3.1 Introduction
Much technological governance scholarship, including that which analyses digital borders, focuses on how digital technologies differentially constitute individuals as members of broader publics. For example, ethnographies evaluate how migrants are sensed and scrutinised by the tools regulating European Union borders and how this treatment is distinct from the experience of EU citizens.Footnote 1 Social histories illustrate how data-centric tools create ‘the norm’ as a category of people and, in doing so, distinguish others as deviant, or apart, from that norm.Footnote 2 While this work evaluates how subjects are individually treated, it also suggests how this treatment constructs subjects as broader publics and political communities.
The digital technologies regulating state borders construct another equally important, but less often explored, subject: the state agency. State agencies are crucial legal subjects. They create geopolitical borders by delineating spaces where sovereign power is exercised. They are also the targets of conventional accountability measures. In administrative law, for example, someone who is mistreated at the border might challenge that treatment in court and seek remedies against the agency or ministry responsible for that treatment. Yet, when digital tools contribute to the governance of borders, state agencies – and the agency or power of the state itself – are reconfigured. This reconfiguration complicates legal accountability mechanisms even as it concentrates the effects of state power on some (often already marginalised) people.
Where and how might we locate digital border agencies? Through which traces, sites, and actors (human and otherwise)?Footnote 3 This chapter takes up the challenge of answering these questions. It investigates how the digital border’s socio-legal-technical infrastructure enables new relations and reshapes institutional practices and dynamics. In doing so, it highlights points where distinctions between human, non-human, and institutional processes are redrawn to suggest how dispersals of authority and reconfigurations of agency through infrastructure play out. Throughout, it emphasises not just the social and technical but also the legal intricacies of border infrastructures, as legal technicalities crucially impact how borders operate.
Drawing on a recent Canadian example, this chapter argues that digital border infrastructures focus and diffuse state agency (or agencies) in ways that fundamentally challenge administrative law and the legal subjectivity of state agencies. In developing this argument, the chapter intentionally plays with the double meaning of the term ‘agency’. Agency evokes the ability of state actors to generate effects; it also encompasses institutions (ministries, departments, authorities) that are agencies in an administrative law sense. Administrative law is the chapter’s focus because, in Canada, it remains the tool that is most often used to hold border authorities accountable for their actions. Yet, as this chapter will suggest, contemporary administrative law may be ill-equipped for this task if we take the infra-legalities of digital borders seriously.Footnote 4
This chapter first explores state agencies as legal subjects, positioning them among other popular subjects in digital or dataist state scholarship.Footnote 5 It then critically assesses how state agency (i.e., the state’s capacity to effect action through its practices) contracts and expands, using a recent digital border conflict to illustrate the argument. Finally, the chapter urges scholars and advocates to take digital infrastructures and algorithmic violence seriously. To that end, it encourages the pursuit of robust horizontal accountability strategies that are as networked and diffuse as the digital governance phenomena that they seek to regulate.
The digital border is a useful example of the simultaneous focus and dispersal of state agency because it contains many features common to digitalised administrative operations. Like initiatives explored elsewhere in this volume, the digitalisation of border administration has been underway for at least two decades.Footnote 6 Digital border infrastructures combine older list-and-algorithm tools and newer anticipatory machine learning tools (or are piloting these tools for near-future use).Footnote 7 The Canadian example fits somewhere between old(er) and new(er) digital border infrastructures. Sophisticated machine learning may operate at some nodes in the infrastructural web, but many of the tools described below rely upon rudimentary risk models and decision tree architectures. Ultimately, when tracing how digital bordering infrastructure fosters new relations and shifts institutional practices, the presence or absence of a specific algorithmic tool may matter less than how multiple tools, together with officials, agencies, and others, interact to focus and disperse authority.
This chapter thus conceptualises digital bordering infrastructure expansively and focuses on infrastructural relations. The infrastructure described below includes ‘technical’ things (databases, software, and interoperability mechanisms, for instance), but extends beyond them to encompass practices (such as the practice of assessing risk), state officials (immigration and border officials, even judges), institutions (ministries, departments), and ‘non-state’ actors (such as airlines). This arrangement of tools, practices, institutions, and people often fades into the background of conventional legal analyses, which privilege some elements (state officials, for instance) while overlooking others. A conventional legal approach, in other words, simply cannot robustly envision or evaluate the simultaneous concentration and diffusion of state agency that is my concern here.
