Big data, surveillance, and migration: a neo-republican account

ABSTRACT

Big data, artificial intelligence, and increasingly precise biometric techniques have given state and private organizations unprecedented scope and power for the surveillance and dataveillance of migrants. In many cases, these technologies have evolved faster than our legal, political, and ethical mechanisms. This paper, drawing on current discussions of justice and non-domination, proposes a non-domination-based ethics of digital surveillance and mobility, in which the legitimacy of these technologies depends on their avoidance of the arbitrary use of power. This allows us to ethically assess new technologies and justify juridical, democratic, and administrative mechanisms.

KEYWORDS:

• Big data
• surveillance
• artificial intelligence
• migration
• neo-republicanism
• domination

State use of biometrics, artificial intelligence, and big data to surveil travelers is ubiquitous and troublingly banal. Travelers routinely hand over their machine-readable passports and submit to iris and fingerprint scans. This routine is quickly evolving. The Emirates airline’s ‘state-of-the-art, contactless biometric path’ has passengers pass through a ‘smart tunnel’ where their faces and irises are scanned (Emirates 2020). Nippon Airways has adopted Face Express’s contactless facial recognition technology and the EU has ambitious plans to expand its surveillance capacities under its Entry/Exit System Regulation and the European Travel Information and Authorisation System (Macola 2021). The Emirates’ promotional materials emphasize convenience, flexibility, and safety; they make no mention of the implications for freedom, privacy, or discrimination. In a 2016 pilot project, the Canada Border Services Agency scanned millions of travelers’ faces without their knowledge as they passed through Toronto’s Pearson International Airport, searching for matches in a database of previously deported people (Cardoso and Freeze 2021). What was recently a dystopian fantasy of preemptive policing and total surveillance has become an emerging reality that ethicists and political theorists must confront.

The surveillance of mobile people serves as a canary in the coalmine, alerting us to the harms of technologies and practices such as facial recognition, risk assessment algorithms, and dataveillance. Technologies such as facial recognition, initially developed to screen travelers at borders, are easily transferred to public agencies and private corporations and deployed in areas such as domestic law enforcement or customer surveillance (Israel 2020). Border control can normalize morally problematic practices before they are imported into other domains of life.

We need to be able to distinguish between types of surveillance that are justified (or could be justified with appropriate safeguards) and types of surveillance that abuse power. This is particularly urgent since advances in artificial intelligence and biometrics have greatly expanded the capacity and range of surveillance. Moreover, technologies have the disturbing tendency of function creep, in which they are applied to new, often unanticipated areas, shifting from corporate surveillance and dataveillance to domestic law enforcement or from domestic organizations to international jursidictions. We need theoretical tools to help us to evaluate future, possibly novel, technological capabilities.

A normative account of just surveillance and dataveillance should be able to offer a framework to held guide us against future abuses. I propose addressing the moral challenges arising from biometrics and big data through a neo-republican account of non-domination. The just employment of these technologies requires institutional mechanisms to prevent arbitrary exercise of power by government and corporations. Freedom from domination in the form of arbitrary exercise of power via surveillance does not mean that agents can never justly exercise power to regulate migration; rather, there must be juridical, democratic, and administrative mechanisms that make the exercise of power non-arbitrary.

Migrant surveillance

John Gilliom and Torin Monahan define surveillance as ‘monitoring people in order to regulate or govern their behavior’ (italics in original) (Gilliom and Monahan 2013, 19), emphasizing how watching (French: surveiller or ‘to watch from above’) serves as a technique for the exercise of power. This is often achieved through biometric surveillance, the collection or generation of biometric data – data about physical characteristics, health, genetics, and other biological characteristics and properties – in order to facilitate governance. Gillion and Monahan define biometric surveillance as ‘the systematic monitoring, gathering, and analysis of information in order to make decisions to minimize risk, sort populations, and exercise power.’ (Gilliom and Monahan 2013; cited in Kalhan 2014: fn4).

Surveillance is necessary for essential state responsibilities, such as national defense, taxation, social welfare, public health, and managing elections. Every aspect of migration management involves surveillance, including border crossings, visa applications, risk assessment, asylum management, and immigrant detention. It is employed before admission to screen applicants, at admission to verify their identity, and after admission to police immigrants within countries.

