The politics of (non-)knowledge at Europe's borders: Errors, fakes, and subjectivity

Abstract From statistical calculations to psychological knowledge, from profiling to scenario planning, and from biometric data to predictive algorithms, International Relations scholars have shed light on the multiple forms of knowledge deployed in the governing of populations and their political effects. Recent scholarship in critical border and security studies has drawn attention to ‘the other side of knowledge’ and has developed a vibrant conversation with the emergent interdisciplinary field of ignorance studies. This article proposes to advance these conversations on governing through non-knowledge by nuancing the analysis of power/(non-)knowledge/subjectivity relations. Firstly, we expand the analysis of non-knowledge by attending to the problematisation of errors and fakes in controversies at Europe's borders. Errors have emerged in relation to border actors’ practices and technologies, while migrant practices, documentation, and narratives are deemed to be potentially ‘fake’, ‘fraudulent’, or ‘false’. Secondly, we explore how different subjectivities are produced through regimes of error/truth and fake/authenticity. We argue that there are important epistemic differences between ‘fake’ and ‘error’, that they are entangled with different techniques of power and produce highly differentiated subjectivities. Finally, we attend to how these subjectivities are enacted within racialised hierarchies and ask whether non-knowledge can be mobilised to challenge these hierarchies.


Introduction
From statistical calculations to psychological knowledge, from profiling to scenario planning, and from biometric data to predictive algorithms, International Relations (IR) scholars have shed light on the multiple forms of knowledge deployed in the governing of populations and their political effects. More recently, 'the other side of knowledge' and its implications for governance have received increased attention across the social sciences. In IR, research on (in)security has unpacked the heterogenous modes of governing unknowns and 'taming uncertainty', whether through risk, resilience, pre-emption or imagination, particularly in the governing of terrorism. 1 Non-knowledge becomes an imperative for the production of more knowledge, even as these 3 Stephan Scheel  Rather than starting with typologies of ignorance or non-knowledge, we propose to start with the practices that problematise different forms of non-knowledge for the purposes of governing. 10 This allows us to attend to the multiplicity of non-knowledge and understand how boundaries are drawn between what counts as knowledge and non-knowledge.
Secondly, we explore the relations between errors, fakes, and power through Foucault's positive model of power as production, making, fabrication, and enactment. Following Foucault, this would entail 'pass[ing] from a technology of power that drives out, excludes, banishes, marginalizes and represses, to a fundamentally positive power that fashions, observes, knows, and multiplies itself on the basis of its own effects'. 11 By analysing problematisations of non-knowledge in relation to power, we highlight the differential subjectivities that regimes of knowledge/non-knowledge/ power produce. 12 Thirdly, by analysing hierarchies of subjectivity, we also aim to contribute to literatures that have underscored the analytical importance of gendered and racial not-knowing. 13 Colonial governmentality does not only produce ignorance and zones of secrecy, but particular imbrications of non-knowledge and (distorted) knowledge. We show how regimes of error/accuracy and fake/authenticity generate racialised hierarchies of subjectivity and ask how non-knowledge might be mobilised to challenge these hierarchies.
Methodologically, we attend to how errors and fakes as modes of non-knowledge are differentially produced and generate distinct subjectivities. In order to connect errors and fakes with power relations and subjectivities, we gather different problematisations of errors and fakes in what can be called, following Celia Lury, a 'compositional methodology'. Lury draws on the meaning of composition as a creative way of 'putting a problem together' that allows us to describe 'the dynamic and methodologically constitutive dimensions of the individuation of a problematic' that is 'always both a composite and compositional' in the sense that a problematic is formed and transformed. 14 Composition also draws attention to the heterogeneity of elements and temporalities that are brought together to constitute a problem space. 15 We form such a methodological problem space of governing borders through non-knowledge by gathering dispersed problematisations of errors and fakes. We bring together problematisations of errors in the practices of European Union (EU) agencies: EASO, Frontex, and eu-LISA. We highlight a complaint brought by the NGO Advocates Abroad against EASO practices in Greece, which problematises the relation between power and non-knowledge. The analysis of error in EASO documents and guidelines transformed the problem space of error by revealing that errors and fakes, making mistakes and deceiving are entwined. Although this article was initially envisaged as an analysis of error in border governmentality, this entwined problematisation meant that we needed to hold errors and fakes together. We subsequently traced the problematisation of fakesin their entwinement with errorthrough a legal case highlighted by EASO in their 2020 annual asylum report. As problematisations of errors and fakes proliferate across institutions, they do not amount to a stable object, but require the methodological flexibility of composition.
We also transform the problem space of border governmentality through the methodological device of juxtaposition. While Lury does not specifically discuss juxtaposition as compositional methodology, juxtaposition is also one of the techniques of collage. 16 Rather than comparison, it is the juxtaposition of these forms of non-knowledge that generates new meanings and allows us to render differential power/knowledge/non-knowledge regimes intelligible. We argue that error is produced in experts' practices as well as in evaluations of technological developments, such as the extension of biometrics across the various EU databases. Fakes, however, are reserved to representations of migrant practices, documents, data, and narratives. Juxtaposing errors and fakes helps shed light on how different power relations and racialised subjectivities are enacted at European borders.
To develop our argument, we start with a discussion of recent scholarship on ignorance and non-knowledge in IR, focusing more specifically on borders and migration. Secondly, we turn to the production of error as both human and non-human in the practices of evaluating asylum claims and identifying asylum seekers. Thirdly, we trace the problematisation of fakes as distinct from errors through a legal case that asks to clarify the conditions under which the refugee status or subsidiary protection can be revoked. Finally, we show how the juxtaposition of errors and fakes can help highlight different forms of power and subjectivity at Europe's borders.