3.2 State Agencies as Legal Subjects
Literature on data, technology, and law tends to address a unique subject: the people who are the subjects of data collection and regulation. This focus may be linked to the recent introduction of the ‘data subject’, a legal category with roots in EU data protection law. This legal device encapsulates and protects the person whose data is captured and stored by a variety of technologies (websites, automated decision-making systems, etc.). While the data subject concept has become so pervasive that it is sometimes shorthand for any (natural) person who is the target of data collection practices,Footnote 8 some scholars question its universality. Fleur Johns, for instance, argues that data-driven tools shift subjectivities in humanitarianism initiatives in ways that are distinct from the European concept of a data subject. She identifies a new ‘tinkering’ subject generated through the novel sensorial capabilities of contemporary institutions like Pulse Lab Jakarta.Footnote 9 Others examine how being the subject of data collection and processing impacts subjectivity itself. To constrain these effects, some scholars propose that the ethical frameworks governing ‘human subjects’ research should extend to protect those impacted by data science initiatives.Footnote 10
At the same time, a growing body of scholarship across the humanities and social sciences analyses the links between subjects, objects, and power with (re)new(ed) vigour. New materialism literature, for example, sharpens how we might distinguish between (legal) subjects and objects.Footnote 11 This work challenges conventional notions of who, or what, might be responsible for outcomes and who has (or ought to have) agential power. Critical new materialism scholarship, in particular, traces how subjects and objects are enmeshed. Its aims are both illustrative and political: the goal is to show how power functions in the world to make critical political economy analyses (and presumably their transformational outcomes) possible.Footnote 12 Similar observations are foundational to the arguments of critical data scholars, which show how data is always constituted in relation to humans, institutions, and technological infrastructures.Footnote 13 Within socio-technical scholarship more broadly, interactionist and infrastructural scholars such as Lucy Suchman and Susan Leigh Star have studied how subjects and objects (from photocopiers to scientific categories) are enmeshed, raising questions about agential relationships and power.Footnote 14 Far from esoteric, such lines of inquiry are vital to reconceptualise how technologies govern, how objects act, and how legal accountability might operate effectively.Footnote 15
Traceable throughout these literatures is a distinct, often understated, subject: the state agency, an equally important subject for digital government and for legal accountability. In digital government initiatives, state agencies work closely with algorithmic tools. They may remain the face of the state even when decision-making authority becomes distributed across a network of actors. Yet, state agencies are less often directly scrutinised. Data subject literature largely assumes that state agencies exist and are (or ought to be) responsible, at least in part, for regulating how data subjects’ data is collected, stored, and used. In new materialism literature, meanwhile, the peculiarities of state agencies may recede behind a flattened topography and network of actors.Footnote 16 State agencies are vital actors in many digital government projects, however.Footnote 17 Their status as such is especially evident in border contexts, where administrative agencies plus other actors exercise sovereign power on behalf of states with often devastating consequences.
State agencies are crucial legal subjects. In administrative law, their subjectivity has two intertwined elements: first, state agencies subject others to the force of law, and, second, state agencies are the subjects of legal mechanisms to constrain the worst excesses of state power. The power or authority of state agencies is tempered by a mix of statutes, legal principles, and constitutional or human rights, each of which may impose conditions that state agencies must meet for their actions to be lawful.Footnote 18 State agencies remain the subjects of these accountability mechanisms even though their activities have evolved over decades of ‘digital-era government’ initiatives.Footnote 19 For example, anyone who contests a digital border decision through court proceedings must name the correct state agency (or ministry responsible for that agency) in their pleadings documents. This naming process is essential for vertical legal accountability. It ensures that any remedial orders issued by a court are legally enforceable against high-level state actors (government ministers, for example). It also assumes that these actors have the authority necessary to ensure that front-line officials within a particular administrative agency abide by the terms of a court order.Footnote 20
The notion of an identifiable, accountable state agency is deeply embedded in domestic legal systems and in theories of public law. For this reason, legal scholars may assume that a (legally) responsible state agency exists and can be found somewhere within a digital border infrastructure. Recent literature on accountable algorithmic administration remains focused on vertical accountability mechanisms that require top-down implementation to ensure fair processes and appropriate outcomes.Footnote 21 Vertical accountability is also a foundational assumption within digital constitutionalism literature,Footnote 22 as well as in mainstream international law literature on algorithmic governance.Footnote 23 When we examine digital border infrastructures more closely, however, the ‘state agency’ quickly becomes a problematic accountability device, not least because many actors contributing to digital border decisions fall outside of its purview.
When state administration is operationalised through socio-legal-technical infrastructures like those that constitute the digital border, a legal accountability paradox arises. This governing arrangement intensifies the visceral effects of state power, while simultaneously diffusing state agency among many agencies, making it harder to ‘name, blame, and claim’ them.Footnote 24 People feel the undeniable sting of state authority facilitated through digital tools, but if they try to identify a responsible entity (in a legal sense), they often find themselves punching a shape-shifting cloud of wasps. Here, the swarm is not nested in the sovereign (the state agency responsible for the border decision).Footnote 25 Rather, the swarm (officials, software, hardware, and other actors channelling state authority onto someone) momentarily holds the form of the sovereign before dispersing when anyone seeks to identify its components as legally accountable state agents.Footnote 26 The ‘sharp edges’ of digital borders are created by no one in particular.Footnote 27 Responsibility and agency are dispersed (or deferred) across a socio-legal-technical infrastructure.Footnote 28
This paradox is likely unsurprising to socio-legal scholars, who have convincingly unsettled the assumption that state power is unidirectional or that conflicting modalities of governance cannot coexist.Footnote 29 Nonetheless, their insights have yet to substantially influence legal scholarship more broadly. Consequently, calls for accountable digital government continue to emphasise traceable vertical delegation and responsibility, which assumes sequential decision-making, distinct human agency and discretion, and clearly identifiable agents or agencies.Footnote 30 These proposals rely on a ‘liberal proceduralism’ that is drifting far from the distributed arrangements they seek to regulate.Footnote 31 Administrative law can evolve. The key challenge is how existing legal accountability frameworks designed to constrain administrative action might become responsive to the emergent realities of digital government infrastructures.