The rise of big data built on increased storage, processing power, and sophisticated algorithms has given state and private organizations unprecedented scope and power. Data is collected from multiple sources, including official government biometrics, private company records, and, in many cases, individuals’ social media use and telecommunications. The ability to comb through billions of social media posts, financial transactions, location data, and much more, creates unprecedented forms of dataveillance, which Roger Clarke defines as ‘the systematic use of personal data systems in the investigation or monitoring of the actions or communications of one or more persons’ (Clarke 1988, 499). Big data policing, built on long histories of racialized surveillance, is particularly troubling (Browne 2015; Ferguson 2017). Not surprisingly, immigration control continues this ongoing legacy.

Biometric and information systems have dual functions of facilitating border crossing (e.g. through preferred travelers’ programs) and preventing them (e.g. through no fly lists) (Van der Ploeg and 2006, 179). These dual functions create normative complexity. Sometimes state surveillance is legitimate and beneficial; without it they could not reliably and fairly deliver services and benefits. But too often, surveillance violates the right to privacy, contributes to unjust forms of migration management, and in some cases risks migrants’ lives, especially when used to screen, serve, and police refugees.

For the normative assessment of these technologies, I draw on work in the ethics and political philosophy of migration that mobilizes republican theory and the appeal to non-domination to evaluate immigration policies and practices (Benton 2015; Costa 2020; Fine 2014; Honohan 2014; Hoye 2018; Rochel 2021; Sager 2014; Sager 2017). Neo-republicans hold that domination occurs in relationships in which one agent has the capacity to unjustifiably exercise power over another agent. On the neo-republican account of freedom, it is not just actual interference that restricts freedom, but also the capacity to interfere.

Neo-republicanism only sees the arbitrary exercise of power as problematically limiting freedom. Though theorists debate about what precisely makes the exercise of power arbitrary, the core idea is that arbtirary power is discretionary power unconstrained by the party wielding power having to answer to the person they are wielding power over or to broader democratic, legal, or social constraints. A slave owner who can beat or kill his slaves if they refuse his orders exercises arbitrary power; an agent collecting taxes for a government agency authorized by the legislature exercises power, but it is presuambly non-arbtirary. Prominent neo-republican accounts dispute whether it is possible to render power non-arbtirary through purely procedural constraints (Lovett 2012); whether it is also necessary that people have control over how power is exercised over them (Pettit 2012); or if the exercise of power must track people’s interests (Pettit 1997).¹ Rather than attempt to mediate between these theoretical differences, I focus on mechanisms that allow people to shape and contest the policies, practices, and decisions that affect them. Domination can be mitigated or eliminated – i.e. made non-arbitrary – when people have sufficient access to democratic institutions (e.g. opportunities to vote, lobby, appeal to decision-makers, and protest), legal protections (e.g. civil rights, property law), and bureaucracies governed by public, well-defined rules with internal procedures.

I follow Dorothea Gädeke in conceiving domination as a ‘structurally constituted form of power.’ (2020, 199) Gädeke draws attention to how domination is not simply an interpersonal relationship. Domination occurs because legal rules, as well as informal social norms and practices, make some people powerful and others vulnerable. In the paradigmatic republican cases, slavery and patriarchal marriage, the dominators (slave owners, husbands) and dominated (slaves, wives) are asymmetrically situated in legal and social systems that deny the dominated tools to shape, contest, or resist how power is exercised over them. These legal and social systems define the subordinate status of dominated classes of individuals.

To legal and social systems, we can add technological systems. Surveillance is embedded in the technologies that enable us to travel (biometrics, passports, facial recognition systems, phone aps). In migration, these technological systems are inseperable from the construction of categories and identities. Whether we are citizens, temporary workers, refugees or ‘illegal aliens’ depends on how our data is captured, categorized, and mobilized. People are reduced to their data and assessed for risk, often with the aid of AI that develops algorithms decision-makers understand only vaguely, if at all. Opting-out of these technological systems is rarely possible without incurring a prohibitive cost. The result is that people are situated asymmetrically to the governments or corporations who use these technological systems to surveil, manipulate, and discipline them.

