Master's Thesis: Refugee resettlement and AI-driven matching hos Institut for Fødevare- og Ressourceøkonomi - Københavns Universitet

In recent years, there has been a significant amount of attention devoted to the field of matching theory, especially since the Nobel prize of 2012. One of the latest applications is refugee resettlement.

Questions typically address how to ideally assign the refugees, given that hosting locations have preferences over refugee characteristics while the refugees themselves have preferences over whereto they get resettled [1, 2]. It has been suggested that data-driven assignment can improve the matching quality significantly, say as measured by overall employment rate [3]. At the same time, even though the matching may improve in the aggregate, there can be concerns that the underlying algorithm systematically favours some participants over others [4, 5].

The topic is very timely and, unfortunately, not unlikely to grow in importance for many years to come.

What to write?

A thesis on this topic should include a formal model with a basic game-theoretical analysis. It should also contain a practical description of some relevant refugee resettlement procedure, say for the case of Denmark, as well as a simple description of some ideas pertaining to deep learning. If relevant, computer simulations can be used to illustrate the theoretical findings.

Contact:

Postdoc Jens Gudmundsson

Section for Production, Markets and Policy (IFRO)

Rolighedsvej 23, 1958 Frederiksberg C, Denmark

+45 35326197 // jg@ifro.ku.dk

References:

[1] Delacretaz, Kominers, Teytelboym ”Matching Mechanisms for Refugee Resettlement” (2020)

[2] Biro and Gudmundsson, “Complexity of finding Pareto-efficient allocations of highest welfare” (EJOR 2021)

[3] Bansak et al ”Improving refugee integration through data-driven algorithmic assignment” (Science 2018)

[4] Obermeyer et al ”Dissecting racial bias in an algorithm used to manage the health of populations” (Science 2019)

[5] Waldrop ”What are the limits of deep learning?” (PNAS 2019)

Bemærk: Du skal ofte bruge forhåndsgodkendelse fra dit universitet eller studievejleder for at sikre, at projekter eller specialeopgaver på SDU Jobbank vil blive accepteret som en del af dit studie. Kontakt de relevante aktører i god tid for at sikre, at du vælger det rette projekt.