Algorithmic autoimmunity in the NHS: Radicalisation and the clinic

Abstract

This article explores the extension of counter-radicalisation practice into the National Health Service (NHS). In the 2011 reformulation of the UK Prevent strategy, the NHS became a key sector for the identification and suppression of ‘radicalisation’. Optometrists, dentists, doctors and nurses have been incorporated into counter-terrorism and trained to report signs of radicalisation in patients and staff. This article explores how calculative modalities associated with big data and digital analytics have been translated into the non-digital realm. The surveillance of the whole of the population through the NHS indicates a dramatic policy shift away from linear profiling of those ‘suspect communities’ previously considered vulnerable to radicalisation. Fixed indicators of radicalisation and risk profiles no longer reduce the sample size for surveillance by distinguishing between risky and non-risky bodies. Instead, the UK government chose the NHS as a pre-eminent site for counter-terrorism because of the large amount of contact it has with the public. The UK government is developing a novel counter-terrorism policy in the NHS around large-N surveillance and inductive calculation, which demonstrates a translation of algorithmic modalities and calculative regimes. This article argues that this translation produces an autoimmune moment in British security discourse whereby the distinction between suspicious and non-suspicious bodies has collapsed. It explores the training provided to NHS staff, arguing that fixed profiles no longer guide surveillance: rather, surveillance inductively produces the terrorist profile.

Introduction

The 2011 review of the Prevent strategy undertaken by the UK’s coalition government introduced three new developments in the relationship between health, security and surveillance. Firstly, it established that healthcare professionals would be trained to recognise signs of ‘radicalisation’ in their patients and co-workers and would have a legal duty to report suspicious individuals (through institutional processes) to the police. Since this review, then, doctors and nurses have been made active participants in counter-terrorism. Secondly, it extended the remit of ideological surveillance to the entire UK population through healthcare (and education), marking a discursive shift away from counter-terrorism practices that profile suspect communities. ¹ Indeed, the policy documentation explains that the extension of ideological monitoring to all citizens is a functional and necessary step. Thirdly, the Prevent review packaged this extension of counter-radicalisation surveillance using the healthcare terminology of the ‘safeguarding’ measure. National Health Service (NHS) safeguarding procedures are codified duties and responsibilities placed upon healthcare professionals to report any concerns about the physical, sexual and economic abuse of patients. By situating the new counter-radicalisation duties as a safeguarding measure to protect patients, the Prevent review explicitly refigured the terrorist as one who has suffered ideological abuse: a victim. It further embeds the notion of radicalisation as a pathological process towards terrorism that results from exposure to abusive radicalisers, and which can be recognised through discernible symptoms of ‘extremism’.

How did it become possible, within the otherwise uncompromising regime of UK counter-terrorism policies, for the figure of the terrorist to be associated with victimhood? And how did it become possible to monitor the ideological beliefs of the entire UK population through their healthcare providers, in a state which otherwise professes to support rights to privacy and freedom of expression? What politics of knowledge has effaced these contradictions, such that the Prevent duty has been implemented throughout the NHS with little public contestation?

This article argues that a mode of calculation associated with algorithms and big data methodologies contributes to these contemporary developments in UK counter-radicalisation. Central to this conclusion is the shift in profiling techniques, whereby extremism monitoring has been extended to all persons. The racial and religious profiles that originally underwrote the Prevent policy ² (which reduced the total ‘N’ to a suspect community and was intended to generate efficiency in surveillance resourcing) have shifted towards calculative modalities drawn from the digital realm. These now address the whole of the population through the NHS. In such inductive calculation, pre-determined ‘characteristics’ of a supposedly radicalising profile are not used to isolate suspects from a mass of population for the subsequent application of surveillance. Rather, the heterogeneous mass of data (patient contacts with NHS services, in this case) is reconceptualised as an asset, not a bewildering haystack that obscures the needle. Indeed, the complexity implicit within massive datasets becomes the modality of calculation, because it is presumed (in the style of big data methodology) that such complexity has the potential to reveal new patterns and connections between previously disparate factors through digital analytic techniques of partitioning and re-assembly (Amoore and Piotukh, 2015; Comfort et al., 2010).