Governing borders through non-knowledge One of the notions most famously associated with Michel Foucault's work has been that of power/knowledge. Dispositifs, apparatuses, or assemblages that entwine power and knowledge have been at the heart of critical engagements with social practices and particularly the 'conduct of conduct'. To make people and things knowable is to make them governable and actionable. IR scholars have studied the heterogenous production of knowledge, proliferating numbers of experts, techniques, and devices that render borders and migration governable. 17 As Foucault notes in the interviews collected under the title Power/Knowledge, '[t]he exercise of power perpetually creates knowledge and, conversely, knowledge constantly induces effects of power.' 18 Work on knowledge societies and expertise in global governance has reinforced attention to the practices of knowledge production, contestation and circulation. 19 16 For Anna Leander, collage as a methodology for the analysis of commercial security is similar to composition: 'Collaging is a way of accepting the messy, sprawling and evolving aspects of composing. '  Recent interdisciplinary work in ignorance studies has supplemented attention to knowledge production with the production of ignorance or non-knowledge. Ignorance is neither a 'lack' of knowledge or a cognitive state, but a social construction. 20 Agnotology was initially coined in science and technology studies (STS) to render the cultural production of ignorance and doubt in techno-scientific controversies. 21 Indeed, ignorance was particularly revealing when strategically deployed by actors who aimed to destabilise the scientific consensus, whether concerning health and tobacco, or climate change. Ignorance studies have expanded the study of how ignorance is generated, how it circulates and how it underpins power relations to a range of social fields and political issues. McGoey has coined the terminology of 'strategic ignorance' to encompass 'any actions which mobilize, manufacture or exploit unknowns in a wider environment to avoid liability for earlier actions'. 22 In a call for an epistemology of ignorance beyond science and technology, Shannon Sullivan has argued that ignorance needs to be understood 'as an active production of particular kinds of knowledges for various social and political purposes'. 23 These approaches that supplement power/knowledge with (strategic) ignorance have also started to make inroads in IR. While scholarship in IR and particularly in critical security studies has offered sophisticated analyses of 'governing unknowns', until recently there has been less engagement with how non-knowledge is produced for the purposes of governance. 24 As William Walters has remarked, even secrecy has been a rather marginal object for security studies rather than a key analyser of power relations in liberal democracies. 25 However, interdisciplinary research on borders and migration has been one of the most vibrant fields of engagement with ignorance studies in IR.
Making migrants knowable, assigning them to categories of risk so that some can circulate, while others are stopped, contained, or confined has been a central concern of literatures addressing border governmentality. Through entanglements of sovereign, disciplinary, and biopolitical power mechanisms, migrants are rendered knowable and legible so that borders can be made governable. Many of the recent developments around creating new border and migration databases in the EU and connecting biometric and other data have been driven by the narrative of insufficient knowledge. If only border professionals knew more and more quickly, then they could govern better. Matthias Leese has located a shift in the state practices of making populations knowable from identity production to identity management through biometric databases and algorithmic analysis. 26 Other critical scholars have highlighted the failure of this quest for knowledge, as it encounters material complexities and migrant agency. 27 Here, ignorance emerges from the encounters between different actors, their practices and knowledges.  Thus, it is not surprising that scholarship in border and migration studies has started to engage with the production of ignorance as a key site of theoretical and political intervention. Critical scholars have offered nuanced analyses of how ignorance is produced and the effects it has upon migrants. For instance, Enrica Rigo explains that What these figures [border crossings] never say, and what the statistics never ask, is how many times the same person crosses the same border and for how long she gets stuck in the border machine during her journeys. In other words, the statistics do not account for migrants' subjective experiences of borders. 28 Statistics and the politics of numbers do not just produce knowledge, but they also produce ignorance, which can have violent and even lethal effects on migrants' lives. Here, the mechanism of producing ignorance is a privative one, leading to a lack, an erasure or otherwise silencing of situated knowledges. Migrants' experiences are effaced, denied or suppressed from the circulation of statistics. Missing numbers of people on the move who died from the COVID-19 virus led scholars to conclude that 'invisibility may mean death' by encouraging exploitation on the temporary job markets or fuelling xenophobic and racist reactions. 29 Lisa Marie Borrelli has aptly rendered this privative and exclusionary effect of ignorance, as she argues that 'ignorance legitimises structural violence as it limits the "other's" agency, for example by consciously creating knowledge gaps and not acknowledging the other's aspirations and thus, voice.' 30 In her nuanced study of refugee governance in Lebanon, Stel has highlighted another form of ignorance through the production of institutional ambiguity that both maximises discretionary power and obfuscates accountability among the multiplicity of actors involved in governance. 31 Focusing on Lebanon's Palestinian refugees who live in unofficial camps, she observes an institutional ambiguity that produces deliberate disinformation and stalling tactics, and reproduces highly asymmetric power relations. 32 In Greece, Katerina Rozakou has highlighted the practices of nonrecording of refugees 'not as an indication of the state's inability to impose order, but as an essential element of statecraft'. 33 Ignorance or ambiguity are intentional and deliberate, they appear as strategies deployed and wielded by particular actors who remove knowledge, or render ineffective the knowledge that is considered suspicious, within the public space. As Martina Tazzioli has recently shown, the interplay between knowledge, non-knowledge, visibility, and invisibility is key to bordering practices more broadly. 34 The literature on border and migration studies has shown the need to attend to different forms of ignorance or non-knowledge: uncertainty, ambiguity, confusion, or opacity. The relations between ignorance and non-knowledge and their meaning vary across authors, partly because the study of ignorance ranges across wide interdisciplinary terrains. For some authors, ignorance and non-knowledge are used interchangeably, with ignorance often being the overarching term. As McGoey has formulated it, non-knowledge needs to be understood 'not as a precursor or an 28 Enrica Rigo, 'Migration, knowledge production and the humanitarian agenda in times of crisis', Journal of Modern Italian Studies, 23:4 (2018), pp. 507-21 (p. 509). 29 Stefania Milan, Annalisa Pelizza, and Yoren Lausberg (3 February, 2020), available at: {https://www.opendemocracy.net/ en/can-europe-make-it/making-migrants-visible-covid-19-counting-dilemma/}. impediment to more knowledge, but as a productive force in itself, as the twin and not the opposite of knowledge'. 35 Jutta Bakonyi, drawing on Michael Taussig, suggests that it is impossible to differentiate non-knowledge from strategic ignorance. 36 For others, non-knowledge is preferable to ignorance, either to avoid the pejorative connotations of ignorance, as Ulrich Beck and Peter Wehling propose, or to attend to the different meanings of the two concepts. 37 As ignorance studies have also drawn attention to other forms of not-knowing such as doubt, uncertainty, ambiguity, secrecy, or ambivalence, these cannot be understood in terms of what is not known.
