Risk scan of Negligent Unemployment

Employee Insurance Agency

Helps us spot possible culpable unemployment.

Last change on 20th of January 2025, at 12:00 (CET) | Publication Standard 1.0

• Publication category: Impactful algorithms
• Impact assessment: DPIA
• Status: In use

General information

Theme

• Social Security
• Work

Begin date

2022-08

Contact information

https://uwv.nl/nl/service-en-contact/contact-uwv

Link to source registration

https://www.uwv.nl/nl/over-uwv/organisatie/algoritmeregister-uwv/risicoscan-verwijtbare-werkloosheid

Responsible use

Goal and impact

When you apply for WW benefits from us, we want to know that you did not become unemployed through no fault of your own. If you did, you are culpably unemployed. You will then not be entitled to WW benefit. To identify possible culpable unemployment in WW applications, we use the Risk Scan Culpable Unemployment.

Considerations

The risk scan was developed to better identify culpable unemployment without having to check every file. This allows us to focus more on customer focus rather than control.

YouTube video: How UWV works with Risk Scans: https://youtu.be/g4qY5mnCfBY

Human intervention

Our staff monitor the use of the algorithm in the following ways:

• The employee himself assesses the WW application and makes his own decision on the possible next steps. The risk scan therefore does not make any decisions about benefit entitlement.

• Every month, specialised UWV employees check the quality of the data used by the risk scan. The proper functioning of the scan is also monitored. In this way, we constantly check whether the scan continues to meet our quality standards. For example, we check whether the population used in developing the risk scan is still representative.

• The risk scan is regularly developed further so that it only uses data that adds value to its operation. We remove data that does not (or no longer) add value. We also regularly check whether there are still opportunities to improve the scan.

Risk management

We ensure that we remain compliant with information security and privacy requirements. We do this in the following ways:

• We constantly check data quality.

• We always ensure that employees do the final assessment and not the algorithm.

• If a scan is (re)developed by us, the quality of our work is reviewed by an independent, reputable organisation. This way, we reduce the chances of errors or sub-optimal quality.

• We use data provided by the client as much as possible.

With the following measures, we ensure the correct use of the Risk Scan Blamable Unemployment:

• The algorithm always uses the same data and behavioural characteristics.

• The risk scan does not use personal characteristics such as origin, gender, age or other privacy-sensitive data.

• The risk scan does not use data from social media or other public sources. This way, we treat everyone the same and human biases do not affect treatment.

• The risk scan produces a selection of signals that staff investigate further. A number (30%) of randomly selected WW applications are always added to this selection. So the employee does not know whether a signal comes from the risk scan or from the applications added at random. In this way, we ensure that employees are not influenced and remain critical of situations they have to investigate.

• The algorithm only signals. An employee assesses whether there is actual culpable unemployment or not.

• Only the applications that pass the risk scan (including the 30% randomly selected WW applications added to it) are additionally checked and examined by our employees.

Elaboration on impact assessments

Instead of the national IAMA standard, a UWV Ethical Impact Assessment was conducted.

Impact assessment

Data Protection Impact Assessment (DPIA)

Operations

Data

The risk scan uses data needed to see whether someone applying for WW benefits may be at an increased risk of culpable unemployment. This data is mainly about someone's behaviour. We therefore call them behavioural characteristics.

We do not disclose all the data and behavioural characteristics used by the risk scan. Otherwise, the outcome of the risk scan may be affected.

The risk scan uses 3 types of data:

• data about (possible) previous WW applications and benefits)

(e.g. how often you have received unemployment benefit, on what date it started(s) and whether you were previously culpably unemployed)

• details of your employment history

(e.g. how long your longest employment lasted)

• details of your current WW claim

(e.g. when you made the application)

Technical design

The risk scan determines the risk of culpable unemployment based on a combination of several data. This is therefore never done on the basis of a single behavioural characteristic or data. High-risk applications are presented to our staff for extensive examination.

It is important to know that the risk scan only signals. A signal does not indicate whether or not someone is culpably unemployed, but is a trigger for staff to carry out a check. Our employees themselves determine whether there is culpable unemployment or not.