GeoMatch: Connecting People to Places Using Artificial Intelligence

“We’ve been using [GeoMatch] for several months now and it’s so nice. It’s able to look into our database and with just a touch of a button, look at case dynamics, as well as seeing the choices available for us to place the case…All of [the] deeper dives that I would have to take for placement, GeoMatch has all in one screen in a matter of minutes…it is giving me back more time to work on arriving cases.”

- Helen Pursel, Assistant Director for Pre-Arrival, Lutheran Immigration and Refugee Service (LIRS), USA

The project in brief

The project is implemented by Immigration Policy Lab - GeoMatch, Church World Service, Dutch Central Agency for the Reception of Asylum Seekers, International Rescue Committee, Lutheran Immigration and Refugee Service, Norwegian Directorate of Integration and Diversity, Swiss State Secretariat for Migration, and U.S. Conference of Catholic Bishops in Switzerland and the United States of America. The pilot launched in 2020. Updates and testing are ongoing.

A team of researchers at the Immigration Policy Lab (IPL) at Stanford University and ETH Zurich developed an AI-powered algorithm that looks at refugees’ background characteristics, resettlement location, and employment outcomes and predicts refugees’ expected employment success. Using this algorithm, the team built a user-friendly placement recommendation tool called GeoMatch, which is in different stages of development, testing, and active implementation across multiple countries.

The aim of GeoMatch is to improve integration outcomes for refugees by empowering governments and immigrants with data-driven insights on the places where newcomers are most likely to thrive.

Main activities of the Good Practice

GeoMatch is a software tool that helps government agencies and nonprofits harness their existing data to identify the communities where refugee families are most likely to thrive. It uses machine learning to build models based on historical data that can incorporate all the factors and synergies that can affect the likelihood that newcomers find success. The tool was developed in partnership with agencies and nonprofits looking for a way to incorporate past outcomes data on current placement decisions. The tool’s design simplifies the existing process of finding appropriate matching communities, improving the administrative workflow for placement case officers. While the current versions of GeoMatch aim to improve employment rates, it is flexible enough to consider other integration outcomes if the data is available. Importantly, GeoMatch is designed as a recommendation tool that provides additional information rather than as an automated-decision making tool.

The tool is currently being used to assist with placements in Switzerland and the United States:

• Switzerland: Since 2020, Swiss State Secretariat for Migration (SEM) has been piloting the tool to help placement officers with the geographic assignment of refugees to Swiss cantons.
• United States of America: Since 2023, Lutheran Immigration and Refugee Service (LIRS) has used GeoMatch to help with placements across its national affiliate network.

Elements which helped facilitate the implementation of the good practice

Close partnerships with governments and resettlement agencies who shared data and acted as co-designers allowed IPL to develop a user-focused tool that has been adapted to fit into the requirements and policy regimes for specific country contexts.

Funding from multiple foundations and institutions enabled the team to establish its initial portfolios, and continue to build capacity and take on additional projects.

“Placing refugees in local communities is one of the most important aspects of our mandate, in addition to offering professional reception and guidance during the asylum process. The community where a refugee is first placed, has an enormous impact on their chances for successful integration. We need scientific insights to help understand what helps determine these successes.”

- Sjef Van Grinsven, Project Leader for Promising Link Using AI, Central Agency for the Reception of Asylum Seekers (COA), Netherlands

Partners involved

• Church World Service
• Dutch Central Agency for the Reception of Asylum Seekers
• International Rescue Committee
• Lutheran Immigration and Refugee Service
• Norwegian Directorate of Integration and Diversity
• Swiss State Secretariat for Migration
• U.S. Conference of Catholic Bishops

What challenges were encountered in delivering the project and how were they overcome?

Challenges

One key challenge has been incorporating existing policies and practices on refugee resettlement and placement for a specific context in order to build those into GeoMatch's algorithmic matching. A second challenge has been identifying the best outcome measure(s) for the GeoMatch tool to target for its recommendations.

How they were overcome

IPL worked closely with staff of the partnering government agencies and nonprofits to identify their existing policies and practices. It used user-focused design strategies to incorporate feedback from the end users to ensure the tool fit into their current workflow and met their requirements.

To address the challenge of selecting outcome measure(s), IPL identified the existing data sources for outcomes, which has varied for different country contexts. In some countries, it is possible to find outcome data in administrative datasets that are available to researchers. In others, the outcomes are collected directly by the partner agencies. In each context, IPL must work closely to understand what data is available to be used by the tool and which integration outcomes its partners consider important measures of success.

Results of the Good Practice

Switzerland: Since 2020, GeoMatch has been used to help placement officers at the Swiss Secretariat of Migration (SEM) with the geographic assignment of over 1000 refugee families to cantons across the country as part of a pilot study. Because the tool is focused on improving three-year employment gains, the outcomes are not yet ready to be analyzed.

USA: Since 2023, staff at Lutheran Immigration and Refugee Service (LIRS) have used GeoMatch to generate placement recommendations for over 1,600 recently-arrived refugees. Staff report that the tool has improved workflows. Data on refugee outcomes is being collected.

In what way does the good practice meet one or more of the four objectives of the Global Compact on Refugees?

Objective 2: Enhance refugee self-reliance

GeoMatch can help match refugees to the communities where they are more likely to find employment. Successful economic integration through employment leads to more self-reliance and financial stability. Ongoing research studies are measuring the tool’s actual impact on employment and wages, but simulations using historical data showed its AI-driven recommendations could increase the chances of employment by 40-70%.

Objective 3: Expand access to third-country solutions

GeoMatch is a tool that can assist governments and organizations in third countries find the locations where refugees are most likely to succeed. The tool provides a user-friendly interface which can help simplify the administrative process of placement. By matching newcomers with places that act as an on-ramp toward integration, countries can unlock their potential to contribute to the economy. When asylum seekers, refugees, and immigrants are employed and integrated into society sooner and in greater numbers, governments have higher tax revenues and lower the costs of social services such as welfare subsidies and unemployment insurance payments.

Next steps

As part of each active implementation, IPL is conducting a randomized-controlled trial (RCT) to measure the effectiveness of the tool. Data for outcomes is still being collected but will eventually be published in future research studies.

There are also plans to expand the tool for use with the Dutch Central Agency for the Reception of Asylum Seekers (COA) in the Netherlands and additional resettlement agencies in the United States.

Are there areas in which support would be required to continue and/or scale up your good practice?

IPL has explored implementing GeoMatch in additional contexts and is interested in offering the tool to more government agencies and organizations.