Inspired by Gavin Sullivan’s infra-legalities approach and its invitation to consider how socio-technical infrastructures reconfigure actors and reshape legal practices and power,Footnote 32 the next section illustrates how digital border infrastructures concentrate and scatter agency across the web of entities governing the interaction between a traveller seeking to visit Canada and state authorities. In the process, it emphasises the socio-legal-technical nature of digital border infrastructures.
3.3 The Concentration and Diffusion of State Agencies in Digital Border Encounters
Encounters involving pre-approved temporary resident visas demonstrate how state agency becomes simultaneously concentrated and diffused. This section first sets out important technical details about Canada’s digital border infrastructure, then describes the case of Smail Khaniche who was forcefully denied entry to Canada in 2018 despite his pre-approved visa.Footnote 33 This experience is then critically analysed to show how digital bordering infrastructures enable new relations and reshape institutional practices. This critical assessment shows how digital border infrastructures intensify and disperse state power in ways that frustrate external review. In the process, responsibility and agency, and relationships between state agencies and other institutional entities, are also reconfigured.
3.3.1 Multiple-Entry Temporary Resident Visas: Legal and Other Technicalities
Today, a network of decision-making agents determines whether visitors to Canada can legally enter the country. For example, most overseas visitors must apply for and receive their visas digitally before arriving at a Canadian airport. Enforcement of this rule is delegated to airline staff, who prevent travellers from boarding a flight if their visa approvals are not already in place.Footnote 34 Travellers who wish to regularly visit Canada (to see family, for example) may apply for a temporary resident visa, a type of visa which allows many entries over a multi-year period.Footnote 35 Like other visas, the application is submitted online to the federal Ministry of Immigration Refugees and Citizenship Canada, or ‘IRCC’.Footnote 36 To support their application, applicants must provide personal information (including proof of relationships if they are travelling to visit family), evidence of financial security, vaccination records, and biometrics (facial photograph, fingerprints). They must also pay a minimum $100 CAD fee. If approved by IRCC, travellers receive their multiple-entry visa documents by email and can use them to travel to Canada numerous times before the visa expires.
An important legal technicality governs these travellers, however. This visa merely confirms that they are eligible to enter Canada. Upon arriving at a Canadian airport, travellers must still complete an in-person interview with a border official, which will determine whether they are admissible to Canada.Footnote 37 Only after a traveller completes this interview will a border official formally find them admissible (or inadmissible) to Canada.Footnote 38 If admissible, the official will stamp the traveller’s passport to show the date by which that person must leave Canada. This stamp is a mark of the Canada Border Services Agency (or the ‘CBSA’, an agency distinct from IRCC and accountable to a different federal Minister). If inadmissible, the traveller may be barred from entering Canada for at least five years.Footnote 39
All of these encounters between temporary residents and Canadian federal officials are recorded in and mediated by the Global Case Management System (‘GCMS’). Some basic elements of this tool, and its links to other digital border components, are crucial for understanding how the wider digital border infrastructure functions.
First introduced in 2011–2012, GCMS is a digital record-keeping system.Footnote 40 It is used by the two agencies already mentioned above, namely the IRCC and the CBSA, and also by Global Affairs Canada (the Canadian Foreign Ministry) and other ‘partner agencies’, a term that government documents use to broadly refer to other federal departments and agencies. Through GCMS, these agencies store, search, and retrieve visa applications, travellers’ entry and exit data, and prior eligibility and admissibility decisions (including the recorded grounds or reasons for such decisions). These agencies also use GCMS to ‘help screen and detect security issues, health issues, fraud, and program integrity risks to Canadians’.Footnote 41 Some of this screening is done through a query function, which enables agencies to search GCMS using data such as a legal name, aliases, date of birth, gender, country of birth, country of citizenship, passport details, and entry/exit information.Footnote 42
To facilitate risk screening and decision-making processes, GCMS interacts with a system called ‘Chinook’. IRCC officials use Chinook to evaluate hundreds of visa applications in ‘bulk’. This bulk sorting and review process is facilitated by Chinook’s ability to ‘flag’ GCMS records for a variety of reasons. A GCMS record might be flagged because a document is missing from someone’s recent visa application, indicating that their application should be rejected. A GCMS record may also be flagged if an application mentions an upcoming academic conference. This practice enables officers to evaluate all individuals who have applied for a visa to attend the same event together.Footnote 43 Chinook also flags GCMS records based on a variety of ‘risk’ factors identified through interoperable databases. For instance, a visa application’s supporting documents may be flagged as potentially fraudulent if they share features with documents previously identified as fraudulent.Footnote 44 Flagging enables IRCC officials to review and decide upon applications recorded within GCMS that are all identified as ‘alike’, enabling faster, more consistent decisions.
But GCMS is more than a database. It also provides data verification and tracking functions, which help IRCC target their immigration services across the country. For instance, GCMS enables state officials to verify biometric data at the border, as well as to track foreign nationals’ travel patterns and compare them to historic travel records.Footnote 45 It also tracks resettlement benefits newcomers receive and links to external databases, including university records, which federal agencies reference as they generate eligibility and admissibility decisions.Footnote 46 Through GCMS, federal agencies share information with each other within minutes, thus facilitating the achievement of ‘real-time’ digital governance.Footnote 47 This data also persists within GCMS, creating a seemingly perpetual memory.