In this article, we use non-knowledge as a heuristic device to attend to the multiplicity and heterogeneity of 'the other side of knowledge'. For instance, Claudia Aradau has proposed to analyse controversies over different forms of non-knowledge such as uncertainty, conjecture, and secrecy. 38 Stephan Scheel and Funda Ustek-Spilda have also highlighted these differences, pointing out that ignorance is a particular type of non-knowledge that 'involves the obfuscation or suppression of otherwise available knowledge'. 39 More importantly, they move from taxonomies of ignorance/non-knowledge to an analysis of practices through which ignorance is produced. In analysing a migration visualisation tool by the International Organisation for Migration, they list four mechanisms for producing unknowns: omitting, compressing, deflecting, and sanitising. 40 In so doing, they extend the analysis of ignorance production to encompass more varied mechanisms beyond repression, erasure, silencing or invisibilisation. Building on this work, we attend to how errors and fakes as modes of non-knowledge are differentially produced and produce distinct subjectivities.
In the formulation of agnotology, Robert Proctor had already enjoined scholars to 'explore how ignorance is produced or maintained in diverse settings, through mechanisms such as deliberate or inadvertent neglect, secrecy and suppression, document destruction, unquestioned tradition, and myriad forms of inherent (or avoidable) culturopolitical selectivity'. 41 The mechanisms that he offers by way of illustration are all focused on a negative model of power that works through exclusion, secrecy, invisibilisation, erasure, and destruction. Therefore, ignorance is often analysed as productive but leading to privative effects or what Foucault has called 'an agency for solely negative effectsexclusion, repression, interdiction'. 42 Feminist and postcolonial scholarship on ignorance has called for more attention to power relations and subjectivity. As Shannon Sullivan and Nancy Tuana have enjoined researchers, 'tracing what is not known and the politics of such ignorance should be a key element of epistemological and social and political analyses, for it has the potential to reveal the role of power in the construction of what is known and provide a lens for the political values at work in our knowledge practices.' 43 Scholars have also analysed how ignorance is deployed by powerful actors and how it entrenches domination. 44 Ignorance as an active practice blurs categories of ignorant and knowledgeable subjects, rendering some credible and others not.  38 Aradau, 'Assembling (non) knowledge'. 39 Scheel and Ustek-Spilda, 'The politics of expertise and ignorance in the field of migration management', p. 6. 40 Ibid. In this article, we want to ask what other forms of migrant subjectivity come to be produced through different problematisations of non-knowledge. Does ambiguity produce the same subjectivity as uncertainty, doubt, or secrecy? Or, in our case, do errors produce the same subjectivity as fakes? By juxtaposing errors and fakes and analysing their composition, we can unpack how different subjectivities are relationally produced, while also being disjoined, and how they are co-constituted with different forms of knowledge and power. In the next two sections, we trace how making errors and detecting fakes have emerged in the practices of border governmentality.

Making errors
Errors have received scant analytical attention in studies of ignorance or non-knowledge. Like ignorance, errors are supposed to be reduced if not effaced in the search for accurate or even true knowledge. At times error and ignorance are used interchangeably. 45 Charles Mills's coinage of white ignorance comprises 'both false belief and absence of true belief', both error and ignorance. 46 At other times, error and ignorance are differentiated, with 'error' a subcategory of ignorance understood as 'distortion'. Here, we consider 'error' as another form of non-knowledge alongside and often entangled with ignorance. Error has come to increasingly feature in the guidelines, policies, and justifications of EU agencies governing borders and migration.
In 2017, the European Asylum Support Office (EASO) introduced a Quality Assurance Tool to provide Member States with a 'common framework for internal quality assessment and assurance'. As part of the tool, EASO developed an 'Assessment form' to evaluate the quality of the personal interviews that national asylum authoritiesor EASO professionals as in the case of Greececonduct with asylum seekers. 47 The form asks asylum professionals to assess a decision based on a scale that ranges from 'correct' to 'significant error'. 48 The descriptive column is replete with the language of valid knowledge: correct; correctly; objectively; accurate; accuracy; facts. Yet, the result of this assessment is not the distinction between true and false, correct and incorrect, accurate and inaccurate knowledge, but a modulation of error. Error is not eliminated, but it becomes integral to how interviews are assessed. From high quality to low quality, each assessment has a percentage of error. EASO then develops a hierarchy of errors, where a 'high-quality' interview is deemed to have under 20 per cent minor errors and no significant ones, while a 'low-quality' interview has one or more significant errors. 49 Shortly after EASO introduced this assessment of error in its Quality Assurance Tool, in April 2018, the NGO Advocates Abroad (AA) submitted a complaint to the European Ombudsman against EASO. 50 In the letter to the Ombudsperson, posted on the NGO's Facebook page but not made public by the office of the Ombudsperson as part of the case record, the Executive Director Ariel Ricker focuses on what she calls 'the most egregious violations of EASO and national office policy'. 51 Based on the analysis of interview transcripts, experience accompanying asylum seekers to interviews, and their complaints post interviews, AA denounce the regime of 'impunity' that allows both 'abusive and poorly prepared EASO interviewers and interpreters' as 45 Proctor and Schiebinger, Agnotology. well as 'the overall inconsistent EASO system'. They argue that EASO falls short of the very standards that the agency promotes through internal training modules and publicly available documents. AA also submitted four letters outlining the abusive behaviour by an EASO caseworker and interpreter during an interview with an asylum seeker, Mr X. 52 Mr X was a young man who had been subjected to a two-day interview and faced with highly inappropriate questions concerning his sexuality. Ariel Ricker pointed out the hefty number of complaints that were raised during EASO's two years of operations in Greece, so that AA could bring over one hundred cases in the complaint to the Ombudsperson. 53 While the AA complaint letter focuses on how lack of knowledge correlated with abuse, the Ombudsperson and EASO shift the language to that of error in their subsequent correspondence. As noted in the decision, the initial complaint by AA describes how the caseworker 'conducted the interview in a hostile and adversarial manner, appeared confused about the facts and ignorant on issues related to sexual orientation and gender identity, and at times impatient, aggressive and insensitive with [Mr X]'. 54 EASO acknowledged its lack of initial response and the misconduct of the case officer, diagnosing it as 'a severe error of judgment when dealing with [that] case'. 55 If the AA letter highlights ignorance and confusion, EASO reproblematises it as error. The Ombudsperson Emily O'Reilly starts correspondence with EASO to both clarify the case and find out what remedy might be available in the case of 'serious errors'. Given that the asylum seeker Mr X had been deported and was by the time of the decision missing (and deemed dead), the Ombudsperson endorses EASO's view that 'it has no mandate to seek or facilitate the return of asylum seekers in such a situation.' 56 Abuse and 'poorly prepared interviews' become rearticulated as 'errors', be those serious or significant.
