New technologies in migration and asylum governance: who benefits?

Use of new technologies across European immigration and asylum systems can expedite some decision-making processes, but also increase vulnerabilities for migrants, meaning new governance frameworks are needed.

AI generated image of people in line enetering a round structure
Image: DALL·E 2 / OpenAI

An AI-generated image of mobility and migration

From dialect recognition systems to matching tools, the use of new technologies is gradually rising in the migration and asylum fields across Europe. Several states have started using or testing such technologies to control who enters their borders or gains access to their protection mechanisms. These technologies, especially automated decision-making systems, can expedite decision-making processes, benefitting governments and some applicants.

Yet they also involve inherent risks of bias, discrimination and potential “machine mistakes”, which pose a significant threat to migrants and asylum seekers – often already disenfranchised and facing challenges in seeking remedies. Their use can also lead to new relationships between the public and private sectors, requiring new governance structures and legislative frameworks to regulate who is responsible for data protection, possible machine mistakes, and related inaccurate or discriminatory outcomes.

As the EU Artificial Intelligence (AI) Act proposal categorises AI uses for immigration, asylum and border control as “high-risk”, there is a need for systemic investigation of current practices and their use across Europe. In a new report, Automating Immigration and Asylum: The Uses of New Technologies in Migration and Asylum Governance in Europe, I map out the existing uses of new technologies across European immigration and asylum systems at both national and EU levels.

Technology uses in migration and asylum

The report explores the technologies used prior to arrival, at the border and within European territories. The term “border” here refers to the physical borders of states and the location where migrants, asylum seekers and refugees are when they are subject to a particular technology used by an authority. This is not to deny that “digital borders” function on and beyond territorial borders. Rather, the temporal approach shows that migrants, asylum seekers and refugees are subject to various technologies at each different stage of their mobility around and inside Europe.

The practices I explore in the report include specifically:

  • Forecasting tools for forecasting future immigration and displacement towards Europe
  • Automated processing of short- and long-term residency and citizenship applications, used in Norway and, to some extent, Sweden
  • Document verification for detecting possible fraud in identity and supporting documents, used in the Netherlands
  • Risk assessment and triaging systems, used to categorise applications for travel to the Schengen zone, and  assess applications to the EU Settlement Scheme, identify irregular migrants and assess marriages and civil partnerships in the UK
  • Speech and dialect recognition, used to help applicants with citizenship applications in Latvia, identify asylum seekers’ country of origin in Germany and transcribe interviews with asylum seekers in Italy.
  • Automated distribution of welfare benefits to asylum seekers, used in Norway
  • Matching tools, used to match asylum seekers with reception centres in Norway, and for screening similar asylum applications in the Netherlands
  • Mobile phone data extraction, used to verify identities and narratives of asylum seekers in Germany, the Netherlands, Norway, Denmark and the UK
  • Electronic monitoring, used as GPS ankle tags in the UK.

Other technologies are under development, including risk assessments and profiling through interoperability between large EU information systems; document verification in Belgium and France, and categorisation of appeal cases in the Netherlands. An EU-funded research project tested controversial lie detection technologies in Hungary, Latvia and Greece. Another project explored the feasibility of behaviour analysis, including emotion recognition for border control. Although currently not implemented, these projects’ findings are available to inform future product development. A number of matching tools are also being tested to optimise migrants’ and asylum seekers’ settlement outcomes.

Some technologies have been tested or implemented, but then terminated. For example, speech and dialect recognition to identify asylum seekers’ countries of origin was tested in Turkey, but not used due to poor accuracy. In the UK, between 2015 and 2020, the Home Office used a potentially discriminatory algorithm to process visitor visa applications, but stopped following civil society campaigning. The Netherlands used risk assessment between 2014 and 2021 to evaluate potential sponsors of highly skilled migrants, storing companies’ details, including the ethnic composition of their boards – a practice halted thanks to internal and external pressures. The country is currently developing a new model to assess sponsors without storing information on ethnicity.

Understanding individual technology risks

This wide range of applications for new technologies implies that each should be investigated independently, including its context and unique stakeholder requirements. The report, therefore, debunks a black-and-white perception of technology uses. Some technologies, such as those that prioritise migrants’ preferences in their settlement processes, can benefit applicants by giving them a say in their settlement trajectory. Others, such as profiling people through risk assessments or monitoring them through invasive electronic tools, can be extremely harmful. This makes it crucial to examine each use of new technology in its own right, including its design and implementation processes, and legal and social impacts.

The need for transparency

Transparency over whether a decision-making process involves automation and, if so, its technical details, is also vital to understanding technology’s impact, yet these details are not always made public or are too difficult for outsiders to comprehend. Yet without transparency, migrants lack access to justifications for decisions affecting their lives and cannot pursue effective remedies.

Transparency is also important for decision-makers. While human caseworkers are involved in each practice I examined, the extent of their involvement and their knowledge of the entire decision-making process varies. For example, some new technologies, such as mobile phone data and speech or dialect recognition tools, are used to produce evidence for decision-makers in the asylum process. These automated reports cannot be a reason to reject an applicant’s claim, but they can impact the process if decision-makers over-rely on reports prepared by these sophisticated tools.

Who benefits?

Many of the technologies I explore in the report are designed to benefit, first and foremost, state authorities. Migrants’ interests and voices are generally not included in their design and implementation. Some technologies aim to benefit migrants above anyone else, such as speech recognition in Latvia to help them with their citizenship applications. However, most are designed to support migration controls or benefit state needs rather than those of migrants. I therefore argue that who benefits from these technologies, who has access to their details, and who is included and excluded remain crucial questions to inform appropriate governance structures for AI use in migration processes.

Further information

Ozkul, Derya (2023) Automating Immigration and Asylum: The Uses of New Technologies in Migration and Asylum Governance in Europe. Oxford: Refugee Studies Centre, University of Oxford.

The report was prepared for the Algorithmic Fairness for Asylum Seekers and Refugees Project, funded by the Volkswagen Foundation.