Hello and welcome to another episode of the Migration Ocean Podcast. I'm Rob McNeil and I'm Jackie brought out. So today Jackie has been talking to two experts on the use of artificial intelligence and other modern forms of technology in the management of rivals in Europe. So, Jackie, what made you want to talk about this?
Thanks, Rob. I feel like AI is kind of everywhere at the moment, from chatty through to, you know, even within the university, how it might kind of impact teaching and learning here. And so it felt really important to kind of hear a little bit about what's been happening within migration governance, both processes that been going on for a long time as ever with these things.
You know, you think it's the kind of new big thing and it actually turns out that some some of the things have been happening a kind of much longer processes. But hearing about some of the new technologies in the way that genuinely they are impacting on decision making and migration governance.
One of the questions for me is, I mean, is there a problem? I mean, is it wrong to use modern technology to try and understand who's arriving, what kind of risk they pose to people and for governments to to to try to ensure that they've got the most efficient mechanisms possible to assess people's claims and to deal with things quickly? Yeah, I mean, one of my favourite programs is Mad Men, and there's a line that Don Draper says in that way he he says, What if change is neither good nor bad?
It just is. And that's something that I thought about quite a lot in this discussion, which is is it's the tools themselves that are the issue or this kind of systems that they're going into.
I think one of the really interesting points in the discussion is if they are going into a system that contains inequalities, biases, discrimination, they have the possibility to amplify and replicate that form of discrimination, and particularly because they allow decisions to be made much quicker than humans might make decisions with fewer moments for reflection. You know, are we walking into a situation where actually we're going to see these kind of biases replaced on a much larger scale?
But I also think it's something about the moment that we're at. It's true that these technologies have been around for a while, but this does feel like quite a transformative moment. And that means that policymakers and governments are kind of writing the rules of the game.
And at the moment that we're writing those rules, it's probably a good idea for us to be having conversations about what we want these types of decisions to be, how much we want human judgement as part of them, you know, as opposed to kind of predictive, you know, how much we want to be If we were making an application for a visa or for asylum, how much we would want that to be done by a person looking at kind of us versus it being done on a kind of predictive algorithm.
All of these sorts of huge questions I think is kind of grappling with are also come into play within migration governance and are happening both for us as citizens in deciding what's acceptable, what might not be acceptable, and then also within governments. So, I mean, there's a question here in that case about just allocation of resources and the scale of situations.
I mean, the UK obviously at the moment, I mean, one of the things that we we talk about all the time is this massive backlog in the asylum and the asylum system in the UK. That's largely a facet of the slow processing of claims by by decision makers. Now I mean surely removing a degree of the kind of of human biases from decision making is is arguably a good thing sometimes, isn't it?
I mean, it's not to say that, it's not to say that automating processes is a perfect solution, but if you've got a situation which is increasing consistently in scale, don't we need technologies like this Sometimes? There's some really interesting examples within the discussion of the kind of tools that have been used. One of the things that really struck me, particularly about asylum is that the tools that are being discussed are actually tools that then will aid human decision makers.
So and there's a really interesting example about dialects that that that we chat about and it's, you know, what are the kind of skills that those decision makers have to be able to kind of deal with this new information that they're getting through.
And it also really reminded me of some fantastic research that Lena Rose, who was formerly of this parish, now elsewhere, did, looking at the way that we make credibility decisions and the fact that these are such incredibly complex and kind of sophisticated decisions on a human level about somebody's well-founded fear of persecution, for example, and the extent to which we are able to understand the kind of new information that's.
Coming in. Whether that is supportive of better decision making or in fact, you know, whether it's just kind of adding to what kind of pile of information and making those decision making process is actually much more difficult. I feel like we haven't necessarily just got to grips with the complexity of of making asylum decisions. And then we're adding this kind of whole extra layer on to it.
I think your point about actually this could remove some of the biases is really interesting, and I think it's something that could be explored a lot further. I know a lot of the worry around A.I. is that it far from removing biases, it kind of amplifies those biases because of the way that it's trained. As a AI, to my layperson's understanding is kind of trained on the stuff that is already out there.