Errors are a productive form of non-knowledge in that they reposition EASO practices and their effects on asylum seekers. As O'Reilly formulated it in her decision, 'EASO should see to it that its work is of the highest possible quality, that mistakes its experts make are identified, and that, to the extent possible, corrective action is taken.' 57 If a serious error is identified, the Ombudsperson asks that EASO do 'its utmost to remedy it'. EASO would also need to inform the national authorities about these errors. Although the Ombudsperson finds a case of maladministration, the corrective actions taken remain thin: proposal for a mechanism of complaint, informing Greek national authorities if errors are discovered and supervising the work of  55 Ibid. 56 Ibid. 57 Ibid., p. 3.
interpreters. EASO responds with a list of modifications to their practices aimed at catching 'errors'. While the training of caseworkers, quality feedback assessment and liaising with Greek authorities are marshalled as evidence for an error-vigilant organisation, nothing is done about Mr X.
Rather than eliminated or eradicated, errors have become integral to EASO's practices of governing asylum. They need to be continually minimised, checked, and filled in assessment forms. Mr X becomes, however, an unknown in this reproblematisation. We don't know what happened to Mr X, except that he is alleged dead. How did Mr X die, how was his deportation implicated in his presumed death, and what kinds of responsibilities are entangled in his death? Nobody knows anything more about Mr X. In her decision, O'Reilly also mentions her concerns that 'the errors identified by EASO may have contributed to the fact that Mr X was deported' and 'that EASO sent its (delayed) reply to the complainant and the Ombudsman the very day Mr X was deported.' Yet, despite a brief acknowledgement that Mr X is 'feared dead', the Ombudsperson concludes that 'While it is difficult to see what could be done at this stage to correct the mistakes that were made in his case, it is of the greatest importance to determine what lessons might be learned for the future.' 58 Only suggestions for the future are proposed with no inquiry into the effect of maladministration and culpability for the potential death of an asylum seeker.
As we have seen, errors are entangled with power relations, subjectivity and responsibility. Making errors enacts a vigilant (and knowledgeable) subject who can minimise, if not correct, their own errors. Making errors is indicative of a 'corrective ethos', 59 where correction becomes a continual endeavour. As John Roberts summarises in his inquiry into error, errors refer to 'missteps, omissions, oversights and mistakes involved in the execution of a particular activity'. 60 Making and finding errors does not undermine the credibility of the subject. Rather, it is exactly a subject who is trained to prevent and react to errors by reducing, neutralising, or even eradicating them. The diagnosis of error by a vigilant subject is thus opposed to systemic ignorance, which transforms relations into abusive ones. EASO, for instance, maintains that the complaint reveals only singular errors rather than a 'systemic problem'. As they are not systemic, errors cannot be associated with abuse as a relation of domination or violence. While we can see the AA letter as reminiscent of Mills's analysis of 'white ignorance' as a systemic effect of not-knowing causally linked to racial privilege in oppressive systems, 61 the EASO correspondence rejects this diagnosis of ignorance underpinning abuse. Rather, errors are envisaged as a complement to biopolitical techniques of improvement.
The EASO documents and the complaint before the Ombudsperson are not the only problematisations of error in relation to European borders. Errors have also become integral to border technologies that identify asylum seekers. We can say that human errors are supplemented or entwined with non-human errors. As asylum seekers are screened and their fingerprints recorded in the European Dactyloscopy Database (Eurodac), error surfaces in many of the reports on the use of fingerprints. Yet, the diagnosis of error does not challenge the use of these technologies, nor does it lead to more credibility for asylum seekers contesting this use of data. The European Union Agency for Fundamental Rights (FRA) has pointed out the multiple errors produced by biometric technologies, from fingerprints to facial recognition. 62 As the FRA explains in one of their reports, 58 Ibid., § 13. 59 We are grateful to Debbie Lisle for this formulation.  In 2016, the Eurodac central system rejected some 54,300 fingerprint datasets due to insufficient quality. This may happen due to the low quality of the fingerprint image or because of a sequence check error, meaning that the fingers were recorded in the wrong order. Conditions to acquire biometrics, such as the physical environment (for example, temperature and humidity) and the characteristics of the person concernedin particular, ageaffect the quality of fingerprints and may possibly lead to a mismatch. 63 In relation to facial recognition, the FRA notes that errors affect more specifically 'women and people of colour, producing biased results, which can ultimately result in discrimination'. 64 They also identify different types of a complement errors in the case of the facial recognition: 'Yet there is a complex relationship between false positives (i.e. stopping innocent people) and false negatives (i.e. not being able to find the person of interest).' 65 In a more recent report on Artificial Intelligence and data quality, they point out that data quality can be affected by 'errors of representation' and 'measurement errors'. 66 These reports show that the deployment of biometric technologies is co-constituted by errors. Once these errors are deemed acceptable, they become recorded as errors, while biometric technologies continue to be expanded. However, in a recent letter addressed to the Members of the European Parliament, thirty NGOs have contested the proposed expansion of Eurodac and highlighted the fundamental rights implications of adding facial recognition, recording children's fingerprints and including more categories of persons in the database. The letter also points out that 'The ability of this technology [of facial recognition] to increase the accuracy of matches, and thus the efficiency of the system, has not been proven and is therefore not a sufficient justification for such serious interference with people's fundamental rights.' 67 The letter questions the error/accuracy doublet of technical and expert knowledge.
These different problematisations show that errors emerge as both human and non-human. Human and non-human errors can be controlled or continually corrected, even if not eliminated or eradicated, by a vigilant and expert subject. Errors produce not only the subjectivity of the vigilant and credible expert, but also relations between these experts, technologies and asylum seekers. Asylum seekers are not included in the assessment or evaluation of errors. Moreover, Mr X's application for international protection was rejected, largely because his testimony was not believed. His narrative and documentation were deemed to be the opposite of truth; yet they were not errors. If caseworkers and border authorities can make mistakes and correct their own errors, this is not the case of asylum seekers, whose narratives, documents, and data are continually under suspicion of being fake. In the next section, we show how a different form of nonknowledge problematises false, fraudulent and fake documents, information, and stories.

