Around the world, wherever governments have received rising numbers of refugees and asylum seekers in recent years, policymakers have wrestled with how to successfully integrate newcomers into host societies. There’s an emerging consensus that place matters: whether or not refugees thrive may hinge on where they are resettled within the host country. A good match between person and place can be the difference between success and failure. But how do you make a good match? The exact formula might be impossible to pin down: the ways refugees’ personal characteristics interact with local conditions are intricate and nearly endless.
This is where artificial intelligence comes in. With an algorithm to find patterns in vast troves of data, governments and resettlement agencies can make the best matches possible for all resettled refugees.
Researchers at the Immigration Policy Lab (IPL) created this matching algorithm tool to improve the way refugees are assigned to resettlement locations. Using data on past refugees and their outcomes, the tool identifies the location where incoming refugees are most likely to thrive, based on their individual characteristics. Thanks to the generous support of The Rockefeller Foundation, IPL launched the first pilot test of the tool in Switzerland last year, and a U.S. pilot program is in development. We have also been working with additional partners in Europe and Latin America to explore implementing the matching tool in their own countries.
With an algorithm to find patterns in vast troves of data, governments and resettlement agencies can make the best matches possible for all resettled refugees.
The breakthrough idea was only the first step. It’s a much bigger project to translate this innovation into everyday practice so that it actually improves the lives of refugees. Here are some of the key lessons we’ve learned so far.
Emphasize Human-Centered Design
In designing the software, we centered the process around on our primary users: the refugee case officers at each of our partner organizations who make placement decisions for newcomers. They provide valuable insight on how the existing process works, identify obstacles and special cases they encounter, and give feedback, which allows us to further tailor the tool to their needs.
Once the interface of the tool is developed and ready for a demonstration and user feedback, target users will also be instrumental in the rapid iteration and further customization of the tool. After launching the pilot program, we schedule regular check-ins and establish open channels of communication to get continuous feedback and updates. Because we are focused on a small group of users—namely, the several placement officers within an asylum or refugee agency who make important choices every day about where to settle thousands of refugees—we can fine-tune the software to their highly specific needs.
Adapt to Different Countries
Each country has its own way of managing migration, often involving a complex interplay of regulatory, bureaucratic, and technical systems. The algorithmic matching tools we develop for each of our country partners must integrate seamlessly into the country’s existing systems and processes. In each case, we train the matching algorithm with a country’s own historical data, which means each matching tool learns from trends in that country’s past integration outcomes.
Regulatory requirements can also require specific adjustments. For example, we recently engaged in a Privacy Impact Assessment process in the Netherlands to ensure that our tool development complies with GDPR standards and regulations regarding the use of data and algorithms in the EU. Another way we adapt is by investing in the professional translation of important materials, such as data dictionaries and codebooks, so that we can precisely understand the systems we are working with. We also share these translations with our agency and government partners, creating a public good for future innovation.
If an innovation is to be easily scalable, it shouldn’t require a bureaucratic overhaul, so investing the time and resources to produce tailored tools for our partners is essential. It minimizes the added burden for the users and supports a seamless implementation.
Align Shared Goals
Even though the goals of the collaboration with our partners may seem obvious, it is helpful to clearly articulate them at the outset. IPL strives to conduct rigorous evaluations and publish high-quality research about immigration policy. With The Rockefeller Foundation’s generous support, we can not only implement these matching algorithm tools in various countries but also bring new knowledge into the world about their effectiveness in improving refugee integration. Our in-country partners share this underlying mission to improve refugee and immigrant integration to benefit newcomers and their host communities.
Another way to learn a partner’s goals is to first understand their concerns. For example, in setting up our pilot projects, the conversations around ensuring secure handling and storage of sensitive administrative data are very important for both IPL and the government partner, so that we can set up robust data sharing agreements and adhere to strict approvals processes. Aligning these mutual priorities helps us move forward and establishes trust between partners throughout the project.
Despite the political realities challenging refugee resettlement programs around the world, it is very exciting to see several countries and U.S. agencies embrace innovation and experiment with promising solutions. These collaborative partnerships bring IPL’s research to life and enable us to meaningfully scale our data-driven and impact-focused work.
The post To Scale Data Science for Refugee Integration, Partnerships are Key appeared first on The Rockefeller Foundation.