By exploring the policy documentation and NHS training modules provided by the UK Home Office, NHS England and the Department of Health, this article documents the transition of the inductive calculative method from the digital realm into policies for surveillance within healthcare. To be clear, this article does not look at algorithms themselves, but rather the translation of calculative regimes from the digital arena to counter-radicalisation safeguarding in the UK. This is important because racial and religious profiles have historically served to limit the field of action for counter-terrorism, making associations between raced communities and the possibility of terrorism. These profiles provided an imagined possible landscape of suspects, less than the entire population, for further investigation. But the 2011 Prevent review has altered and augmented this calculative operation: in NHS policy documentation we can see evidence that surveillance is being rethought around a massively extended ‘N’. Its remit is now described as the entire population. However, this is not a replication of historical totalitarian surveillance, whereby certain characteristics (such as perceived dissent) lead, in a linear process, to arrest. This new population-wide surveillance demonstrates the translation of inductive calculative methods from the digital realm to the non-digital. This induction is associated with a non-linear method: pre-existing profiles are not used to identify the targets of security intervention; instead complexity is operationalised to inductively produce new terrorist profiles and subjectivities. In this rendering, the profile no longer guides the deployment of surveillance; rather, surveillance of the population now guides the production of the terrorist profile (Leese, 2014).

Louise Amoore’s work on the UK e-borders program has been pivotal in exploring algorithmic security calculation, showing how digital analytics combine discrete and unrelated pieces of data (such as travel histories and methods of payment) to constitute possible futures of risk or normality (Amoore, 2011). The mode of inductive calculation associated with algorithmic prediction involves the ingestion, partitioning and machinic reassembly of vast amounts of data – transforming them into knowledge ‘spoken’ into being by the algorithmic process (Amoore and Piotukh, 2015). The ‘data derivative’ is the name given to the judgement that emerges from this analysis of large amounts of data (Amoore, 2011; Leese, 2014). Such inductive methods usually reside in the digital realm, but the contribution of this article lies in mapping the transition of inductive, digital discourses of calculation to the non-digital arena. The algorithms themselves do not move into the non-digital arena, but the calculative regimes in which they are embedded do.

The Prevent review replicates the calculative modality associated with the algorithm by operationalising whole-of-population surveillance and reporting around unfixed and ambiguous indicators ³ of radicalisation. The surveillance is not interested in the reporting of pre-identified profiles; rather, it actively distances counter-radicalisation training from a reliance on static indicators. The training of NHS staff in counter-radicalisation is intended to impart a ‘professional judgement’ such that ‘risk’ can be identified despite the mutability and ambiguity of indicators (NHS England, 2015). The training describes the importance of developing an intuition, such that ambiguous profiles and combinations of risk factors can be managed. An attitude of creative alertness to radicalisation (and compliant reporting) is now required of NHS staff, as emphasised in official NHS England and Department of Health documents, rather than the prescribed reporting of set profiles. Such discourse represents a translation of inductive calculative discourse from the digital to the non-digital realm because safeguarding policies imagine the vast amount of NHS patient contacts as a productive asset which can generate previously unseen leads and connections. In the policy framing, the intuitive professional judgment of NHS staff is envisioned as the mechanism by which inductive calculation can occur in a situation of complexity. The calculative regime of digital analytics has been translated into policy documentation for non-digital surveillance.

It is important to note, however, that such a policy discourse does not necessarily lead to either: a) the replacement of other profiling techniques within UK security practice (such that race and religiosity are rendered irrelevant in surveillance judgments); or b) changes in the practice of NHS professionals. While the author is keen to develop future research within the NHS to ascertain the dynamics and effects of counter-radicalisation training, this article utilises only official policy documentation and online training modules to explore the framing of counter-radicalisation safeguarding within healthcare discourse.

The transition of inductive calculative logics from the digital arena to NHS staff policy indicates a framing shift in how the UK government thinks about biopolitical security. No longer does security policy strictly separate the realms of data and human. Now the calculative modalities applied to financial transactions and travel histories are used to frame the ideological surveillance of people by healthcare professionals. Human is data, and data is human. This is reminiscent of Katharine Hayles’s discussion of post-humanism, especially regarding the reflexivities which developed during ‘waves of cybernetic development’. In How We Became Posthuman (1999), Hayles discusses the progressive breakdown of Cartesian assumptions about subjectivity, mind and body when faced with technological advancement. Indeed, the reconceptualization of the subject as a material-informational entity is related by Hayles to the progressive insertion of the subject into digital knowledge-production: rather than standing outside the system under observation, the human has progressively been recognised as being a systemic component. Life and the digital have increasingly become understood as intertwined, mutual and inseparable (Hayles, 1999: 9–12). Hayles’ work seemingly anticipated the Prevent review, whereby the security calculation now relies on data produced by a human need for healthcare – a primary feature of human embodiment. The imagination of the subject within the Prevent review conceives of a material-informational entity: part embodied subject who feels pain and requires medicine, and part constituent of an informational flow.