The data stored and tracked within GCMS is shared internationally, including with the United States, facilitated by legal and technical mechanisms that enable interoperability. Pursuant to the US-Canada Beyond the Border declaration, travellers’ entry and exit data, as well as their biographic and biometric data, is automatically shared between the agencies of both states.Footnote 48 American and Canadian agencies are also able to query each others’ databases as they conduct risk assessments.Footnote 49
While GCMS already links to other digital tools (risk assessments, Canada’s no-fly list), it is presently undergoing a ‘digital platform modernisation’ process. Among other things, this process will update GCMS’ underlying hardware and software infrastructure to make GCMS compatible with new(er) machine learning functions. While publicly available materials on this initiative remain vague, Canadian authorities appear focused on enhancing GCMS’s automated decision-making, data sharing, and batch processing capabilities. Digital platform modernisation is being advanced by a range of federal agencies and consultants, including McKinsey, and anticipates a combination of tools and systems that will be developed in-house and procured externally.Footnote 50
3.3.2 Smail Khaniche and the Experience of Digital Border Infrastructure
The experience of Smail Khaniche illustrates how state agency becomes concentrated and dispersed at and through the digital border. In 2017, Mr Khaniche, an Algerian citizen, was granted a multiple-entry temporary resident visa so that he could regularly visit his brother in Canada. This visa was valid for eight years. Mr Khaniche’s job as a railway mechanic enabled him to pay for his trips to Canada. Though his visits meant being apart from his wife and two young children, they were important to stay connected with his sibling. Mr Khaniche travelled to Montreal on this visa for the first time in January 2018 without incident. In November 2018, Mr Khaniche made a second trip, arriving at Montreal’s international airport from Algiers for a one-month visit with his brother. Mr Khaniche had purchased his return plane ticket in cash shortly before this trip, the only way to finalise the transaction at the time.
Upon his arrival at the Montreal airport, Mr Khaniche was examined by a CBSA officer. The officer asked about his employment and financial status, as well as that of his brother. Mr Khaniche explained that his brother lives in Montreal but works remotely in Fort McMurray, a major oil sands operation in northern Alberta, Canada.Footnote 51 He stated that they were going sightseeing, but he was not sure where. The officer looked at his computer. Mr Khaniche offered to call his brother, who was waiting in the arrivals area, to verify his account. The officer refused, and instead demanded that Mr Khaniche hand over his cell phone and its password, the contents of which the officer then examined. The officer said, ‘You are doing pretty well in your country. Why did you really come to Canada?’Footnote 52 The officer then seized Mr Khaniche’s passport, airline ticket, and other identification. He ordered Mr Khaniche to be quiet and called the airport police, stating ‘Tonight you are going back to your country’.
At this point, the officer also pushed a voluntary departure form in front of Mr Khaniche, telling him that he must sign it.Footnote 53 Mr Khaniche refused. The officer yelled at Mr Khaniche, attempting to physically force his signature, but Mr Khaniche continued to protest. In the end, the officer signed the form alone; the space for Mr Khaniche’s signature remained blank. The officer, assisted by airport police, then escorted Mr Khaniche back to the aircraft that had transported him to Montreal, telling him to ‘get the hell back to your country’.Footnote 54 A combination of CBSA officers and police boarded Mr Khaniche onto the flight. His identification documents and cell phone were given to airline staff by another CBSA officer. Mr Khaniche arrived back in Algiers the following day.
After Mr Khaniche’s departure, the CBSA officer finalised his GCMS notes. These notes stated:
~ Subject is not a canadian citizen – subject is not a permanent resident of canada – subject is a citizen of algeria – subject requests admission for 999 days, has no hotel reservation, mentions that he is visiting his brother for 1 month who lives in alberta. mentions that he is going to stay in quebec for the month and that his brother is coming here for 1 week only. subject bought his ticket 2 days ago in cash, which is the equivalent of one month’s salary. subject mentions that he wants to visit the landscape and canada but does not know what he wants to see or what he is going to visit. subject mentions that he has a salary of $500/month but that he cannot prove the financial means to travel. subject does not have much connection with his country. subject’s luggage is not representative of the duration of the trip. – subject failed to satisfy the officer that he was a bona fide traveller and would leave at the end of the authorized period.Footnote 55
At the end, the officer wrote that Mr Khaniche failed his security screening assessment and was ‘allowed to leave’ Canada.Footnote 56
Upon returning to Algeria, Mr Khaniche hired a Canadian lawyer who successfully challenged the inadmissibility decision. But during the Court proceeding, government lawyers tried to use legal technicalities to render state agency invisible and thus unaccountable. Despite the very physical encounter between agents of the state and Mr Khaniche, government lawyers argued that there was actually no decision here to challenge. This argument relied upon a key administrative law technicality: namely, that courts will only review administrative action when a clearly identifiable decision exists.Footnote 57 According to government lawyers, ‘no admissibility was found and no removal order was made against [Mr Khaniche]’.Footnote 58 In other words, because the officer did not record Mr Khaniche’s status as ‘inadmissible to Canada’ within GCMS, and because the officer did not issue a removal order (a mechanism that would have provided Mr Khaniche with access to a statutory appeals process), the officer made no decision at all. This description suggests that Mr Khaniche removed himself, despite the altercation detailed in Mr Khaniche’s evidence.