Detecting fakes
In their 2020 Report on Asylum, EASO reviewed key jurisprudence by the European Court of Justice (CJEU) concerning the asylum legislation and procedures in the EU. The revocation of 63 The literature on biometrics has also drawn attention to error in biometric technology and has particularly highlighted that errors are 'intrinsic to biometric recognition'. See Shoshana Amielle Magnet, When Biometrics Fail: Gender, Race, and the Technology of Identity international and subsidiary protection for asylum seekers shows how a different form of nonknowledge is produced in relation to the subjectivity of asylum seekers and refugees. One of the cases in the report, Mohammed Bilali v. the Federal Office for Immigration and Asylum in Austria, was focused on the revocation of subsidiary protection that the Austrian state had granted Mr Bilali. The EU's 2011 Qualification Directive specifies not only the conditions under which asylum seekers are granted the refugee status or subsidiary protection, but also when these statuses can be revoked. 68 The attribution of danger or the commission of a 'serious crime' would not be surprising to IR scholars studying bordering and securitisation practices. 69 But particularly relevant for our analysis of power/non-knowledge is the revocation of protection due to 'his or her [an asylum seeker's] misrepresentation or omission of facts, including the use of false documents, [which] was decisive for the granting of refugee status'. 70 Here, misrepresentation, omission, and false documents pertain to a different category of nonknowledge, that of fakes and deception, and practices of faking, counterfeiting, and lying. The case of Mr Bilali problematises the distinction between two forms of non-knowledge: errors and fakes. The case did not implicate Mr Bilali as providing fake, fraudulent or otherwise deceptive information, but discussed status revocation given a change of 'the state of knowledge of the authority'. 71 The Austrian Court sought to clarify the conditions of revocation of subsidiary protection by asking whether 'a simple error by the authorities' would be considered within the scope of 'false pretences' and false claims on the part of the asylum seeker. 72 Is error a subcategory of fakes, asked the Austrian court that the CJEU clarify, so that revocation of status could be extended to this case? At first sight, errors and fakes might not appear that different. Yet, they problematise the relation between knowledge and non-knowledge differently, as we will see in the CJEU decision.
Mohammed Bilali, a stateless person, had his application for asylum refused by the Austrian Federal Office for Immigration and Asylum, as his identity could not be established, while the Federal Office held that 'he was probably an Algerian national'. 73 On the assumption that Mr Bilali was an Algerian national, the Federal Office granted him subsidiary protection 'on account of the high level of unemployment, lack of infrastructure and continuing instability in Algeria'. 74 Subsequently, this status was revoked on the basis that the Federal Asylum Office deemed Mr Bilali to be a Moroccan national, and subsequently 'adopted a return decision indicating Morocco as his country of destination'. 75 As further information came to light, the Federal Office decided this time that Mr Bilali was a national of both Morocco and Mauritania. According to the referring court, 'there is nothing to indicate that the delay in gathering information was attributable to Mr Bilali.' 76 What was at stake here was whether the revocation of status could be applied 'notwithstanding the fact that the person concerned cannot be accused of having misled the Member State on that  71 Ibid., § 30. 72 Ibid., § 27. 73 Ibid., § 18. 74 Ibid. 75 Ibid., § 21. 76 Ibid., § 26. occasion'. 77 Different practices of taming and producing non-knowledge are problematised in this case: on the one hand, how to decide on Mr Bilali's identity; on the other, whether Mr Bilali used fake information or deception. Indeed, Bilali's identity had been reclassified multiple times. In several rounds of interviews between October 2009 and September 2012, Bilali consistently reaffirmed his biographical account. 78 He was originally from Tindouf, a refugee camp in the Western Sahara that was located in Algeria but administered by the Frente Polisario (Polisario Front), a national liberation movement seeking to end Moroccan occupation of the Western Sahara. His father was a founding member of the Polisario and his mother was originally from Morocco, a claim that the Federal Office relied on in determining his citizenship. As he was born and grew up in Tindouf, like all other refugees there, he was not considered an Algerian citizen. In 2004, he escaped to Mauritania. Despite claiming to be stateless, the Federal Asylum Office assumed that he was probably Algerian and denied his asylum application on the basis of country information on Algeria provided by the American and German Foreign Ministries. His account did not indicate that he would have faced any state-sponsored persecutions in Algeria, and a distant possibility of persecution was not sufficient.
However, this decision was shortly reversed by the Austrian Asylum Court, highlighting institutional tensions between the Asylum Court and the Federal Office. The Court pointed out that the Federal Asylum Office did not conduct any research on the country of origin of the claimant and based its assessment on mere assumptions that he was from Algeria. 79 Moreover, the Office did not interrogate the claimant further, nor did it conduct any further investigations into his identity or origin. Finally, it did not conduct any research on Tindouf or Morocco and the Western Sahara. The Court also explicitly called upon the Federal Asylum Office to increase its standards and check statements given by asylum applicants, in a move to reduce ignorance and replace it by accurate knowledge. Following new rounds of interviews and considering country information on Algeria, but not Bilali's own narrative, his application was subsequently denied in October 2012. The Federal Asylum Office argued that the claimant's testimony was not credible, not consistent, not plausibly comprehensive or without contradictions, and not probable, particularly regarding the situation in his presumed country of origin and reasons of possible persecution. 80 A series of knowledge/non-knowledge relations are generated here: credibility/incredibility, consistency/inconsistency; plausibility/implausibility; probability/improbability. These attributions of producing non-knowledge to Mr Bilali are underpinned by a deeper problematisation of the relation fake/authenticity.
Although the Austrian Asylum Office made an error, the CJEU concludes that 'the fact that the error made by the host Member State when granting that status is not attributable to the person concerned cannot alter the finding that that person was never in fact a person eligible for subsidiary protection.' 81 The question of what caused this error becomes moot. As we have seen in the previous section, state authorities and experts are deemed to be in error, to be in possession of incorrect information. There is no requirement to shed light on why Mr Bilali's claim of not being an Algerian national had been dismissed by the Austrian authorities, or why he was disbelieved, or his testimony deemed to be incredible, implausible, improbable, inconsistent, partial, and contradictory. Even though Mr Bilali has not deceived the authorities, his claim is ignored in favour of the supposedly correct information that the Austrian authorities assign to him as a Moroccan national. As with Mr X, Mr Bilali is also excluded from the problematisation 77 Ibid., § 31. of error. It is the authorities' self-assessment of error that leads to the revocation of his status, even when fake information could not be proven.