Simultaneously, this post-human shift in governmental discourse indicates an autoimmune tendency. The datafication of life blurs distinctions between constitutions of the insecure, risky body and the hurt body requiring healthcare, rendering everyone potentially suspect by virtue of their embodiment. The inductive calculative method implies the radical potentiality that everyone’s embodied existence is implicitly productive of security data, revealing previously unknown patterns and connections. As subjects, we are thus constituted through an autoimmune gaze whereby the system cannot differentiate between healthy and radical cells, and turns on those cells previously afforded protection.

The article begins by exploring the historical relationship between health and security sectors, contextualising the novelty of the Prevent review within Foucault’s reading of biopolitics. While health and security share biopolitical functions in the constitution of population as an object of government, the move towards total ideological surveillance of the population (in a self-professed liberal state) indicates autoimmune tendencies – where, as Derrida teaches us, security turns upon itself to efface political aporia (Borradori, 2003; Mitchell, 2005). Subsequently, the article interrogates policy documents and NHS counter-radicalisation training modules to build the argument that inductive calculative modalities have here translated into the non-digital realm, replicating assumptions within big data methodology about induction, complexity and the making visible of previously unseen patterns.

The unhealthy matrix of health and security, and the coming of autoimmunity

The Prevent Strategy has constituted one quarter of the UK’s counter-terrorism strategy, ‘CONTEST’, since 2004, and outlines the responsibilities and roles of government and partner agencies in the detection and suppression of ‘radicalisation’. ⁴ The 2010–2015 UK coalition government undertook a review of Prevent in 2011, arguing that previous formulations of counter-radicalisation by the Labour Government had failed to tackle the ‘extremist ideology’ responsible for terrorism. In the resulting reformulation of Prevent, the NHS became a key sector for the identification and suppression of ‘dangerous’ beliefs and ‘pre-criminal’ trajectories (NHS England, 2015; see also Home Office, 2011). ⁵ Optometrists, dentists, doctors and nurses have been incorporated within security practice.

It is important to contextualise these recent developments within the historic relationships between health and security. This section will elaborate on the impact of the 2011 Prevent review upon the NHS, indicating that particular novelties that emerge in the relationship between health and security are situated within a historical affinity between the sectors. Health and security have never truly been separate, but the 2011 Prevent review marks the moment of a crucial development in their relationship. At this point, ‘radicalisation’ was refigured as a process involving the ‘abusive grooming of vulnerable people’ – effectively refiguring it as a safeguarding issue (much like domestic violence and sexual abuse) and state employees took on a duty of vigilance to identify and report such abuse of the vulnerable. Political violence, in short, became framed as a grooming and safeguarding issue for the medical community to recognise and report. While other sectors of society (especially police, but more recently schools) have been responsibilised for spotting the supposed signs of ‘radicalisation’ since the original Prevent strategy, counter-terrorism is now practiced by those with extensive experience in spotting symptoms – medical professionals.

To begin assessing the novelty and significance of the deployment of counter-terrorism through healthcare, we must first understand the historical relationship between sectors of health and security. In several regards, it would appear wholly unprecedented for counter-terrorism to be performed through the delivery of health. For instance, the 2011 Prevent review was the first piece of British legislation to introduce the overlap between counter-radicalisation and healthcare, although previous iterations of the CONTEST strategy had developed the capacities of other sectors of social security for the task. Before piloting counter-radicalisation with several health trusts in 2010, Prevent utilised ‘partnerships’ with police, charities, prisons, schools and communities to spread awareness and understanding of ‘radicalisation’ as the ideological symptoms and social vulnerabilities which precede terrorism.