The Federal Court rejected federal authorities’ attempt to make Mr Khaniche the agent of his own removal, concluding that ‘this was a disguised decision to deny entry’.Footnote 59 The Court’s ultimate concern was the wider impact of its decision. If it accepted the government’s argument that the officer’s actions (interviewing Mr Khaniche, evaluating his answers alongside GCMS data, concluding that Mr Khaniche should not be admitted to Canada, forcibly boarding him onto a return flight to Algeria) were not a ‘decision’ in administrative law terms, doing so would shield large spheres of administrative action from legal accountability mechanisms, such as statutory appeal processes and judicial review. The Court also noted that accepting the ‘non-decision’ argument might prompt other border officials to reach ‘disguised decisions’.Footnote 60 But the Court’s reasoning relies on the GCMS note that Mr Khaniche ‘failed to satisfy the officer’ that he was a genuine traveller as evidence that the officer reached a decision. It thus underappreciates the distributed agency present in this interaction and signalled by the notes themselves (which we will return to below).Footnote 61
The Court decision in this case is a rare legal victory for someone with a temporary resident visa, albeit a ‘victory’ that only partially addresses how Mr Khaniche may continue to be targeted as ‘risky’ in the future as a result of this encounter. The decision also simultaneously hides and reveals how dispersed state agency is within digital border processes. On the one hand, the Court identifies a single state agency (the federal Ministry of Public Safety and Emergency Preparedness, and the CBSA officer exercising delegated authority on behalf of the Minister) as legally responsible for the inadmissibility decision. The Court even orders this agency to provide Mr Khaniche with a unique legal remedy, discussed further below. However, the judgment also suggests how widely agency is distributed throughout bordering infrastructures, and how elusive accountability might be when vertically oriented mechanisms are used to achieve it. The next section examines these issues more closely.
3.3.3 A Critical Reassessment: Concentration, Diffusion, and Digital Border Infrastructure
If we critically reassess the above interaction, the simultaneous concentration and diffusion of state agency becomes clearer. Digital border infrastructure concentrates state power in three ways: it intensifies state power on Mr Khaniche at the border, it formats knowledge to shield the inadmissibility decision from external review, and it attaches a risk status to Mr Khaniche that persists into the future despite judicial intervention.
First, in the physical encounter described above, Mr Khaniche experiences a focused form of state agency. The digital border connects the actions of many agents exercising state power: from IRCC and CBSA officers to GCMS and its data sources to risk analyses and beyond. Through their coordinated acts, Mr Khaniche’s multiple-entry visa and his previously verified status as a trustworthy traveller become meaningless. State agency applies with full force to Mr Khaniche: his cell phone and identification documents are seized, officers attempt to coerce him into signing a voluntary departure form, multiple officers reboard him onto the plane, he is returned to Algeria. Throughout, Mr Khaniche is the subject of and subjected to algorithmic violence.Footnote 62 The digital border infrastructure (risk assessments, data sharing tools, and the representations of risk flagged in GCMS for the CBSA officer’s review) casts doubt on whether Mr Khaniche is a bona fide traveller. It also targets him as potentially risky and singles him out for scrutiny based on intersecting features such as his country of origin, his lack of access to electronic banking systems, his anomalous amount of luggage, his brother’s remote work arrangements, and the fact that his plane ticket was purchased last-minute using cash. The unfair treatment that follows is produced by this combination of officials, tools, risk indicators and state agencies that together deny Mr Khaniche’s admission to Canada and return him to Algiers.
This infrastructure intensifies state agency in a second way, by making (or attempting to make) the state’s inadmissibility decision legally unreviewable. Specific infrastructural elements format knowledge – the finding that Mr Khaniche is inadmissible to Canada – to shield it from external scrutiny. This formatting occurs through a combination of actions by the officer, the voluntary departure form, and the GCMS entry/exit record that would indicate Mr Khaniche chose to leave. When state agents characterise the conclusion that Mr Khaniche was inadmissible as a non-decision, and when they use the voluntary departure form to support this characterisation, they intensify state agency by making it unaccountable. In doing so, they try to mobilise legal technicalities to make this (non-)decision ‘final’ and shielded from judicial scrutiny.Footnote 63 Here, the Federal Court judge doubted the ‘non-decision’ categorisation because the border infrastructure itself expressed conflicting knowledge (the partially signed voluntary departure form, the GCMS note that the officer ‘was not satisfied’ of Mr Khaniche’s traveller status). In most other encounters, however, those affected would likely not seek judicial review and the non-decision illusion would persist within GCMS. Here, these tensions (the conclusion that Mr Khaniche is inadmissible, which is a non-decision) appear to be resolved through judicial oversight. Nonetheless, traces of the inadmissibility decision will persist throughout the digital border infrastructure, distinguishing this situation from previous forms of border decision-making.