Gädeke’s radical republican account helps highlight the problematic ways in which immigrants are put in positions of vulnerability. First, the political status of immigrants renders them vulnerable to abuse by depriving them of tools to shape and contest decisions. This observation is true for most, if not all immigrants, but starkest for many refugees, especially those in camps and in transit with the goal of seeking asylum and official refugee status. Second, the legal safeguards necessary to mitigate the potential abuses of surveillance are often absent. Immigration enforcement operates under a security paradigm where core rights are abrogated and procedural safeguards fail to meaningfully curb abuse. Furthermore, many of the technologies are deployed without appropriate regulation. Third, the implementation of biometric surveillance creates epistemic barriers to effective oversight. There are rarely measures that adequately preserve transparency and accountability. Moreover, in many cases in which AI interacts with large databases, we do not understand how AI arrives at its outputs. Fourth, the biometric surveillance of immigrants occurs under conditions of structural injustice in which policies and broader social norms have frequently served to uphold white supremacy.

Domination and migration

Biometrical surveillance and dataveillance take place within structures that make immigrants vulnerable to political domination. A central claim in republican political thought is that popular control is necessary to render state interference with people’s freedom non-arbitrary. For contemporary neo-republicans, popular control requires a combination of the power to elect representatives and significant opportunities to contest decisions (Lovett and Pettit 2009, 25). More broadly, republican freedom demands that people subject to power have the right and the legal and social status necessary to actively participate in decision making.

Non-citizens are typically excluded from the franchise and thus unable to elect representatives.² While they may have recourse to juridical and bureaucratic procedures and can influence elections through proxies, popular appeal, protest, and direct action, their lack of equal status and standing significantly impairs their influence. While we should not underestimate immigrants’ political agency, they often incur significant risk in exercising it, in many cases risking deportation. Philip Pettit refers to the ‘eyeball’ test in which people ‘can look others in the eye without reason for the fear or deference that a power of interference might inspire; they can walk tall and assume the public status, objective and subjective, of being equal in this regard with the best.’ (Pettit 2012, 84). Immigrants are rarely in a position to look decision-makers in the eye, knowing that these decision-makers have broad discretion to deny their visa application, detain, or deport them.

The asymmetry of status and political rights is particularly stark for refugees, for whom Hannah Arendt’s observation of the importance of ‘the right to have rights’ is critical (Arendt 1943). For example, biometrics are embedded in the UNHCR’s turn to ‘results-based management’ as a tool that improves accountability to donors and – allegedly – to refugees under its mandate (Jacobsen and Sandvik 2018). The UNHCR’s decision to make humanitarian assistance for Afghan refugees repatriating from Pakistan contingent on their participation as subjects in an experimental iris-recognition technology is troubling (UNHCR 2010). This initiative was justified on the grounds that iris-recognition technology would allow the UNHCR to weed out ‘recyclers’ or ‘false refugees’ attempting to receive a second aid package (Jacobsen 2015, 149). At the time, there were no previous attempts to employ this technology at the scale and under the environmental conditions of Afghanistan, raising concerns about its reliability. Indeed, attempts to anonymize biometric refugee data had the unfortunate side effect of creating an obstacle for refugees who received a false match; since the iris template could not be connected back to the person who gave it, refugees could not contest errors (Jacobsen 2015, 152).

The UNHCR failed to uphold basic standards of research ethics in conducting their experiment on a vulnerable population. As this example illustrates, their rights are protected – or, as is too often the case, violated – at the whim of receiving states and international organizations. Refugees camps, in which such experiments with tech are often initially deployed due to the presence of a ‘captive’ population, are aptly described as spaces of domination, stripping political agency from refugees, stranding them in isolated areas and subjecting them to surveillance in the name of humanitarianism. Michel Agier has observed that in the humanitarian governance and refugees, ‘there is no care without control’ (Agier 2011, 4). To this we can add that those controlled are rarely granted means to contest and shape their care.

Political participation needs to be combined with rule of law, due process, and a robust set of constitutional rights that limit and check government and private interference. People must be equal under the law. Legal rights need to be vigorously protected, so that their exercise is transparent and agents exercising power face meaningful accountability. Non-domination mandates access to clearly enforced procedures that give people meaningful recourse when immigration authorities violate their rights, incorrectly apply rules, or abuse power.

