Promising Link with GeoMatch

Central Agency for the Reception of Asylum Seekers

The most ideal (relocation/)labour market region with greatest chance of paid work for the asylum seeker

Last change on 26th of August 2024, at 9:41 (CET) | Publication Standard 1.0

• Publication category: Impactful algorithms
• Impact assessment: DPIA, IAMA
• Status: In development

General information

Theme

Migration and Integration

Begin date

2024-08

Contact information

Voor vragen kunt u contact opnemen met de COA Informatielijn: 088 - 715 70 00

Link to publication website

[e-mail]

Responsible use

Goal and impact

The algorithm provides advice on where in the Netherlands a status holder has the greatest chance of finding a job. A COA employee uses this advice to determine which labour market region a status holder will be linked to. With the algorithm, we want to help status holders find a (suitable) job faster. With a job, they can settle in faster in the Netherlands.

Considerations

An algorithm can use and incorporate much more data into an opinion in a shorter time than a human (in this case, a COA employee). Such as data from the Central Bureau of Statistics (CBS). This data is about whether previous status holders have found a job. The algorithm can thus ensure that status holders move out to a region where they are most likely to find paid work. Having paid work contributes to faster and better integration in the Netherlands.

Human intervention

We use an algorithm to determine where status holders can best live to find suitable work. The algorithm analyses status holders' data, such as their origin and their work experience and education. In addition, the algorithm looks at where people with similar characteristics have successfully found work in the past. The algorithm then advises on the three most suitable regions for them. A COA employee reviews that advice and also takes other factors into account. For example, where the status holder has a social network and where he/she/they themselves would like to live. The COA employee decides in which labour market region the status holder will live. If the COA employee chooses something else than the algorithm advises, we note the reason for this. This allows us to improve the algorithm in the future.

Risk management

With a monitoring plan, we check every month whether the algorithm is working as intended. For example, we look at the quality and speed of the recommendations, how many of the recommendations are followed and whether there are any unintended effects, such as discrimination. By looking at this every month, we as COA can quickly intervene if the algorithm does not work as intended. The moment CBS publishes new data, the algorithm is adapted to the latest trends in inflow and employment.

Impact assessment

• Data Protection Impact Assessment (DPIA)
• Imact Assessment Mensenrechten en Algoritmes (IAMA): AI Impact Assessment (AIIA) op basis van Toetsingskader Algoritmes van de Algemene Rekenkamer

Operations

Data

The following categories of data are used for this algorithm:

- Data about the person such as gender, marital status, date of birth, country of origin and mother tongue

- Data on education and work experience

- Data on reception status

- Historical data from CBS on whether status holders have found a job and what income they have.

We regularly check that the algorithm does not bias or discriminate improperly. For this, we use data on ethnicity and religion. These are not used to develop the algorithm, but only for verification.

Technical design

We are working with the algorithm to increase the probability of a status holder finding work within a year of moving into housing in the municipality. Here, the two targets are the probability of finding a job and the level of taxable income. For each labour market region, using supervised learning, a model is developed of which characteristics of an individual determine success on the target variables. It is then determined per family or family unit which labour market region gives the best result.

External provider

Immigration Policy Lab - Stanford University and ETH Zurich