Communications of the ACM
AI Supports Displaced Peoples, Refugees in Ukraine and Beyond
Researchers and technology companies have been working with relief organizations, non-governmental organizations, and governments to develop solutions, some based on artificial intelligence, to support internally displaced persons and refugees worldwide.
Credit: SMU Centre for AI & Data Governance
The United Nations (U.N.) reported in May there were around 7.7 million internally displaced persons (IDPs) in Ukraine, people who have been forced to leave their homes but who remain in the country. Meanwhile, the U.N. High Commissioner for Refugees (UNHCR, the U.N. refugee agency) recorded over 5.9 million Ukrainian refugees who have fled their country to locations across Europe.
The humanitarian crisis in Ukraine is part of a global situation in which over 89 million people worldwide were found to be "forcibly displaced" due to persecution, conflict, violence, or human rights violations at the end of 2021, according to UNHCR.
Displaced peoples and refugees face a multitude of difficulties, from poor living conditions and loss of identity documents to longer-term resettlement, discrimination, and employability issues. Researchers and technology companies have been working with relief organizations, non-governmental organizations (NGOs), and governments to develop solutions, some of which are based upon artificial intelligence (AI), to support IDPs and refugees worldwide.
Ukraine: Responding to the immediate crisis
Palo Alto, CA-based Orbital Insight is using AI and geospatial data to monitor the movements of displaced peoples and refugees in Ukraine. The observations support agencies involved in the relief effort— including the U.S. Department of State and the U.S. Agency for International Development (USAid)—to deploy aid and emergency resources where and when they are needed.
Orbital Insight operates a cloud-based platform that uses data including vehicle, foot-traffic, and land-use observations to provide insights on global economic, societal, and environmental trends. Its clients typically come from sectors like defense and intelligence, energy, and financial services.
The platform's raw data is obtained from multiple sources. It includes high-resolution electro-optical (EO) satellite imagery (which includes information from beyond the visible spectrum) supplied by satellite manufacturer and operator Satellogic, as well as data from connected vehicle sensors and anonymized geolocation sensors on mobile devices. Customized algorithms are used to analyze the data and perform specific tasks, like computer vision-based classification and object detection.
When war broke out in Ukraine, Orbital Insight CEO Kevin O'Brien and his team adapted the platform's capabilities to the conflict. Foot-traffic algorithms, for example, can track the volume, direction, and speed of displaced peoples' movements. The company's analysis showed how foot traffic dropped off dramatically in eastern Ukraine as people fled intense fighting in the region— and how it then grew in cities in the west of the country as displaced peoples moved to those safer areas.
Such insights support humanitarian organizations' decision-making, such as whether emergency shelters and food should be redeployed to areas that people are moving towards. "Some of the first use-cases were to help manage relief resources both in Ukraine and at border locations," explained O'Brien. "The next was safe exit, safe passage: how do you get people further away from the fighting?"
Satellite imagery gives an aerial view of the situation at borders as refugees cross into neighboring countries like Poland and Moldova. Computer vision provides further insights by automatically counting the number of vehicles at border crossings. The results can be downloaded as geoJSON files, loaded onto a Geographic Information System (GIS), and visualized on maps for further analysis by relief organizations.
Alongside contracts with governmental agencies, Orbital Insight donates analytics via the Data Partnership, a collaborative platform that allows tech companies to share data with international development organizations for research and humanitarian purposes. The company also works directly with partners in Ukraine, including with Reface, a Kyiv-based AI startup that developed a popular face-swapping app and content creation platform. Reface has redirected its focus towards the conflict via numerous initiatives, including developing algorithms to identify Russian troops in satellite imagery and using data to fight disinformation.
Optimizing resettlement
AI can also support longer-term humanitarian processes, like the resettlement of refugees.
For instance, researchers at Stanford University's Immigration Policy Lab (IPL) and Switzerland's ETH Zurich have developed a tool called GeoMatch to help resettle refugees in geographies where they are most likely to succeed in their new lives.
IPL data scientist Elisabeth Paulson explained that at its core, GeoMatch is a stochastic matching algorithm that uses existing demographic and administrative data such as an individual's age, gender, family size, education, or job experience, collected via sharing agreements with partner organizations to provide "recommendations of locations where newcomers, so either refugees or asylum seekers, are likely to thrive."
The two-stage algorithm uses gradient boosted trees to generate predictions, then a matching algorithm identifies synergies between arrivals and locations. Every location has a separate predictive model, Paulson said. "We have to take into account capacity constraints that often exists at locations" since each new person assigned to a location means one less slot for future arrivals.
GeoMatch also draws on historical data about previous arrivals' outcomes. "Most countries have one or more key outcome variables that they're interested in maximizing. So, in the U.S., for example, it's employment after 90 days." Other locations may look for some educational attainment, or an integration outcome, Paulson said, so "the metric that we try to predict and then also optimize on" varies.
Paulson's new extension to GeoMatch improves its "allocation balancing," she said, building on the tool's base algorithm to balance "the goal of improving refugee outcomes with the desire for an even allocation over time."
Identity matters
Lucia Gallardo is the founder of Emerge, a development lab that uses technology to tackle global and societal challenges. Honduras-born Gallardo said she became frustrated with the "incredibly slow and bureaucratic" processes that refugees can face when she was working at the General Consulate of Honduras in Canada.
Homeward, Emerge's AI and blockchain-driven solution, is an identity management and intelligent resettlement platform currently under development, with pilot studies in progress in resettlement cities including Toronto, Canada, New York City, and Atlanta, GA, in the U.S., as well as Paris, France, and Bonn, Germany.
Homeward's algorithms match people to the locations most conducive to their successful integration, based on socio-cultural preferences alongside biographical and biometric data, and education and employment experience. Gallardo wants the platform's AI matching capabilities to make immigration more dignified and effective.
For some, Gallardo said, refugee resettlement is the best pathway, but this may be subject to closed quotas. For others—with certain skillsets—more open-ended migration programs are a better fit. "It allows them to come into a country not under a narrative of burden, but rather as a narrative of value-add," she said.
Proof of identity is necessary for resettlement, but the realities of fleeing crises mean documents may be left behind or lost; Homeward tackles this problem with a blockchain-enabled public key infrastructure that produces a unique digital identification (ID) that can be self-managed and accessed by a user via an app.
Gallardo said this "generative identity" builds trust as an individual uses and adds to it. It is produced via what she calls "attestations," instances such as receiving service at a refugee camp or crossing an international border where one's identity has been verified. "The idea is really to capture all of the instances in which you have consistently proven that you are who you say you are to build that legacy."
The Emerge team is now building public and private partnerships, including with the UN Development Program and several banks, to further develop Homeward's AI and blockchain framework. Planned extensions will include access to support, like family reunification, training and education, as well as services like banking.
Karen Emslie is a location-independent freelance journalist and essayist.
No entries found