The fact that immigrants as non-citizens do not enjoy equal status is intensified by the nature of legal regimes at borders. States predominately treat immigration under a security paradigm, in which biometrical and information technologies are justified as serving state security or combatting terrorism (van der Ploeg 179: 191–192). For example, UN Security Council (UNSC) Resolution 2396 (2017) calls on Member States to establish advance passenger information (API) systems and require that airlines provide this information to national authorities, collect passenger data, develop watchlists of known and suspected terrorists, as well as develop systems to collect biometric data (paras 11–13 and 15) (OSCE Office for Democratic Institutions and Human Rights 2021). The upshot of this resolution is that all passengers are surveilled and scrutinized as potential security threats (in contrast to an approach where people would be considered potential security threats only if there is cause to suspect them).

Furthermore, many legal mechanisms and rights that are considered fundamental within state territories are waived or abrogated at the border. Most immigration applications require biometrics, which are typically retained after immigrants become citizens.³ Immigration officials and border agents have extraordinary discretion and travelers have limited opportunities to contest decisions. Travelers in airports must submit to having their person and belongings searched without reasonable suspicion. Merely possessing a visa does not protect a traveler from being turned away at the border by border guards without explanation. Again, refugees are particularly vulnerable to involuntary biometric collection and to forms of algorithmic governance (Achiume 2020). They are rarely in a position to meaningfully consent to the collection of their biometric information, since refusal means the denial of (sometimes lifesaving) services.

Further barriers to accountability come from what Anil Kalhan calls the ‘migration border’, which is distinct from the territorial border (Kalhan 2014, 9). States have moved border controls far away from their territories, deploying legal maneuvers to prevent migration, while abdicating responsibility for their human rights responsibilities (Shachar 2020). Borders have been displaced, deterritorialized, delocalized, and/or extended into airports and consulates around the world where people are denied visas or permission to travel abroad, into other states where governments or private security forces detain migrants in transit, or across the ocean where coast guards intercept refugees and return them to places where they cannot lodge a claim to asylum (Casas-Cortes, Cobarrubias, and Pickles 2015; Longo 2018).

The border is in many respects constituted by the set of technologies and practices that allow for the surveillance and sorting of people thousands of miles away from territorial boundaries. Biometric information feeds into often problematic forms of algorithmic governance, creating epistemic obstacles that compound the political and legal disabilities that immigrants confront. Artificial intelligence can create an illusion of objectivity that protects decision-makers from scrutiny. Algorithms are often incorrectly presented as ‘objective’ and not subject to the biases and errors of human judgment, encouraging authorities not to scrutinize their reliance on risk-assessment tools (Molnar 2018).

A particularly troubling example is the Department of Homeland Security’s ATLAS software, which takes case files and runs them through various federal databases to check for fraud. This can be used to justify deportation or to denaturalize US citizens (Biddle and Saleh 2021). The Department of Homeland Security has refused to share with the public how the algorithm works. In another case, reported by the American Bar Association, risk assessment software meant to inform decisions about whether to release immigrants from detention recommended that everyone be detained (Tashea 2018). As in many cases of governance powered by machine learning, literally no one knows exactly how the outputs are generated, undermining accountability (Leese, Noori, and Scheel 2022, 9)

Technology is only one source of epistemic obfuscation. Since border control and immigrant surveillance crosses territories, its governance is necessarily transnational. Immigration authorities rely on documentation from other countries and on sharing information and intelligence (Longo 2018). A further complication comes from how the boundaries between public and private blur and meld. Private actors (such as airlines) are deputized to collect data. State intelligence agencies incorporate private databases, often by purchasing them from data brokers, which allows them to circumvent legal restrictions on surveillance.

Serious obstacles to accountability arise when databases are constructed and rendered interoperable from multiple data sources. Leese et al. refer to ‘the opacity of knowledge production’ (Leese, Noori, and Scheel 2022, 9). Not only will a border guard not know why a particular traveler has been flagged by the system, but it may be that no one knows how the system assesses risk. The right to know if one is included in the database and to contest one’s inclusion is a necessary safeguard. It can become practically impossible to correct false information when databases are shared and combined. Not only may immigrants not have the standing to contest decisions made by state authorities, they may not even be able to determine who or what is responsible for a decision to deny them a visa or entry. These epistemic barriers compound the issues of political status and rule of law.