Furthermore, before the ‘radicalisation’ discourse came to prominence in 2005 (Sedgwick, 2010), previous counter-terrorism campaigns also abstained from deploying counter-terrorism through healthcare. In Northern Ireland, for example, the British government eventually utilised certain provisions of social security to try and draw the allegiance of Belfast’s divided population away from paramilitary groups and towards a common civic identity organised around consumerism (Neill, 2010). Alongside brutal measures of internment, the suppression of civil rights and the extra-judicial execution of suspected paramilitary members, the British government attempted to foster the economic growth of Belfast in the 1980s and 90s through measures such as the construction of the new Castle Court shopping centre (Needham, 1998: 171), making a gesture towards an overlap between providing services for a population and practicing counter-terrorism. But there was no place for healthcare within the deployment of counter-terrorism in contemporary British history until 2010/2011.

Yet, despite this novelty of performing counter-terrorism through counter-radicalisation safeguarding in the healthcare sector, the overlap between security and health in international politics is fluid and long-running. Since the creation of the World Health Organisation in 1946, the health of all peoples has been recognised as a cornerstone for the attainment of security (Elbe, 2010: 1). While Stefan Elbe shows that this bold constitutional statement went largely ignored throughout much of the 20th century, the securitisation of viruses in the contemporary era as harbingers of potential catastrophe demarcates a clear shift in political attention. Health issues relating to Ebola, AIDS and pandemic flus have risen to occupy prime political terrain on the security agendas of international organisations. A growing literature in the field of global health analyses the ways in which health issues have been successfully securitised by actors, shedding light on the ways in which (inter)national security and social security sectors are merging (Davies, 2008; Heymann, 2003; Howell, 2011; Kamradt-Scott and McInnes, 2012). Given the prominence of these concerns, calls are now made to reclaim global health from its association with nation states (and Copenhagen School theorising) and to refigure the concept through emancipatory theory to expose relations of domination in the international system (Nunes, 2014).

Furthermore, as Alison Howell (2014) clearly shows in her Foucauldian contextualisation of health, medicine and security, healthcare has never simply been about curing illness, nor security about protecting the nation state – rather, both are intertwined in a symbiotic historical relationship whereby they have constituted population as a governable object. Drawing upon the centrality of medical knowledge and disciplines within Foucault’s oeuvre, Howell shows that it is redundant to think of health and security as two separate spheres that have recently come together. Health has not been ‘securitised’ (Davies, 2008; Kamradt-Scott and McInnes, 2012), nor security ‘medicalised’ (Elbe, 2010); instead, apparatuses of national and social security have historically permeated through and co-constituted each other. Howell provides historical examples of ‘public hygiene’ campaigns and psychiatric diagnoses being used to discipline and constitute the social body, demonstrating that medicine has never been separate from statecraft, before exploring the development of medical techniques during warfare which later crossed back into the domestic polity. As she succinctly summarises:

It seems easy to assume that curing and killing are at opposite ends of the spectrum of human activity. However, as I will argue, the ‘threat-defence’ logic does not get imported into ‘society’, health, or medicine through the so-called process of ‘securitisation’. Rather, modern medicine at least from the nineteenth century has always already functioned through this logic, precisely because it has been a science of social security and public safety at the level of the population. (Howell, 2014: 971)

This historical co-constitution is also evident in the contemporary era. Consider, for example, the viral framing of militancy used within counterinsurgency policies of the United States Army in Iraq and Afghanistan, which framed the colonisation of local ‘hearts and minds’ as inoculation against viral militant incursion. Similarly, to serve their counter-terrorist mission, the CIA invented a vaccination campaign against Hepatitis B in Abbottabad to collect genetic material that might indicate the presence of Osama Bin Laden’s family (Rubenstein, 2015). As Howell demonstrates, counter-insurgency and medical treatments of infection share a common logic.

As such, the UK’s deployment of counter-terrorism through healthcare could be contextualised as a relatively minor development in the longstanding relationship between security and health. So what is important about the British case, beyond the discrete novelties of the particular context? I will argue that two significant developments are evident here: the translation of calculative modalities associated with digital analytics, and a broader indication of autoimmune tendencies.

The importance of counter-terrorism performed through the NHS stems from what it reveals about the universalisation of counter-terrorism surveillance to an entire population, and the underlying translation of algorithmic calculative modalities from the digital realm to the non-digital practice of security. Specifically, the surveillance of the whole of the population through the NHS indicates a transition away from traditional profiling logics (which create suspect communities, reducing the total ‘N’, for the attention of the security services) to a scenario where every recipient of healthcare is subject to security screening. As I will discuss in the next section, this transition to a whole-of-population inductive calculative model is indicative of big data methodology and its transition to the non-digital realm.

