ANNIE™ MOORE: INCREASING EMPLOYMENT OF RESETTLED REFUGEES USING MATCHING, MACHINE LEARNING AND INTEGER OPTIMISATION

 

Tens of thousands of refugees are resettled from refugee camps around the world every year. The evidence shows that the initial community into which refugees are resettled can dramatically affect socioeconomic outcomes for them and their children. However, not all communities are able or willing to host refugees. A key research question is: how can we best use the hosting capacity of communities for the benefit of refugees and local residents? This is a crucial question for social and migration policy at the time of a record-breaking number of refugees around the world. To answer the question, we need to understand what makes a good refugee-community “match”. By studying how outcomes of refugees depend on their characteristics and the features of local areas, we can understand which communities given resettled refugees the greatest opportunities to thrive.

https://player.vimeo.com/video/392893650

Annie™ MOORE: Improving refugee resettlement from Refugees AI on Vimeo.

 

Download the Video Transcript

Refugee resettlement constitutes a very unusual “marketplace”: it involves no money, but it does require procedures for matching refugee families and local communities. Refugee resettlement is a prime example of the model of matching with complex, multidimensional constraints because families require different support services, such as training and child care. The conceptual idea of “refugee settlement” as a matching problem was outlined in a series of papers and policy briefs by Alex Teytelboym and  Will Jones, formerly of Oxford’s Refugee Studies Centre (Jones and Tetelboym, 2018). Teytelboym, with co-authors with David Delacrétaz, a Prize Postdoctoral Fellow at Nuffield College, and Scott Duke Kominers, set out a basic theoretical model of refugee resettlement as matching with multidimensional constraints in a 2016 working paper. Subsequent research has refined and extended that initial analysis (Nguyen et al., 2019; Delacrétaz et al., 2019)

The matching model provides the theoretical foundations for the pioneering software Annie™ MOORE (Matching and Outcome Optimization for Refugee Empowerment). Developed with researchers at Worcester Polytechnic Institute in Massachusetts and Lund University in Sweden, Annie™ MOORE uses advanced machine learning and state-of-the-art integer optimization methods to suggest placements of refugees that (i) maximise their employment chances of refugees (ii) ensure that the needs of the refugees (e.g., child care or language support) are met, (iii) guarantee that the service capacities (e.g., housing or places in training programmes) of hosting communities are not exceeded (Trapp et al, 2018).

HIAS, a US refugee resettlement agency, has been using Annie™ MOORE since 2018, and thus far has matched over 1,100 refugees with hosting communities. Estimates suggest that Annie™ MOORE has raised the employment rate for refugees from 30 to 40 percent. It has reduced the proportion of refugee families placed in communities which lack the services necessary for their support from around 20 percent to essentially zero. This has dramatically improved the quality of refugee integration in communities. Finally, Annie™ MOORE has improved HIAS capacity, empowering HIAS staff to make better and faster decisions about placements and to focus their time on helping refugees that require individual attention.

Other agencies have shown interest in the potential of matching models to improve refugee resettlement. A recent report by the UK Independent Chief Inspector of Borders and Immigration recommended that the Home Office “improve the geographical matching process” for refugees in the Syrian Vulnerable Persons Resettlement Scheme. Recently, the Swedish government recommended the adoption of carefully designed optimisation matching systems for resettling asylum seekers living in Sweden.

 

Visit Refugees.AI

 

In June 2020, Alex Teytelboym was Highly Commended in the Policy Engagement category of the University of Oxford’s 2020 Vice-Chancellor’s Innovation Awards for his work on Annie™ MOORE.

References

Delacrétaz, D., S. Duke Kominers and A. Teytelboym (2019)

“Matching Mechanisms for Refugee Resettlement”

Jones, W. and A. Teytelboym (2018), “The local refugee match: Aligning refugees' preferences with the capacities and priorities of localities”, Journal of Refugee Studies.

Nguyen, H., T Nguyen and A Teytelboym (2019). “Stability in Matching Markets with Complex Constraints” 

Trapp, A.C., A. Teytelboym, A. Martinello, T.Andersson and N. Ahani (2018) “Placement Optimization in Refugee Resettlement” Working Paper 2018:23 Department of Economics School of Economics and Management. Lund University