Third, the encounter between Mr Khaniche and digital border infrastructure produces knowledge about Mr Khaniche (i.e., that he is a risky traveller), seeds that knowledge within the border infrastructure, and intensifies future scrutiny of Mr Khaniche and others like him. This encounter begins long before Mr Khaniche arrives in Montreal. He is likely flagged for attention through a prior risk analysis that evaluates his travel data to identify travellers with anomalous luggage, in-cash ticket purchases, and connections to high-risk countries. The physical interaction between Mr Khaniche and the officer, GCMS notes, and risk assessment tools (including the software used for analysing Passenger Name Record (‘PNR’) data to identify anomalous travellers pre-arrival) further co-produce Mr Khaniche as a ‘risky’ subject. Through interacting with Mr Khaniche, the officer came to doubt the purpose of Mr Khaniche’s trip.
This doubt was partly generated because of Mr Khaniche’s responses to questions about sightseeing plans and his brother’s work and living arrangements, but it may have been supported by data gleaned from Mr Khaniche’s phone.Footnote 64 Risk assessment(s) almost certainly reinforced this doubt. The officer’s notes suggest that he failed to hear many of Mr Khaniche’s answers to his questions, perhaps because he focused on data appearing on his computer screen.Footnote 65 The GCMS notes list factors common to border security risk assessments: suspicious luggage, a long stay (mis-recorded as ‘999 days’), a ticket purchased in cash.Footnote 66 If these notes were written on paper and filed in a cabinet, their impact might be limited. When recorded in GCMS as machine-readable text and stored in a database accessed by other agencies, however, they will influence Mr Khaniche’s future interactions with Canadian border authorities and those of partner states, such as the United States, especially as these states enhance the interoperability of their digital border infrastructures.Footnote 67
Mr Khaniche is also likely to be placed on a watchlist. This technical re-use of notes and other data ‘captured’ during the altercation at the Montreal airport thus further intensifies the force and focus of state power on Mr Khaniche and on other ‘risky’ travellers with similar attributes. Over time, this data may also refine and modify the software used by Canadian border agencies to prevent pre-approved visas from being granted to risky travellers like Mr Khaniche in the first place. State authority thus projects into the future in a way that is distinctive to this context and shaped by these border infrastructuring practices and arrangements.
Conventional legal tools fail to combat the seeding and future use of this knowledge and data. If digital border infrastructure components were less connected, a judge-ordered remedy might mitigate the impact of Mr Khaniche’s status as a risky traveller. For example, the Court might order that the inadmissibility decision was of no force and effect and would ‘stay’ (or freeze) any state actions put in motion by that decision, such as a removal order.Footnote 68 Similarly, the Court might order that details in Mr Khaniche’s GCMS record should be purged. If the chain of events put in motion by digital border infrastructure could easily be stopped, and problematic data easily deleted, these remedies would be useful. Yet, conventional remedies cannot prevent infrastructural elements from focusing authority into the future, allowing traces of riskiness to remain in circulation potentially indefinitely.Footnote 69 Perhaps recognising this dilemma, the Court issued a unique order in Mr Khaniche’s case. It required the Minister to delete the non-decision traces from GCMS notes (the text stating that Mr Khaniche failed his security screening and was ‘allowed’ to leave Canada).Footnote 70
This order opens new legal avenues for Mr Khaniche, enabling him to challenge the inadmissibility decision through a formal hearing.Footnote 71 But until Mr Khaniche uses that process (and there is no indication that he did so here), GCMS notes continue to identify Mr Khaniche as suspicious. This data remains accessible to future algorithmic and human reviewers. In addition, the Court order to delete notes does not bar a machine learning model from inferring risk as a result of the GCMS encounter data showing Mr Khaniche was returned to Algeria.Footnote 72 These infrastructural features bolster state power vis-à-vis Mr Khaniche because they deem him a risky traveller. This status is one that he cannot easily escape now or in the future, especially if it is relied upon by interoperable border governance systems elsewhere.
As these infrastructural arrangements concentrate state authority onto Mr Khaniche, they also disperse state agency so widely that identifying a responsible state authority for conventional legal accountability purposes becomes difficult. This spreading out amplifies pre-existing bureaucratic agency and accountability challenges. In earlier iterations of border infrastructures, multiple officers, files, and forms might impact a decision. In such cases, officials would dispute who was responsible for the outcome.Footnote 73 But today, the spreading out of agency occurs at a spatiotemporal scale that makes it nearly impossible to formally ‘name’ or pinpoint the many actors who materially contribute to a border decision using available legal tools. Administrative law and its vertical and temporally limited accountability pathways are thus rendered ineffective. The remainder of this section will detail how this phenomenon occurs.
Although state power clearly targets Mr Khaniche, as soon as we search for a responsible state agent we find agency and responsibility to be widely distributed. The actions of many entities materially contribute to the decisions that led to Mr Khaniche’s removal from Canada. If we extend the search beyond those actors who exercised formal delegated authority on behalf of ministers (a technical administrative law approach), it becomes evident that many actors who contribute to the digital border infrastructure are key agents in the events leading Mr Khaniche to be returned to Algeria. Rather than ask how boundaries between the state agency, individuals, and other entities like tech companies are redrawn or stabilised, which is one of the aims of an infra-legalities approach, here I trace relations to see how widely the boundaries of state agency are spread. This is an adjacent, but equally important, task for grasping the legal dimensions of a socio-legal-technical infrastructure.