Yet, the suspicion of faking or deceiving does not disappear. The distinction between errors and fakes, which we have started to outline in the previous sections, is enabled by the production of subjectivity. Errors and their correction remain the prerogative of authorities, even when these have previously failed in their responsibilities. Finding an error entails a 'change in the competent authorities' knowledge'. 82 On the one hand, state authorities appear as subjects lacking accurate knowledge, but able to correct this lack even when they are shown to have relied on vague assumptions and conjectures about Mr Bilali's identity. On the other, an asylum seeker is always assumed to present fake information, to mislead the authorities by omitting or misrepresenting facts.
As human and non-human errors are increasingly problematised in border and asylum practices, the 'fake' has been extended from asylum seekers' narratives to counterfeit or fraudulent devices and documents they mobilise to evidence claims to international protection. Passports, photos, and even biometric data are under suspicion of being 'fake'. In 2019, the annual conference of the European Union Agency for the Operational Management of Large-Scale IT Systems (eu-LISA) was dedicated to the implementation of interoperability among EU's border and migration management databases. One of the justifications that was repeated by several speakers was that of 'multiple identities' across databases and within a database. 83 The existence of multiple identities leads to a question about fake and authentic. Who or what is the authentic identity? Multiple identities are not simply disorderly, the effect of the multiplication of databases, sites of data extraction, and actors who process this data. Rather, the problematisation of multiple identities relies on the problem of the 'fake'. These identities are not simply multiple, but they are multiple 'fakes'.
The European Border and Coast Guard Agency (Frontex) also sees document fraud as a key challenge to European border security. The fraudulent document that Frontex highlights is not only a forgery or a fake document: New methods include obtaining authentic travel documents under false pretences, using such documents as fake birth or marriage certificates to obtain real documents, or using authentic documents of look-alikes. Both strategies use real documents, thus making it increasingly difficult for law enforcement officers to spot. Some forgers have also begun to use advanced counterfeiting techniques to circumvent automated border control. 84 Fakes become deeply embedded in the materiality of documents, as migrants allegedly pass authentic documents as fakes by forging or counterfeiting them. Detecting fakes then requires technologies to authenticate and reveal forgeries and fakes.
Fakes, forgeries, frauds, and counterfeits rely on a distinction between authentic/real and fake, and the ability of some to manipulate this distinction. The distinction authentic/fake produces a different subjectivity, as Stephan Scheel has analysed through the 'trickster narrative'. This narrative circulated by border professionals renders migrants as always under suspicion of producing fakes and it imbues them with 'excessive agency which needs to be tamed through security technologies like biometrics'. 85 The excessive agency of faking and counterfeiting can then only be tamed by experts equipped with technology. Discerning the fake from the real requires devices and methods of authentication. Yet, as anthropologist John L. Jackson reminds us, 'no matter 82 Ibid., § 26. 83 eu-LISA, 'Annual Conference "the New Information Architecture as a Driver for Efficiency and Effectiveness in Internal Security"', available at: {https://www.eulisaconference.eu/} accessed 12 October 2021. 84 Frontex, 'Document Fraud Is a Key Challenge in Border Control: Frontex Is on the Case', available at: {https://frontex. europa.eu/media-centre/news/focus/document-fraud-is-a-key-challenge-in-border-control-frontex-is-on-the-case-acJ0cG} accessed 16 February 2021. 85 Scheel, Autonomy of Migration?, p. 112. how sophisticated our authenticity tests, no matter how elaborate our analytical strategies, the slippage between what we see and what we can trust, between what is obvious or self-evident and what must be taken on faith about other people (based on partial and every-changing shreds of evidence), always remains.' 86 Errors, however, are underpinned by a subject who is not only vigilant but also trustworthy.
Starting from the analysis of how different forms of non-knowledge are problematised by authorities in the processing of asylum claims, we see how disjunctive subjectivities are produced through non-knowledge. Migrants are the ones who know too much so that they can forge, fake, falsify, and counterfeit. Underpinning the discourse of lack of credibility is the constitutive production of the subjects capable of fakery and deception. Fakes as a form of non-knowledge are also about an 'ethic of the relationship'. 87 To fake is to falsify or counterfeit both information and identity. By contrast, border professionals are the ones who appear in search for accuracy and authenticity, often aided by experts in technology, forensics, and biometrics. They can make errors but do not fake. Migrants are the ones who possess excessive knowledge, while border professionals are always described through a 'lack' of knowledge, needing more and more knowledge and a multitude of devices and technologies to correct errors. In the next section, we reflect on how these errors/fakes and related disjunctive subjectivities are inscribed within gendered and racialised power relations.

Non-knowledge, power, and subjectivity
The previous sections have shown that errors and fakes are underpinned by two different regimes of (non-)knowledge. Errors emerge in a co-constitutive relation with truth, which is to be pursued by means of objective methods and by vigilant subjects who have been trained into a 'corrective ethos'. Errors are mobilised in relation to what counts as valid knowledge and the distinctions true/false, objective/subjective, accurate/inaccurate. Fakes imply a relation to authenticity and a subject who knows too much confronted by authorities claiming to know too little. Fakes are considered inevitable or at least not exceptional among migrants who wish to settle in Europe.