And if the stuff that's already out there contains these these pre-existing biases, then it is also going to have those. So that's one of the ways in which it kind of becomes almost a little bit more human. So I think I just feel like these are incredibly interesting and important discussions and the fact that this is not something that we're projecting forward into the future. These are technologies that are being used right now. And so it's a discussion that that should be had.
In the same way that air is trained and learns so to people.
I mean, like air is not the only thing that learns. And so, I mean, we're constantly going to be in a situation surely, where there has to be a balance struck between the way that people learn how to how to either use the system or to manipulate a system or to play a system against the kind of the against the sort of the idea of the perfect decision making process that may exist either with humans making solid sound choices based on complete understanding of a situation which is never there,
or machines doing something based on perfect automated system. There's always got to be some kind of of balance between these things. I'm joined by Derek Ozekhome, senior research fellow at the Refugee Study Centre at the University of Oxford, and Catarina Ridley. EU Policy Analyst Access now. Daria, we hear so much about how I might transform our lives. Can you tell us about some of the ways in which it might transform and is already transforming migration governance?
Yes. So yeah, we started hearing more about A.I. when it started impacting our lives. So, for example, AI is transforming the workforce with many types of jobs now being automated. But it's also used in many other areas, including migration. And we see various practices that state authorities have started using on migrants. So right now from in fact, from the moment that someone thinks about even a migrant, the possible migrant thinks about even migrating to somewhere else.
There are data that Google searches on the news that they're looking for are all being recorded and monitoring monitors. And so, for example, predictive analytics systems helps states to forecast the number of arrival of migrants on their borders. And then these systems can then lead to more increased border controls and push backs of people who may need protection. We also see a variety of other types of new technologies being used.
For example, drones or various types of sensors in border areas for surveillance purposes, apart from those in recent years. Several states have started using new technologies to automate some of their casework processing. So these are often in the form of automated systems that take inputs from other databases. So, for example, in Norway, the immigration authority has automated the processing of citizenship applications.
And the way that they can do that is they, by taking data from that is already available against that individual from all the databases that they have, and then checking whether the person fulfils all the requirements to become a citizen. So the process can be fully automated if they have enough data about a person, basically. But just the more dangerous form of automation in this field is when immigration authorities are using what is called risk assessment systems.
And we have seen these types of systems, for example, in the UK or in Netherlands as well. In the UK case, thanks to huge efforts from civil society, it has been found that applications from certain nationalities were automatically categorised as high risk and that those applications were receiving a higher level of scrutiny from officers.
The civil society organisations, particularly the Joint Council for the Welfare of Immigrants and Foxglove as well, They found that the system had also a feedback loop problem, which means that applications from a certain nationality being rejected at a higher rate will also influence future applications from that nationality. So obviously these types of systems can create discrimination on the basis of nationality and it's quite risky and important to monitor them.
Absolutely. It's amazing how much is already happening. Catarina There are some big changes that are happening at EU level, but not seemingly with very much regulation. Can you tell us about how some of the technologies that debris are talked about are changing things in the Schengen zone and and why we might want to kind of regulate things to reduce some of the discrimination that we've heard about? Yes. Thank you. So I think that's an important starting point for orientate ourselves.
And this discussion is to reflect on the fact that new technologies in the era of EU migration policies are not that new. And in fact, they were actually rolled out and developed together with the new design of migration policies in the nineties. So in the nineties the Schengen area was, was started with 95, the abolishment of internal frontiers between European Member States and the a new European common external frontier was developed.
What civil society nowadays calls fortress Europe and digitalisation of borders and technology was since the beginning, since the get go crucial for the functioning of Fortress Europe, the first large scale database. In fact the Schengen Information System was deployed in 95 and it had the purpose of facilitating police cooperation among Member states as well as border management.
And it had information on different categories of people such as missing persons or people wanted for arrest in the same way for an implementation of the Dublin Regulation at the beginning of the 2000 stadia. That database was implemented and the Europe that year.
That database holds in source biometric information on asylum seekers, and it's fundamental for an implementation of the Dublin Regulation that should support Member States in distributing asylum seekers that arrive at the first countries at the European southern border.
Usually this is important to say that technology was key from the beginning and is important for implementing punitive migration policies that are very much focussed on enforcing deportation or impeding people from entering the European Union, even outside of the European borders. For example, in embassies, while a request applying for a visa, as there was explained before.
