In recent years, European states and international organisations have developed various technologies to detect migration patterns and predict the number of people from third countries seeking asylum in the EU. But doubts have been raised about both the effectiveness and desirability of using predictive technologies to impede migration.
Around 2.3 million migrants arrived in the EU in 2021, reflecting a resurgence of flows since the outbreak of the Covid-19 pandemic.
The ongoing carnage in the central Mediterranean and the increasingly frequent reports of violence and pushbacks perpetrated by the border police in the Balkan route highlight on the one hand, the failures and struggles of European countries to manage the ongoing situation and, on the other hand, the willingness of migrant people to find better living conditions in Europe.
Developments in forecasting and risk assessment techniques have gained appeal within European institutions. For instance, the Danish Refugee Council developed, in cooperation with IBM and funded by the Danish ministry of foreign affairs, the Foresight Project, a forecasting tool to predict forced displacement worldwide.
In Germany, the federal foreign office is working on the PREVIEW project, a tool for monitoring ongoing conflicts with the long-term goal of developing a tool to predict future displacing events. The promise of predicting with high accuracy the evolution of migratory flows is attracting institutional investors and solution-oriented policies. Yet, the science behind these technologies is not always precise, and the risks to migrants’ safety are not fully understood.
Among the upcoming projects is ITFLOWS, an EU-funded project under Horizon 2020’s Secure Societies programme supported by up to €5m from Horizon 2020 funds and developed by a consortium of 14 members, including universities such as Universidad Autonoma De Barcelona, NGOs such as Oxfam Italia and the Italian Red Cross, and a private company, Terracom.
The consortium aims to develop the EUMigraTool (EMT), an AI tool that, through a combination of agent-based modelling and deep learning architectures, is intended to provide predictions on the number of migrants arriving in a specific European country and, as the website states, will help “end-users predict the number, gender and age range of asylum seekers/non-recognised refugees entering several countries of the EU, as well as showing real-time information regarding the camps, some major cities and the conflict zones in non-EU countries.”
Who are the end users that EMT addresses?
“We understood that one of the main problems in migration management is that organisations working with migrants don’t have enough information in advance to manage their arrival properly”, states Cristina Blasi Casagran, assistant professor in EU law at the Universidad Autonoma de Barcelona and Project Coordinator of ITFLOWS.
“They don’t know, for example, what the real needs are when these migrants arrive. We are providing a tool that provides a range of arrivals according to the migrant’s country of origin, gender, and age group. This allows organisations to prepare in advance and provide sufficient resources for the arrival of these migrants”.
According to the consortium, ITFLOWS only uses open-source data. “We do not rely on government data or other types of non-public data. We check databases such as Eurostat, ACLED, and Frontex data, which provide us information on the number of illegal border crossings”, states Blasi Casagran.
Although the project aspires to help operators and organisations better manage the flow of arrivals, many issues place such projects under the watchful eye of migrant protection and security observers. Since its announcement, ITFLOWS has attracted the attention of civil society. An open letter drafted by Access Now outlines several points on the possible risks of this software. The first point, in particular, states that “predictive technologies risk being repurposed for the securitisation and criminalisation of migration.”
Along the same lines is Derya Ozkul, signatory of the letter, senior research fellow at the Refugee Studies Centre at Oxford University, and author of the AFAR report on the use of new technologies in the governance of migration and asylum in Europe, which adds: “Concerning predictive analysis, the research conducted so far shows conflicting results. For example, even the predictions of the Foresight Project, which is supposed to be one of the best ones available, are not entirely correct. It did well in some years, but the higher change in displacement, the higher inaccuracy the tool had. So we should not take these tools as the silver bullet. Instead, we should concentrate our efforts to understand the real causes of conflicts and prevent them.”
One of the main concerns regarding risk prediction and forecasting tools is the possibility of these technologies being reproduced and reused for risky purposes. Blasi Casagran insists, however, that “we want to provide technology that benefits migrants and the people who help them. That is our priority. So we don’t want this at any level and in any way tool to be for governments”.
In the current environment of increased securitisation of European borders through establishing ‘Fortress Europe’, tangible and digital borders become a testing ground for new technologies.
Yet, no matter how advanced these technologies may be, the question arises about their actual usefulness and to what extent the research conforms to a joint European determination not to pursue systemic and integrated solutions to the migration issue but instead to an emergency and short-term approach. As Ozkul says, “It is very vague, those boundaries where science stops, and politics begins.”
Systemic and integrated crises such as climate change and wars are the main reasons people get on the move. Jan Tobias Muehlberg, professor at Université Libre de Bruxelles, an expert in the AI Act and member of #Protect Not Surveil, a coalition of civil society partners that drafts amendment recommendations to the AI act in the migration context, emphasises “that it is economic activity in the Global North, including modern forms of (digital) colonialism and our development of all these AI tools, that drives climate change. We are not addressing the underlying problem here, but we are escalating a crisis by building harmful technologies that allow us to hide behind a notion of perceived objectivity. Managing these numbers by means of technology is feasible; trying to manage these numbers by means of discriminatory technologies is a disaster”.
The AI Act currently under discussion classifies artificial intelligence technologies in migration management as “high risk”. Yet they are not banned, and the minimum controls are still weak enough not to guarantee adequate protection for people on the move.
According to Muehlberg those AI systems deployed in migration management — specifically those for making predictions, assessments and evaluations in the context of migration claims — are “inherently discriminatory and judge people based on factors outside of their control, based on data that subjects are often not able to query and rectify, and with no possibility to question decisions or life or death”.
An uncontrolled application of these technologies could facilitate illegal push-backs in order to prevent people from exercising their right to seek asylum. But the implications could extend to very different contexts. “The borders are often a testing ground for technologies”, says Muehlberg. “Yesterday it was detention centres, in 2024 it’s going to be everyone at the Olympics in France, and with that we have a precedent for domestic use of AI-powered surveillance tech.”
The EMT is in the development phase, and it will still take time to fully operational: “It is not easy to find new sponsors who can help us to continue developing the tool,” says Blasi Casagran. “We don’t want money from Frontex; we don’t want money from governments. It is not easy to find new sponsors who share our intentions and will help us continue developing the tool”.
What is certain is that risk assessment and forecasting technologies will become increasingly widespread among European countries. If the need to alleviate and make sustainable the migratory flow within the EU is crucial; certainly the risk of dehumanisation and discrimination of people on the move is right around the corner.
“We are Black and border guards hate us. Their computers hate us too”, is the answer #Protect Not Surveil got from a migrant in Brussels in early 2020 at the onset of the pandemic. “It’s a good summary of the migration issue and how the EU deals with it,” states Muehlberg, “human dignity and human rights as key values of the EU don’t seem to apply to people on the move, certainly not if they aren’t white.”