In this section, we unpack how these regimes of (non-)knowledge are entwined with different forms of power and subjectivity. We draw on postcolonial scholarship to situate errors and fakes within colonial histories of science and capitalism, respectively. We also engage with feminist scholarship, which has questioned and disturbed Foucault's distinction between power and domination. We argue that we can understand the two regimes of error/truth and fake/authenticity through the feminist distinction between power and domination. Indeed, feminist scholars have recast domination to account for Foucault's analytics of positive power 'whose operation is not ensured by right but by technique, not by law but by normalization, not by punishment but by control'. 88 As we saw with the Advocates Abroad complaint, EASO's actions were initially framed as 'abuse' and connected with lack of knowledgeunprepared interviewers and their ignorance. When abusive decisions or administrative practices are rearticulated as 'significant errors', power relations can no longer appear in the form of 'abuse'. itself.' 89 She shows how 'failure' can even be perceived as an 'instructive' step of improvement rather than a problem with dramatic consequences for the professionals who are developing technological devices at the border. 90 While there are differences between errors and failures, they both enact accountability transfers and enable power relations. 91 The AA complaint points out systematic ignorance or what we have called, following Mills, 'white ignorance', which operates in oppressive systems. When it comes to making errors, we are closer to Foucault's understanding of reversible power relations, where power can be carefully calibrated between experts themselves and between experts and technologies. Yet, this calibration of power through the modulation of error does not extend to migrants themselves. Rather, it is fakes that appear to endow asylum seekers with excessive knowledge and power, while the authenticity of their narratives, documents, and biometrics needs to be established by experts aided by technologies. In this case, the authorities' lack of trust in Mr Bilali's testimony is identified as an error, while asylum seekers' testimonies are a priori under the suspicion of being fake or deceptive.
Expert and migrant subjectivities are differentially related here to knowledge and nonknowledge and determine what kinds of errors can be acceptable and what kinds of errors are not. Subjectivity shapes the distinction between what should be considered an 'error' or a 'fake'. As Roberts has put it, errors are 'the very grounds of subjectivity, belief and judgement'. 92 Subjects at the border do not occupy the same hierarchical positions, while errors and fakes do not have the same status, meaning, and effects. Experts are at the top and migrants at the bottom because the former can 'spot' and correct their own errors, while the latter are excluded from the remit of error. Error is a fundamental element of professional practices, but unlike the famous Latin formula 'errare humanum est' (to err is human), it does not apply to all humans.
In its connection to expert, accurate, and objective knowledge, error is situated within a quasi-scientific endeavour. Indeed, errors have been deemed inevitable in science, inherent to the process of knowledge production. As Bruno Latour and Steve Woolgar explained in studying the production of scientific facts, what we call objective facts are always social accomplishments that include subjective decisions, apparatuses, and inscription devices. 93 Errors are part of processes of making true and objective knowledge, even when they emerge out of subjective failures. Moreover, science and technology were 'an intimate and inextricable part of the colonial machinery'. 94 Thus, science also produced 'the Black Man as a racial subject and site of savage exteriority, who was therefore set up for moral disqualification and practical instrumentalization.' 95 When error activates a regime of quasi-scientific 'truth' as 'a system of ordered procedures' in the governance of asylum, migrants need to acquiesce to this regime if they want to be epistemically evaluated as refugees. 96 Achille Mbembe has argued that colonialism displays a very particular form of ignorance, which the colonisers passed for knowledge: 'a casual and frivolous ignorance that destroys in advance any possibility of an encounter and a relationship other than one based on violence'. 97 Colonialism joins (distorted) knowledge and ignorance under the guise of facts and objective 89 knowledge. More than 'white ignorance', the 'casual and frivolous ignorance' diagnosed by Mbembe blurs the boundary between knowing and not-knowing. Error, however, redraws the boundary between objectivity and truthful knowledge, on the one hand, and non-knowledge, on the other. Experts are supposed to know, even when they govern through error. Migrants are not allowed to err, but only to fake.
Power/error/truth constitute a regime where experts correct their own errors and those of technologies, where they calibrate their actions, so that 'abuse' becomes unthinkable. This has concrete political effects for Mr X in the EASO case and Mr Bilali in the Austrian Office case, as the 'errors' made by the authorities might have led to the death and/or deportation of two asylum seekers. In the case of the use of facial recognition technology, the FRA highlights how common technical errors consequently render suspicious certain groups, especially 'women and people of colour'. 98 They cannot make claims either against abuse or forged data but need to rely on experts to recalibrate technologies and undergo training to correct their own errorseven if it is too late for the lives of migrants.
Migrants and asylum seekers are relegated to a different regime of power/knowledge/nonknowledge, which we can specify as abuse/fake/authenticity. According to Benjamin Muller, identification procedures for asylum seekers have been turned into authentication processes. The politics of asylum are precisely now about discriminating the 'bogus fraudster' from the legitimate claimant through the verification of a 'well founded fear'. 99 Authentication tests are needed to dispel fraud, forgery, and fakery and produce the 'authentic' figure of the refugee. Bruno Magalhães has also aptly noted two types of registers that shape the ascription of 'clearly abusive' asylum applications in Brazil: a legal-evidential register and a socio-psychological register, which are often entangled in the authorities' decisions. 100 Unlike error/truth, fake/authenticity is underpinned by relations rendered as abuse rather than 'actions upon actions', as Foucault defined power relations.
Abuse needs to be understood within the history of 'racialized arbitrations of fake or real'. 101 Fakes and forgeries express modern anxieties about originality and copy, particularly as enabled by technologies of reproduction under capitalism. 102 The nineteenth century, especially, was haunted by fears of fraud, forgery, and counterfeiting, which were mobilised to justify the need for metricising the body to render populations knowable. 103 As historian Simon Cole explains, at this time '[v]iolent criminals were less objects of fear and fascination than con men and counterfeiters.' 104 Rather, it was 'crimes of identity and authenticity', which consumed late nineteenth-century societies in Europe. The fake assumes an alteration or perversion of the original and the authentic. When migrants fake, there is an assumption that they know too much, while authorities know too little. But more importantly, as we have seen, fakes are not just about hiding, imitation, or counterfeiting for one's advantage, but are also 'a certain way of giving the interlocutor the impression that he or she is similar to the listener and thus within their trust'. 105 98 FRA, 'Data Quality and Artificial Intelligence'. 99 Faking can be seen as entangled with imaginaries about the boundaries between self and other, what and who is similar and therefore trustworthy.
Guidelines and training provided by EASO are based on a regime of suspicion since they prepare experts to detect deception and decide on credibility. 106 The case of Mr X appears quite emblematic of this power relation since AA highlights how 'aggressive' was the EASO expert during the interview. But while AA designates what happened with Mr X as a question of 'abuse', the Ombudsperson's statement shows that she agrees with the EASO's re-formulation as 'error', and it endorses one subjectivity over another. Migrants' speech is not taken into consideration in the same way as the experts and technologies.