And if we look at the European policy landscape, there is a variety of different legislations that have allowed for the introduction of different technological solutions. We have European, we have we have migration policies such as policies, regulations that underpin migration databases such as fear that this which is a database for the information sharing on visas,
the Schengen Information systems and others. But we also have a regulation about interoperability of all of these migration databases. We have the new pact on asylum and migration that allows for the introduction of new technology, but we also have regulations on other type of labour, such as digital policies such as the Artificial Intelligence Act that also regulates the use of technology in the migration context.
We also have it in the context of law enforcement, such as the Europol regulation, that allows for more power for police forces to use air based systems and surveillance technology, and also legislations or initiatives about funding that allow for the money to invest on these type of technologies.
Thanks, Katerina. Daria, I am really interested in this difference between some of the processes that Catarina has spoken about, which I guess are around surveillance and collecting information and some of them that are kind of predictive and about making decisions. And one of the areas that we know where policymakers have to make huge decisions is in the asylum process, deciding whether somebody has a kind of well-founded fear of persecution.
And if you could tell us a little bit about how these technologies are affecting the asylum system, both for asylum seekers themselves and also for the people making the decisions. So in asylum decision making, we see we can see the introduction of some of the new technologies not to automate the decision, the whole process itself, but to automate some of the evidence that is used in decision making. So for example, one of them is used to determine the applicant's identity.
The immigration authority in Germany has been using this what is called a dialect recognition tool to identify the applicant's language. So this technology. It's quite very much like all the types of biometric technologies. It can identify a person's voice data in terms of percentages. So it generates the report basically saying that this person speaking speaks 64%, 17 Arabic and 16% golf Arabic and so on.
And it's really fundamentally I mean, this this particular technology is fundamentally changing the way language is assessed in asylum decision making, whereas in previously it was used, obviously it was always human linguists doing these kind of analysis, and they were recommended to make a qualitative analysis of a person's language.
So they would say, okay, this person is very likely speaking Lebanese in Arabic and then leave it there, whereas this new technology is giving percentages and is changing how then the decision maker is perceiving the applicants because it's 60 44% levels in Arabic, but that not so much the rest and so on. And another technology that is used more widely is mobile phone data extraction.
So a number of countries in Europe and elsewhere, such as Germany and Netherlands, for example, in Europe, have introduced this technology again to determine the applicant's identity, but also to identify their travel route and then use that as an evidence to determine whether or not the person is telling the truth. So it's of course, this technology is very invasive to the person's private life, extracting all the data available on their mobile phone.
But also, research in this area shows us that mobile phones are often exchange among migrants. They can be sold. They can be just used by their friends and family members. And all types of contradictions that come out from that can potentially be used against the applicant's credibility during the decision making process. So overall, we see that these technologies really create additional grey areas for applicants to address during the interview process.
Thanks so much, Terry. And Katrina, the Protect Not Surveil campaign is looking to make a human rights based argument for regulating some of these forms of new technologies, but it also banning some others, including some of the kind of predictive and profiling systems that we've talked about. Can you explain why the campaign is so worried about these forms of profiling and and what areas actually shouldn't be used in migration governance at all,
in your view? Yes, The particular survey campaign fits into a progressive society effort to throw some red lines on the use of artificial intelligence based systems in a way that irreversibly violates fundamental rights. And we see that in the migration context. People on the move and third country nationals are usually used as the testing ground for some type of technology.
So it's crucial that they are found some type of systems that should be banned not only relate to poor fighting type of systems, but also systems that amount to biometric mass surveillance.
So there is a called for a ban on remote biometric identification systems that could identify people in in on a remote basis and that could be used outside detention centre or at the borders to identify people that might have their biometric data, really not the basis systems such as the motion recognition systems that claim to infer the emotional status of a person's from their facial micro gestures.
And that reinforces this to make the suspicious suspicion against non European citizens and then profiling systems as the area was explained before, these systems that claim to assess the risk that a person posts are inherently biased. The example that Dario was making before from the UK, but also from the Netherlands of automated risk assessment used in the context of three ageing visa applicants.
Well, these type of systems are assessing how much the person might pose a risk to either public security or at risk to overstay or their visa based on some categories that are predetermined by some people that by nature cannot be objective because they build a system of risks that is embedded in some specific assumptions. So when it comes to risk assessments, you have some risk indicators that are pre-decided.
