The global refugee population has surged dramatically in the past decade, overwhelming resettlement programs that struggle to match millions of refugees with host countries. A recent study from Harvard Business School suggests that artificial intelligence and machine learning might offer much-needed solutions for these programs, which are run by both governments and NGOs. The study proposes two new algorithms to better match refugees and asylum seekers to host countries, focusing on their likelihood of finding employment.
The current process of refugee resettlement is complex and slow, with agencies overburdened by the sheer volume of refugees due to conflicts in places like Ukraine, the Middle East, Africa, and South America. Economic fragility and political tensions over immigration have only worsened the situation.
Time-consuming process
“It’s a very time-consuming process that requires analyzing vast datasets to find suitable placements for everyone,” the researchers explain. They sought to create algorithms that could streamline the process, helping refugees integrate faster by matching them to areas where they’re likely to find employment.
Using data from Switzerland and the U.S., the researchers tested two algorithms. One significantly improved employment outcomes—by up to 50%—compared to existing methods, while the other focused on balancing refugee placements across various locations. The first algorithm, which prioritized job placement, achieved 96% efficiency in the U.S. and 98% in Switzerland. However, it led to resource imbalances, leaving some resettlement sites underutilized or overwhelmed. The second algorithm, which focused on even distribution, maintained a steady refugee flow and only slightly reduced employment outcomes.
The researchers stressed that job placement is a key factor in successful refugee integration. The goal is to place refugees in areas where they are most likely to find work, ensuring a smoother transition into their new lives. However, balancing refugee distribution across locations is crucial to avoid overburdening certain areas while underusing others.
Resettlement challenges
Globally, there are about 50 million refugees, with most coming from Afghanistan, Venezuela, Syria, Ukraine, and Sudan. Resettlement workers face the challenge of evaluating factors like education, work history, and cultural background to determine the best placements for refugees, which these algorithms aim to simplify.
The study looked at U.S. data from 2015-2016, focusing on refugees placed by the UN High Commissioner for Refugees and distributed among 10 resettlement agencies. In Switzerland, it examined asylum seekers assigned to one of the country’s 26 cantons. The results showed that AI could dramatically improve the speed and accuracy of placements in both countries.
The researchers are optimistic about further applications, including pilot tests in the Netherlands and Canada. While country-specific contexts differ, the methodology used in the algorithms could be adapted to other resource distribution challenges, from healthcare to business.
“In any situation where resources are limited and need to be distributed based on differing skills or needs, these tools can be incredibly valuable,” the researchers conclude.
