AI for Immigrant Integration
More than 270 million people have left their home countries for better lives abroad. The Immigration Policy Lab is working with government agencies to design evidence-based immigration policies that are beneficial for immigrants and their host countries. They are one of six labs to receive start-up funding from Stanford Impact Labs for 2021-2023.
Jennifer Fei, Stanford Immigration Policy Lab’s senior program manager, Patrick McEvenue, Acting Senior Director, Strategic Policy and Planning Branch at Immigration, Refugees and Citizenship Canada (IRCC) and Sjef van Grinsven, project leader at the Central Agency for the Reception of Asylum Seekers (COA) in the Netherlands describe their work to design, test, and scale an algorithm-based tool for government partners to match new immigrants to locations where they are likely to be most successful.
Q. What is the social problem you are working on?
JENNIFER FEI (Immigration Policy Lab): No matter why people migrate – whether to escape persecution, or to pursue better opportunities abroad – the place where they settle within a host country matters significantly. The destination can affect their ability to find employment, maximize earnings, and access education and healthcare. Location decisions therefore also affect the extent to which immigrants benefit the local economy and society. When immigrants are unable to gain a foothold, climb the economic ladder, and integrate in this holistic sense, there is potential for public backlash and policies restricting immigration. Our goal is to provide governments and immigrants new data-driven tools to help them to choose the most beneficial location, using insights from rich historical data and human-centered artificial intelligence (AI).
Our matching algorithm, GeoMatch, identifies synergies between immigrants’ personal characteristics and locations, develops models to predict how each newcomer will fare at various potential locations, and matches them to one that gives them the best chances of integration. In Switzerland, for example, we’ve found the ability to speak French (i.e. among French-speaking African refugees) results in a larger payoff for refugees assigned to French-speaking regions than for those assigned to German-speaking parts of the country. This is just one of the ways individual characteristics interact with local conditions like social networks and labor markets; it’s why an algorithm-based approach that can account for multiple kinds of interactions can be so helpful.
The tool has the potential to improve outcomes for both immigrants and host communities. When immigrants integrate more rapidly, they become less dependent on the social safety net and better contribute to the local economy through labor, entrepreneurship, and consumption. Our evidence suggests these tools are more cost-effective than resource-intensive integration programs like skills training. They can be deployed at scale in a number of different locations to benefit large cohorts of incoming newcomers at a low cost.
SJEF VAN GRINSVEN (Central Agency for the Reception of Asylum Seekers, Netherlands): 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 for scientific insights to help understand what helps determine these successes. Through our cooperation with Immigration Policy Lab we will have access to the information and tools to make the best location match possible for every incoming refugee in the Netherlands.
PATRICK MCEVENUE (Immigration, Refugees and Citizenship Canada): Canada’s Express Entry system selects economic immigrants based on factors that are empirically shown to contribute to better long-term wage outcomes. At the same time, Immigration, Refugees and Citizenship Canada and its partners work to ensure that the benefits of immigration for both newcomers and host communities can be enjoyed across the country. We’re interested in ways to leverage data to support these aims. Evidence suggests an immigrant's initial arrival location plays a significant role in shaping their integration success, and personalized recommendations have potential to benefit newcomers to Canada. Immigration Policy Lab’s research surrounding a “matching algorithm tool” could enable us to add to the information already sought out by newcomers. We expect this research collaboration to add to our growing evidence base, and test innovative applications of data-driven tools in the economic immigration context.
Q. What will the start-up lab funding and team approach help you do?
JENNIFER FEI (Immigration Policy Lab): Our research projects with implementation partners in Canada (Immigration, Refugees and Citizenship Canada) and the Netherlands (Central Agency for the Reception of Asylum Seekers) are co-designed at each step--from initial pitch meetings to propose a potential collaboration to the analysis through historical backtests to identifying important research questions to guide our design, we work with our partners at every stage to ensure that the project meets our shared objectives for research and policy impact.
As part of our commitment to a partnership-centered applied policy research model, I am the full-time program manager on the Immigration Policy Lab team who serves as a primary point of contact for our partners, leading working groups and weekly meetings with our partners and stakeholders to ensure that partnership engagement is consistent and meaningful. This ensures support from Immigration Policy Lab members and partner organizations throughout the project and that the outcomes are based on a common set of goals: to improve integration for newcomers and produce knowledge and evidence about how our AI tool for immigration integration works in the real world.
SJEF VAN GRINSVEN (Central Agency for the Reception of Asylum Seekers, Netherlands): The Immigration Policy Lab will be able to provide us with access to analytical methods and capabilities that will unlock insights about historical refugee integration and help inform future program and policy design. Partnership between the Immigration Policy Lab and our Central Agency for the Reception of Asylum Seekers is highly collaborative. We co-create materials and proposals, give feedback to each other often, and design the timeline/projects together. Although there have been studies about the characteristics of refugees and their relation with integration outcomes in the Netherlands, these studies haven’t led to concrete tools to use these insights directly in the work process. Until now.
PATRICK MCEVENUE (Immigration, Refugees and Citizenship Canada): Canada has in its Longitudinal Immigration Database (IMDB) a rich source of data to learn about immigrants and their outcomes. Partnering with the Stanford research team gives us access to analytical methods and capabilities to unlock insights about historical immigration integration which can inform future program and policy design. Along with the Immigration Policy Lab’s technical expertise in machine learning, data science, and evaluating impact, working with them also allows us to draw from its experience setting up matching algorithm pilots in other countries. The partnership between Immigration, Refugees and Citizenship Canada and the Immigration Policy Lab positions us to realize and learn from these innovations in the context of the Canadian immigration experience. All of this supports our evidence-based approach and the value we place on partnerships towards the successful settlement and integration of new Canadians.
Q. What are you most excited about with this work?
JENNIFER FEI (Immigration Policy Lab): Given our strong relationship with our implementation partners in Canada and the Netherlands that we’ve built over the last few years, the potential real-world policy impact of our research is significant. Since our partners are invested in learning and gaining insights from the backtest analysis and randomized controlled trial pilot testing initiatives of the GeoMatch tool, we have the opportunity to develop widely relevant evidence about the matching algorithm’s efficacy and impact. These insights can directly inform programs and policy at the Immigration, Refugees and Citizenship Canada and the Netherland’s Central Agency for the Reception of Asylum Seekers, as well as those of other immigrant-receiving countries around the world looking to improve integration outcomes for their incoming immigrants and refugees.
SJEF VAN GRINSVEN (Central Agency for the Reception of Asylum Seekers, Netherlands): We are convinced that the insights from working with the Immigration Policy Lab research team on this project will result in improved practice and that our government is therefore prepared to invest long-term in this international policy research cooperation and relationship. To innovate our work process, which will lead to gains in terms of integration, will benefit the target group we work for, our direct and indirect stakeholders and society at large.
PATRICK MCEVENUE (Immigration, Refugees and Citizenship Canada): We’re excited to explore further with our partners the positive results that came out of the Immigration Policy Lab’s initial study and the potential for economic immigration clients, while building our policy and data innovation capacity through this collaboration. This work has the potential to deliver the insights needed to help immigrants go where they can thrive in their new communities, and support the distribution of the benefits of economic immigration across Canada. This sort of personalized resource for immigrants would be the first of its kind, and we continue to explore how this work could inform economic immigrants coming through the Express Entry process. This could also help us better understand how immigrants make decisions and learn more about the determinants of successful immigration.