Finally, surveillance technologies are embedded in histories of racial domination and exclusion (Browne 2015). Margaret Hu points out that ‘systems of subordination, such as Jim Crow and Apartheid, depend on classifying who is privileged and who is not … once a clarification is emplaced, laws can require separation, subordination, and screen on the basis of that classification.’ (Hu 2017, 657) Big data opens up new forms of classifying, surveillance, and discriminating against classes of people. Immigration decisions affecting individuals draw on aggregated data. Risk-profiling software will flag individuals not because of their behavior or unique characteristics, but because they share on characteristics with a ‘profile’ constructed on the basis of aggregated or group information.

Algorithms give a veneer of objectivity which disguises the ways in which forms of social discrimination are built into the methods and datasets used to create them. Even if the techniques used do not explicitly attempt to group people according to race, ethnicity, nationality, sex, sexual orientation, or religion, machine learning may latch onto proxies such a zip codes, education level, or occupation that sort people according to these characteristics. Using big data to vet immigrants will not be limited to people seeking to visit or settle in the country (Hu 2017, 641). It invariably involves surveilling groups of people, including citizens, connected to the target population, either because they share an ethnicity or other forms of connection (e.g. on social media). Since facial recognition has well-known racial and gender biases, its use to scan travelers builds discrimination into border control (Israel 2020). These technologies interact with institutional and structural racism, which amplifies prejudice and systemic injustice.

Neo-republicanism and just surveillance

What would be needed to render the biometrical surveillance of migrants non-dominating? For the most part, biometric surveillance of migrants would need to be brought in line with just surveillance of domestic populations. The protections that migrants need so that surveillance does not dominate them are often similar to the protections that everyone needs. I stress just surveillance, since a great deal of current surveillance of non-migrant populations is unlikely to pass muster. Despite a few promising (if perhaps limited) measures such as the EU General Data Protection Regulation (GDPR), mechanisms to mitigate domination created by biometrics and big data are not in place. Even in examples such as the GDPR, there are significant limitations. For example, the GDPR focuses on personal data, leaving largely touched the ways in which aggregate data may be mobilized to dominate groups and individuals (Hoofnagle, Van Der Sloot, and Borgesius 2019). Furthermore, immigration law and policy create legal spaces that disempower the people caught inside them.

In all cases of biometric surveillance or dataveillance, we need to ask if they are justified in the first place. Just because we have the capacity to collect and indefinitely save massive amounts of data doesn’t mean that we should do this. Does the collection of this data serve a legitimate purpose? Can the purpose be reconciled with demands on privacy and the risks of abuse? Is the type of biometric surveillance or dataveillance required to achieve this purpose or are there other less invasive means available? Are the tools employed proportional to the task or do they open up other, problematic applications? Given that immigrants do not enjoy the same rights and protections as citizens and because many of them are members of groups routinely subjected to discrimination and abuse, there is a strong case for heightened scrutiny of their surveillance.

It is often the use of technology, rather than technology itself, that is the problem. Technology may play a role in warding off domination, for example the use of social media to monitor politicians or to organize. Nonetheless, we should be wary of techno-utopian solutions. As Margie Cheesman warns, states and corporations can reappropriate highly touted digital identity tools that use blockchains to sidestep the need for centralized identification, especially when these technologies are depoliticized and torn from the contexts in which they are used (Cheesman 2020). Technology needs instead to work in conjunction with political and legal measures to enable migrants to shape regulations and to contest power.