The interaction that generated the inadmissibility decision appears to be largely between Mr Khaniche and the CBSA officer, but digital bordering infrastructures distribute key relations beyond the physical encounter at the border. Mr Khaniche’s multiple-entry visa, for instance, is issued by officials and assistive devices within IRCC, a department that is bureaucratically and legally separate from the CBSA. GCMS data is maintained and updated largely by officials from both IRCC and CBSA, but GCMS’s technical infrastructure is under the purview of the Treasury Board and shaped by the Privacy Commissioner and private developers. The digital tools that draw ‘lines of association’ between Mr Khaniche’s PNR data, travel patterns, and high-risk indicators share data between federal and provincial authorities and intelligence units elsewhere.Footnote 74 Mr Khaniche’s airline transmits personal information to border officials while he is in transit, which the CBSA and GCMS rely upon to assess Mr Khaniche’s admissibility.Footnote 75 Border officials may also use social media monitoring resources, especially after confiscating Mr Khaniche’s phone.Footnote 76 Many agencies together remove Mr Khaniche from Canada. Municipal police officers work alongside border officials to escort Mr Khaniche to the plane. Airline staff return his travel documents and phone. The decision-making tasks determining a single border crossing outcome are therefore horizontally dispersed across a web of officials, government agencies, airlines and other firms, databases, and algorithmic tools. Their effects ripple into the future through the traces of ‘riskiness’ left in GCMS and accessible to other agencies. For example, Mr Khaniche may be enrolled onto Canadian and international watchlists like the American no-fly list (if he is not already), which will impact his future ability to travel to Canada and to other states that have integrated this risk governance technology into their digital border infrastructures.Footnote 77
This diffuse state authority challenges the vertical legal accountability pathways common to administrative justice systems in Canada and elsewhere. Legal accountability fails to flow easily when authoritative decision-making agents are spread horizontally across different agencies and throughout a digital border infrastructure.Footnote 78 Even lawyers who are expert in immigration law may have trouble identifying and formally naming the correct federal ministry in a judicial review application, as did Mr Khaniche’s own legal team. Perhaps relying on conventions commonly used to challenge removal orders, his lawyers named the Minister responsible for the IRCC, whose agents issued Mr Khaniche’s multiple-entry visa. During the court proceedings, however, government lawyers argued that the Minister of Public Safety and Emergency Preparedness was the more appropriate party because the CBSA officer exercised delegated authority on behalf of this Minister when determining whether Mr Khaniche was admissible to Canada. The Court ultimately amended the hearing documents to name the second Minister so that its remedial order would legally bind the appropriate federal authority.Footnote 79 This change ensured a vertical flow of accountability channelled from the Court, through its order, to the Minister and presumably to the front-line CBSA officers who would implement the Court order.
Many responsible entities and data governance practices involved in this particular border encounter remain untouched by this vertical legal accountability model, however. If Mr Khaniche’s legal team tried to name every entity that materially contributed to the inadmissibility and removal decisions, or to expunge all data traces from which the potential ‘riskiness’ of Mr Khaniche might be inferred in the future, they would encounter real legal and practical challenges. A court would likely resist naming all of the responsible government agencies due to administrative law conventions. Judicially crafted legal principles might bar attempts to expand the network of named parties horizontally to include tech companies, airlines, and other entities that substantively contributed to Mr Khaniche’s removal.Footnote 80 Practically speaking, the opaque relations between the actors that co-create digital border infrastructures would impede lawyers from learning about other entities responsible for the inadmissibility decision. Canada, like the UK and EU, continues to build a global border governance infrastructure that is partially grounded in advanced risk analytics and data-driven insight sharing.Footnote 81 This infrastructure, including GCMS, relies upon exchanges between different data systems and data sharing between public and private actors.Footnote 82 When layered on top of a multi-agency bureaucracy like the one governing Canada’s borders, this infrastructure may exacerbate the fractured responsibility already characteristic of administrative governance.Footnote 83
Ultimately, conventional legal accountability enables the diffusion of agents described above to strengthen state agency because it places powerful border decision-making elements beyond the reach of accountability mechanisms central to administrative law. At present, only a small part of the expanding web of agents, relations, and data practices involved in digital border governance can easily be reviewed or subject to public law remedies: specifically, the actors who exercise statutorily delegated authority on behalf of a single ministry.Footnote 84 This prevents administrative law principles, such as ensuring a fair hearing or a substantively just outcome, from reaching pivotal components of digital border infrastructures and poses serious legal accountability challenges, as I discuss below.
3.4 Tackling Challenges for Legal Subjects and Legal Accountability
Building on the insights into how the digital border’s socio-legal-technical infrastructure operates, how should scholars and advocates respond? Specifically, how might we address the challenges that a network of agencies and actors with few responsible – or willingly responsible – principals raise for conventional accountability mechanisms? To return to the metaphor of the swarm nesting in the sovereign, what is a swarm without a ‘queen’? This concluding section proposes three main points of action. It urges scholars and advocates to take decision-making infrastructures seriously, to take algorithmic violence seriously, and to develop robust horizontal accountability mechanisms that better address both decision-making infrastructures and algorithmic violence.