So some categories that are decided to be a factor of risks, for example, country of origin or level of education or type of employment. And then for each risk indicator, a screening rule is associated. So for each indicator, a different, the risk rate will be calculated.
So in the case of the Netherlands, for example, in the Three Eyes system for visa application, applicants from Suriname were systematically received systematically a higher score when it comes to the risk based on the country of origin. And the same went with the case from the UK. Applicants coming from certain African countries were systematically receiving a higher score when it came to the risk indicator of country of origin. But the same could apply also for level of education.
When people go to universities that have religious background, they might receive a higher score when their university is Muslim. University could give this idea that it's related to Muslim affiliation and there are many other thought of examples. So providing systems as well as other type of systems that protect not surveyed is calling for it, but are all those type of systems that would reinforce systemic oppression and forms of discrimination under the guise of technical neutrality.
A final question for you both. I guess a lot of this discussion is quite scary and difficult to hear about some of the risks in particular in relation to discrimination. But I wonder, are there any causes for optimism in terms of the use of these new technologies in the in the right hands? Do you think there could be any ways in which they could solve some of the long standing issues within migration governance? Is there any positive note on which we can end? Derek, I'll start with you though.
Some of the technologies that we have mentioned, such as automated processing of visa applications, can of course bring benefits around speeds, but those benefits are not equal for all. So some applicants can, of course, received a speedier response to their applications in one day because it's automated and it's low risk and they receive a positive response. But others can be under more scrutiny and probably receive a rejection.
But overall, we can see that most of these technologies have been designed in a way to benefit state authorities themselves to kind of ease and lessen their workloads. And they are not. I mean, the only way that they can be beneficial for migrants if they are, can it can happen if they are first thought about and designed out of migrants or needs.
And we haven't mentioned in this talk, but there are some technologies that are designed, for example, to match refugees with municipalities and areas that are best for their future employment or specific needs. There is, for example, the Merchant Imagine project in Germany, which is trying to basically match refugees needs with municipalities capacities across the country.
So that's a really good example of what technology can achieve if they centred around migrants and refugees needs themselves. But apart from this, I guess, I mean, we haven't talked about at all in this talk, but it's worth remembering that even if it's for the best intentions, every technology also brings new demands on natural resources and has consequences for the environment.
So we really need to be always asking ourselves whether we really need them and whether this is really cannot be done any other way. Catarina Any grounds for optimism? I think I want to be the one to bring the optimism in this conversation, but because I don't think the the framework where this debate is has is happening is, is the one that we bring safer solutions.
And I always think from the abolitionist movements, the idea that we don't need a reform in a system that is already is created to oppress certain people. So I don't think the question is around the positive use of technology, but about the policies themselves. Of course, the technology is not a problem. It's not a problem per say. We see some uses of technology, the same type of technology used in different contexts, for example, drones to detect people in distress.
If you put it in the hands of Frontex, it can lead to the facilitation of pushback through the Libyan coast guard, whereas also the NGO Sea-watch is using drones to detect the presence of people that are escaping from Libya towards Italy. And these, based on the detection of the drone, will then look locate their their vessels to start the search and rescue mission. But the thing is that technology is always used by someone that is always advancing their own priorities.
So when it comes to migration management, what needs to be done is rethink the whole infrastructure, the whole objectives behind the policies. And if we want to address the the problem of discrimination, the problem of violence that is happening also through the use of technology, we should be more bold and more courageous in calling for a systems that prioritise justice, that prioritise the ability and the freedom of people to move.
And that doesn't. That's not tied beyond technologies or opaque systems that in a way, the responsibilities, the authorities that are in fact legalising very violent policies that are closing that I thought were and within our borders. So not only optimism, but for for sure, I believe there is a way to to improving that. And I would like to close to refer to a report. Very interesting that actually brings some optimism, which was written by the Equinox Project.
And it's it's about Fortress Europe. And it actually details three ways in which we can challenge and change discriminatory EU migration policies. That seems like as good a place as any for us to leave it. Thanks so much to you both. You've been listening to the Migration Oxford podcast. I'm Robert Neill. And I'm Jacqui Broadhead.