W.2d 699 (1996). In conducting the balancing test, the Department considered the strong public interest in disclosure and the public interest in protecting the integrity of the Unemployment Insurance (UI) Division's fraud detection efforts. As mentioned in the press release cited in your request, DWD worked with Google to use predictive analytics to shorten adjudication decision-making. In doing this, DWD used a rules-based analytics engine to identify claims with a low risk of improper payment, including using fraud detection and identification service scans to identify potentially fraudulent claims based on key indicators. Sharing information about how this predictive model worked, including a list of the variables the model considers, how those variables are weighed by the model, explanations of the machine learning techniques employed, the results of accuracy validation tests, and the results of disparate impact tests, would give the public enormous insight into the methods and indicators that UI uses to detect fraud, which in turn could be used to avoid or thwart UI's fraud detection methods. Similarly, de-identified lists of flagged and unflagged cases could be used to gain similar insights into UI's fraud detection methods. Because release of these records could ultimately be used to circumvent the law, the Department determined that the public interest in preventing the release of predictive model documentation and de-identified lists of flagged and unflagged claims outweighs the public interest in disclosure of those records. See Democratic Party of Wisconsin v. Wisconsin Dep't of Justice, 2016 WI 100, ¶ 24, 372 Wis. 2d 460, 888 N.W.2d 584, 593.

Pursuant to Wis. Stat. § 19.35(4)(b), this determination is subject to review by mandamus under Wis. Stat. § 19.37(1) or upon application to a district attorney or the Attorney General.

Thanks!

Tyler L. Tichenor
Communications Specialist, Adv.
DWD Communications Office
Office: 608-267-4780
Cell: 608-630-6391
201 E. Washington Avenue
Madison, WI 53703
[new DWD logo]
Twitter: @WIWorkforce<https://twitter.com/WIWorkforce>
Facebook: /WIWorkforce<https://www.facebook.com/WIWorkforce>
YouTube: /wiworkforce<https://www.youtube.com/user/wiworkforce>

Files

pages

Close