The necessary blend of law, policy, and institution building that would be needed to render biometric surveillance non-dominating is complex and no doubt can only be full worked-out through implementation and careful assessment. Nonetheless, we can gesture at types of political, legal, and epistemic protections and mechanisms of empowerment. First, political measures are needed. Domination cannot be alleviated without establishing equal status. This poses a major ethical challenge to immigration policy and enforcement: the categories of immigrant and non-citizen are constructed to ensure that non-citizens do not have equal status. One way forward would be to adopt an open border position, where differences in rights and obligations would come from situational factors (e.g. present place of residence), rather than individual status (e.g. citizenship) (Sager 2020). Such a policy would not mean that states cannot exercise control over their borders (for example, by requiring people cross borders at official checkpoints or by limiting entries in any given time frame); it would only mean that any actions taken at the border must treat everyone as having the same status as citizens, and that the same mechanisms that citizens have for redress are extended to everyone regardless of national origin.

Less radically, equal status might be achieved through the creation of institutions that allow for genuine participation and contestation that fall short of opening borders. It is possible that a combination of stronger domestic protections for immigrants, meaningful representation from countries of origin, and supra or transnational oversight could ameliorate the most problematic aspects of biometric surveillance. Migrants would at a minimum need a meaningful voice in decisions to collect and use their data, as well the nature of the legal, social, and technological systems that surveil them. These systems would need to take into account their needs and interests, rather than prioritizing the needs and interests of humanitarian organizations, tech startups, or states (Lemberg-Pedersen and Haioty 2020; Madianou 2019). The collection of biometrics should not be a default or a matter of convenience, but rather the result of a deliberative process that amplifies the voices of those most affected. If biometrics are collected, immigrants would also need some form of access to data collected through biometric surveillance; some recourse to appeal, request reconsideration, or at least lodge a protest over how these data are used in making decisions about status, entry, and the like; and some say in how these data are deployed in policing and controlling borders in order to meet the non-domination condition.

Second, laws should be designed with the goal of minimizing the potential for domination.

Legal mechanisms need to be implemented that define what information can be collected, how it can be used, how it is stored and protected, who can access it, and how long it is kept. When there are exceptions to rules, there needs to be clear and impartial oversight. The legal system provides opportunities to mitigate and contest domination, for example, through the right to be forgotten, right to obtain information that companies or the government have collected about us. Of course, these opportunities are only meaningful if people can actually access and use them, so implementation is crucial.

Third, many of the obstacles are epistemic. We cannot contest abuses if we don’t know about them. Transparency and the accessibility of information should be the default, with secrecy subjected to rigorous judicial and political scrutiny. There should be clear and straightforward procedures that allow people to know what is being collected (e.g. mandated notification), verify its accuracy, and demand, when appropriate, that information be removed. The use of AI to screen applicants for suitability or risk assessment should be continuously monitored and supplemented with independent human oversight; when the potential for abuse is widespread or severe, it should be abandoned. This cannot be achieved without international cooperation and supranational oversight to set international standards and rules for the collection and use of this data, and for monitoring how states conduct biometric surveillance at their borders.

We have yet to learn if the use of biometrics and big data for the surveillance and dataveillance of migrants will morph into a full fledged dystopia. The domination-based analysis presented in this paper offers a framework for evaluating these technologies and their applications and for guiding our thinking as we develop legal regimes and forms of democratic control that empower people – mobile and otherwise – to shape and contest how surveillance is carried out in social, legal, and technological systems. This analysis may also be helpful when thinking about how innovations developed for immigration enforcement are applied to other domains of life, often for the surveillance of other vulnerable groups (Kalhan 2014, 10–11). When faced with a new technology or novel technical applications, we can ask: does this technology or it application directly dominate any individuals or groups or amplify broader structures of power that dominate? What mechanisms are available or could be created to allow the individuals or groups to shape or contest the power relations that the technology enables? If we are unable to address these questions, just surveillance will not be possible.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Alex Sager

Alex Sager is Professor of Philosophy and Executive Director of University Studies at Portland State University. He is the author of Against Borders: Why the World Needs Free Movement of People (Rowman and Littlefield International, 2020) and Toward a Cosmopolitan Ethics of Mobility: The Migrant’s-Eye View of the World (Palgrave Pivot, 2018).

Notes

1 For discussion, see Arnold and Harris 2017.

2 The connection between citizenship and political rights is not a necessary one and there are many examples of non-citizen voting. Pedroza (2019) provides an important overview and analysis of immigrants’ voting rights around the world.

3 Margaret Hu (2013, 1523–1528) provides a valuable overview of government biometric database programs and immigration-related biometric screen programs.