Could a tech start-up helping refugees resettle in the U.S. and Europe help Canada?

This independent journalism on data, digital transformation and technology for social impact is made possible by the Future of Good editorial fellowship on digital transformation, supported by Mastercard Changeworks™. Read our editorial ethics and standards here. 

According to its founders, a Canadian platform helping refugees resettle in the United States and Europe could help Canada find homes and support for refugee newcomers more quickly and effectively. 

It’s still a “complex intergovernmental problem,” says Craig Damian Smith, who co-founded Pairity with Gotam Bhardwaj and Radboud Reijn. Pairity is a technology platform that uses a series of algorithms to match refugees to sponsors, community organizations, volunteers and cities that might best fit their needs. 

Executive director and principal investigator Smith is a researcher at the Centre for Refugee Studies at York University, and was previously a senior research associate at the Canada Excellence Research Chair in Migration and Integration at Toronto Metropolitan University. Bhardwaj serves as the director of engineering, and Reijn as the director of operations.

In Canada, there are immigration targets at federal, provincial and local levels, with incentivized pilot schemes to encourage migrants to move to rural and Northern areas, as well as small and medium-sized cities. Canada, in particular, sees a lot of “secondary migration,” Smith says, because there isn’t room for everybody in the big cities, which are often the natural choice for newcomers. 

Pairity carries out a matching process alongside organizations in the settlement sector, and refugee populations are followed longitudinally for at least a year. 

Information about their outcomes and the quality of their integration into their new communities are then fed back into the algorithms that drive the initial matching process. 

The team, comprised of human rights experts, social scientists, technologists, data analysts and scientists, and monitoring and evaluation specialists, has built bespoke versions of this technology in the U.S. and Europe. 

In Germany and the Netherlands, pilot programs have been matching refugees to local and municipal governments that can take people in. Re:Match is the platform that Ukrainian refugees use to find which one of the six participating cities in Germany they would like to settle in. 

In collaboration with a local NGO in Krakow called Salam Lab, Pairity’s algorithms have successfully matched 77 refugees to partner cities in three rounds, with the most recent round of matching taking place in September.

While resettlement work in Europe is more city-focused, it is a federal mandate in the U.S. Here, the algorithm matches refugees to private sponsors, gathering data about sponsors’ capacity and refugees’ own preferences. 

Verified third-party data from the U.S. Census and other datasets is also fed into the algorithm, especially supplementary data around employment opportunities and housing / living costs that can better match refugees to sponsors. 

“Our platform has core code, which is then adapted for each jurisdiction,” Bhardwaj says. “Our partners [in each place] are administering surveys and giving us information when it comes to housing and diaspora communities, which then feeds into the data machine.” 

Expanding their technology to multiple national jurisdictions means Pairity can also carry out “interesting comparative work between federal and municipal models” of refugee resettlement, Bhardwaj adds. 

What sorts of data and information are fed into the algorithm to start the matching process?

An algorithm relies on a series of statistical processes and calculations, which it then analyzes to repeat a specific outcome. As such, algorithms require a lot of data – in this context- about refugees, their preferences and the capacity for specific cities and municipalities to receive them. However, this part of the process cannot rely solely on technology or automation. This information is gathered through very non-technical means, say Smith and Bhardwaj: in-person surveys.

The pair stress how vital this part of the process is. Settlement organizations in each location are trained on how to administer the survey, which collects information about refugees’ preferences and requirements. This baseline data then allows organizations to match refugees up with private sponsors or cities with the capacity to welcome newcomers. 

The matches are based on six variables: geographic distance, household composition, newcomer vulnerability and volunteer capacity, labour market experience, language and culture, and hobbies and interests. 

“Hand-matching will often match the best possible fit and then go down the ladder with what is left,” Smith says, to the point where the ‘final’ match might not be a good match at all.

The data is then monitored over time to measure metrics around an individual’s social cohesion, sense of belonging and integration into their new community. Follow-up surveys, focus groups and semi-structured interviews form part of the longitudinal research and analysis. 

“In some cases, there are several hundred outcome variables,” Smith says. Using data science and social science methodology, the team plans to grow enough data points to allow for statistical processes showing correlation and causation – and eventually being fed back into the algorithm.

It’s also vital, he says, to recognize the relative vulnerability of refugee populations – not only do those conducting interviews and collecting data have to be aware of that, but the appropriate security protocols also have to be built into the technical architecture of the matching platform. 

Why the technology won’t replace workers in the settlement sector 

According to Smith, Canada has always had “remarkably positive attitudes to humanitarian responses and refugee resettlement, and there has been a big demand for everyday people to be involved in that.” 

Before developing Pairity, Smith co-founded Together Project, a matching service that pairs refugees with community and social support to improve integration. 

There is a three-fold application of the technology, Smith adds. It can be adapted to particular national jurisdictions and help match refugees with the appropriate services and support at a larger scale than hand-matching could. 

The crucial application, however, is in reducing bias. 

“Bias comes into hand-matching or one-to-one matching. You’re not equitably distributing scarce resources among the population,” he says.

“There are millions of possible matches in a cohort of 50 people, which can’t be done by one person.”

Smith and Bhardwaj stress the technology is not there to replace the work of staff in the settlement sector but to help them make more informed, just and equitable decisions. Every project starts with a co-designing stage, where local organizations that work with refugees can build their goals into the technology. 

“Everything we do starts and ends with human interactions,” Smith says. “A big part of what we do is labour intensive. We’re working on a very thorny social and political issue, and trying to maintain the idea that an algorithm is not secretly doing the work is a struggle for us, to be frank.” 

Training staff in ethical data collection methods encourages them to consider that the people they gather information from – newly settled refugees – are usually “busy and stressed out,” Smith says. 

“We’re bringing tech and tools that complement the existing ecosystem, as opposed to bulldozing or wiping everything off the table,” Bhardwaj adds.

Once the algorithm suggests matches, resettlement agencies can vet the suggestions. 

Pairity’s technical systems and roadmap are built so that eventually, Pairity might be able to step back from the matching process and have local resettlement organizations running their own. The analysis process requires not only buy-in from local resettlement agencies but also refugees themselves. Seeking informed consent means that refugees “understand how matching works and understand how [expressing] their preferences is going to inform the matching,” Smith says. 

He adds that refugees understand the bigger picture of how sharing their data and preferences can improve ongoing advocacy to improve outcomes for refugees and asylum seekers.

“The evidence we’re collecting isn’t purely academic or for the algorithm to get better,” Smith says. “It’s a social impact policy area for which we need more evidence-based policy.” 

Tell us this made you smarter | Contact us | Report error