First, decision-making infrastructures, including those governing borders, must be taken seriously. This proposal is deceptively simple. It challenges accountability strategies that centre on human action and those that focus on regulating a single node within a decision-making process. Human-centred and tech-centred approaches are standard in the growing literature on how to regulate algorithmic tools, but they underappreciate the interwoven nature of infrastructural relations. For example, a human-centred response to Mr Khaniche’s experience might propose that CBSA officers’ discretion ought to be better guided. It might also suggest that human officers ought to act as a constraining force ‘in’ or ‘on’ the loop of risk assessments and automated capabilities incorporated into GCMS.Footnote 85 A tech-centred approach, meanwhile, might identify GCMS as the key ‘actant’ in the network that forcibly removed Mr Khaniche from Canada and recommend a wholesale re-evaluation of GCMS’s functionality to regulate data flows between GCMS and other systems.Footnote 86 On their own, however, both approaches fail to address how humans, technologies, and agencies co-produce the conclusion that Mr Khaniche is inadmissible to Canada. The authority of state agencies, from the CBSA to the intelligence units constructing no-fly lists, and of other agencies whose actions support the outcome here (airlines, banks, tech firms) grounds the functionality of digital border infrastructure. Taking infrastructure seriously arguably requires us to analyse the proliferation of state agencies to ensure that state power remains governable particularly in dataist state initiatives.Footnote 87
Taking infrastructure seriously also enables us to use infrastructural knowledge to identify new sites where strategic action might impact outcomes and to trace how such action affects an infrastructural web.Footnote 88 This approach would require that we commit to long-term horizontal accountability strategies and to follow-up studies of their effects. Both features disrupt contemporary academic and regulatory conventions, which are governed by relatively tight three-to-four-year cycles based on the temporal constraints of external funding and election cycles. To address this challenge, we must creatively envision, scope, and support both research and regulation to achieve long-term goals despite the pressure for short-term deliverables. For instance, researchers and advocates alike might undertake projects to expansively trace infrastructural elements in a format that is easily accessible and intelligible. These initiatives could mirror ‘open government’ practices and could be regularly archived and updated. Government agencies relying upon digital infrastructures could revise their consultation practices to adopt ongoing, robust, agonistic processes that continue past the initial adoption of any given algorithmic tool.
Second, we must better recognise and respond to how data-driven infrastructures enable and shield algorithmic violence. While this term may lack a clear definition,Footnote 89 what is clear is that socio-legal-technical infrastructures tend to target the same marginalised people and communities for unfair, even violent, treatment.Footnote 90 For these reasons alone, we should heed Rocco Bellanova and others’ call to critically assess ‘how algorithmic systems feed (into) specific forms of violence, and how they justify violent actions or redefine which type of violence is considered legitimate’.Footnote 91 Doing so necessitates reflection on what counts as ‘violence’. The forced removal of Mr Khaniche fits, as does the officer’s attempts to coerce Mr Khaniche’s signature on the voluntary departure form. But Mr Khaniche also spent one month’s salary on his plane ticket believing his multiple-entry visa would ensure his admission, only to be denied entry to Canada. To what degree does algorithmic violence extend to family in Algeria, or his brother in Canada? How might the digital traces of ‘riskiness’ resulting from this border encounter affect his life or the movements of those ‘associated with’ him in the future?
Taking algorithmic violence seriously requires that we interrogate how infrastructural relations extend the state monopoly on violence to other actors, and with what effects. In addition to examining computational force (the force required to enable computation and the force leveraged by computation),Footnote 92 we must also investigate how infrastructural arrangements distribute opportunities for violence through computational and other digital border components. To what end are airlines, municipal police, financial institutions, and even voluntary departure forms sources of algorithmic violence? To what degree do (or ought) legal accountability mechanisms extend to these actors? Put differently, how does algorithmic violence become shielded from recognition and scrutiny when performed by an infrastructural network that seems infinitely expansive? How might we recognise and critically assess other components of a socio-legal-technical infrastructure, including the violence of infrastructure maintenance (i.e., biometric data collection, or forced access to Mr Khaniche’s phone).Footnote 93
Finally, to take both infrastructure and algorithmic violence seriously, we must develop accountability strategies that are as horizontal and networked as the phenomena they seek to regulate. Public administration and digital governance scholars highlight the urgency of this task.Footnote 94 It is arguably a necessary companion to infra-legalities research, as infra-legal studies suggest that only a more robust set of tools will effectively govern the web of actors that constitutes digital border infrastructures. Specifically, socio-technical solutions that ensure responsibility flows across infrastructural nodes must be paired with legal mechanisms. Administrative law updates, for instance, might allow multiple actors to be named in legal proceedings, expand available public law remedies, or even impose joint and several responsibility for implementing court orders. We might also imagine how legal standards themselves might be encoded into the networks and nodes described above to alter their affordances.
Horizontal accountability theory is one place to start. As a theoretical approach, horizontal or shared accountability has two main benefits. First, it encompasses the diverse actors (front-line officials, managers, ‘non-human’ software, databases, and state agencies) that co-create outcomes.Footnote 95 Through its attention to these actors, this theory may suggest how responsibility can be shared between many agencies of the state. Second, as a theory of public administration, horizontal accountability generates potential regulatory responses that extend beyond law and policy amendments or technological tweaks. For example, it may reveal other responsibility pathways and accountability mechanisms that would be highly effective components of a larger toolkit to govern state agency and agencies.Footnote 96
Open access