Fakes, in relation with excessive knowledge and agency by migrants, reverse the directionality of abuse. Migrants are now seen to abuse the system and even the authorities' trust. While abuse does not directly enter Foucault's analytics of power, it can be considered akin to his analysis of domination. As he argues, 'in speaking of domination I do not have in mind that solid and global kind of domination that one person exercises over others, or one group over another, but the manifold forms of domination that can be exercised within society.' 107 Feminist scholars have reinterpreted domination in order to analyse the forms of power implied in abuse. Vikki Bell has argued that, unlike reversible power relations, domination relies on a 'perpetual asymmetry'. 108 However, domination is 'still based on the operations of unstable tactics of power, but where a reversal in power relations appears to be almost impossible'. 109 Following this conceptualisation, we suggest that detecting fakes is akin to the exercise and effects of domination.
As a form of domination, abusesuch as in the case of child abusemeans that 'silence is the basis of subjugation', both in the sense that the child cannot speak and also that the child's subjective truth is ignored or denied: 'if the child asks for help, s/he is usually ignored, his/her truth and experience denied.' 110 This leads to a relation of 'co-dependency', which is based on the belief that 'only one discourse is valid and that there are no alternatives'. 111 In our case, the discourses of migrants and asylum seekers remain suspicious and untrusted. At the same time, as migrants are ascribed the power to 'abuse', they are also summoned into an inverted hierarchy of power. They are able to 'fake', as they appear endowed with excessive knowledge and power. It is in relation to this inverted hierarchy that experts can be rendered as lacking in power and striving to dispel the 'fog' of fakes with training and technology. Experts need to deploy authentication tests to distinguish the fake from the real. These authentication tests simultaneously objectify migrants and render them as excessively powerful. In so doing, authentication enacts a perpetual asymmetry between experts and migrants, as migrants can only be authenticated by someone else.
These different regimes of power/(non-)knowledge/subjectivity also raise the question of how they can be contested. For instance, Scheel has rejected the 'trickster narrative' associated with fakes and faking by rereading migrant practices as appropriating 'mobility within and against a security dispositive, whose institutionalised distrust, unpredictability and restrictive requirements render the appropriation of mobility a cumbersome and not always successful challenge'. 112 Given the 'unpredictable institutionalised regime of distrust', migrants appropriate mobility by having recourse to these practices of 'faking'. Other scholars and NGOs mobilise 106 the accusation of abuse against the authorities, as we have seen in the AA letter. Our juxtaposition of errors and fakes can also highlight a different tactic, one that extends the possibility of making errors to migrants themselves. The possibility of making errors could be a claim to equal humanity and political capacity.

Conclusion
The quest for more and more knowledge in the governing of populations, borders, and security has been well documented in IR. This article has proposed to contribute to interdisciplinary conversations about the production and problematisation of non-knowledge in different governance sites. While questions of ignorance, uncertainty, and doubt have shaped the field of ignorance studies and work on agnotology across the social sciences, scholarship on borders and migration has foregrounded the production of ignorance and its political effects. Rather than a quest for knowledge, scholars have shown the active and often strategic production of ignorance on the part of experts and authorities, and the co-production of knowledge and ignorance in governance.
This article has analysed the production of two different forms of non-knowledge at Europe's borderserrors and fakes. Through the juxtaposition of errors and fakes and a compositional methodology that brought together their problematisations across different sites of border and migration governance, we have shown how they are part of two different regimes of power/ (non-)knowledge/subjectivity. Starting with how different forms of non-knowledge are produced for border governmentalityand how they are contested in controversies over bordering practices -, we have traced a regime predicated on error and another on fakes. There are stark differences in how errors and fakes are deployed and differentiated at the border. We have examined how experts are deemed able to correct their mistakes, while migrants cannot err but only be responsible of faking. Errors are constitutive of a range of governing practices, from the processing of migrants' data to collecting information about countries of origin up to decisions on asylum applications. Technologies can also have errors, yet these non-human errors can be calibrated by experts themselves. Fakes, however, are characteristic of migrant stories, behaviours, and documents. Migrants are therefore subjected to repeated tests of authentication to dispel the supposed fog of inauthenticity.
Secondly, we have analysed the production of different subjectivities in relation to nonknowledge and not just knowledge. Starting with non-knowledge has allowed us to attend to the racialised hierarchies of subjectivity that are produced in regimes of error/truth and fake/ authenticity. Errors presuppose and foster a vigilant, quasi-scientific subject who is deemed capable of generating objective and valid knowledge while self-correcting. Fakes, frauds, and counterfeits produce a racialised hierarchy of subjectivity, where some subjects endowed with devices of authentication can decide on the in/authenticity of data, documents, stories, and bodies. Moreover, as anthropologist John L. Jackson has pointed out, distinguishing the authentic from the fake implies 'an unbalanced relationship between the powerful seer and the impotently seen'. 113 Finally, we have also engaged with the distinction between power and domination as reformulated by feminist scholars in order to show how errors and fakes are underpinned and reproduced through relations of power and relations of domination. As we saw in the AA complaint, errors cannot be intelligible in a regime where power is rendered as 'abuse', violence, or domination. Abuse is akin to domination, where relations of power are stabilised in a 'perpetual asymmetry' that places migrants at the top of a perverse hierarchy of excessive knowledge and power.
Errors cannot be extended to migrants, and fakes cannot be extended to experts given the production of these disjunctive subjectivities. Even if denunciations of deception, lying, faking, or abuse have been raised against experts, these are difficult to sustain, particularly in the absence of authentication tests. Claiming migrants as subjects allowed to make errors can be a potentially subversive, even if partial, intervention in the politics of (non-)knowledge at the border. It would only be partial, as any political claim to error would also simultaneously need to undo the strictures and histories of objectivity, impartiality, and validity that underpin assumptions of (quasi-) scientificity in border and asylum governance.
The analysis we have proposed here has wider resonances for IR in the way in which we envisage interdisciplinary conversations and research on entanglements of power/(non-)knowledge across dispersed sites of governance. While IR periodically resurfaces anxieties about being late to theoretical debates, we have shown how a vibrant interdisciplinary conversation about the study of ignorance and (non-)knowledge has developed in IR. Furthermore, the problem space of governing through non-knowledge, whose contours we have traced through a compositional methodology of errors and fakes, can highlight different arrangements of power and (non-) knowledge, their tensions and gaps, thereby multiplying critical interventions.