But this mass surveillance simultaneously signifies an autoimmune development. There is an autoimmune logic at play in the transformation of counter-terrorism surveillance from a targeted practice to a whole-of-society practice. The UK has moved from the traditional security logic associated with immunity, which invokes an external other that could contaminate the national body, to a regime of practices that posit the mutating, unpredictable profile of the contaminant – and the escalation of protective measures against all cells, regardless of their healthy appearance.

The Derridean logic of autoimmunity shows us how the medicalised metaphor of immunity captures the binary oppositions that traditionally inform national security. The outside is posed as a threat to the body politic. According to this biopolitical logic, the health of the population is susceptible to contamination and infection by the other – a perfect metaphor for most identity discourses operating within national security practice. Yet, as both Derrida (Borradori, 2003) and Roberto Esposito (2008) show us, the immunity paradigm is actually constituted through an aporia – the dangerous outside is simultaneously taken into the body to produce the immunological security response. We must be exposed to contaminants in order to produce resistance. Therefore, the supposed ‘boundary’ between outside and inside is leaky and porous.

To conceal this conceptual deceit, the discourse of immunity folds the aporia back into itself, obscuring its presence, and the binary is maintained as sensible (Mitchell, 2005). But when this discursive ordering fails, and the separation can no longer be maintained, we encounter autoimmunity. Autoimmunity refers to the backfiring of a protective measure (the immune system within the body, or national security) such that it destroys that which it tries to protect. Derrida’s concept mirrors the operation of autoimmune diseases where the function of the immune protective system radically backfires. The protective system targets the internal healthy cells and becomes killer.

The Derridean concept of autoimmunity (Borradori, 2003) helps us to understand the radical extension of counter-radicalisation surveillance to the entire UK population. Both the NHS and the security apparatus of the British state were designed to effect the protection of the body politic. The discourses of both health and security utilise the metaphor of protecting the body from external threats and viruses, thereby constituting population as an object of government. They are, in this biopolitical reading, immunological systems that separate bodies-to-be-protected from unqualified bodies. However, after the 2005 bombings of London by British citizens, rather than foreign nationals, the radicalisation discourse became a prevalent mode of understanding terrorism (Sedgwick, 2010). Its operations signalled a turn towards the monitoring of the domestic population for internal signs of threat. This retasking of security towards the detection of rogue internal cells and the deployment of suppressive measures upon them speaks directly of an autoimmune response. Counter-radicalisation then targets healthy cells (parts of the national body) and attacks them, in its desperate search for the radical. The distinction between the inside and outside, the threat and the referent object of security, is radically disrupted in the autoimmune response (Mitchell, 2005).

Such autoimmune responses deny all knowledge of mistaken identification of threats. They reject the voices which assert the misdiagnosis of threat, such as the cellular assertion of normality and the critical assertions that ‘radicalisation’ has no predictive power and is not a real representation of the transition towards violence, but instead inappropriately stigmatises communities (Baker-Beall et al., 2014; Githens-Mazer and Lambert, 2010). This rejection of rational evidence and knowledge has elsewhere been called the epistemological crisis of counter-terrorism (Jackson, 2015) to highlight the silencing of knowledge regarding the inefficacy of counter-terrorist measures, and the relentless push to continue them anyway. The immune system refuses to heed the evidence that it is attacking healthy parts of the body or, in this case, its own citizenry. It is instead compelled to destroy through its corrupted logic of protection.

So, while health and security have long co-constituted population as an object of government in the biopolitical era, the radicalisation agenda and the incorporation of the NHS into counterterrorism practice are novel in their demonstration of the autoimmune, self-destructive culmination of this logic.

The significance of counter-radicalisation’s implementation through healthcare resides in the implosion of inside/outside and healthy/unhealthy distinctions into total ambiguity. The immunological distinction between bodies-to-be-protected and dangerous bodies, which long underwrote both health and security discourses, functioned to conceal the aporia that plagues identity distinctions (Mitchell, 2005). As Derrida (1982: 1–28) has shown through his notion of différance, self and other are never separate but are paired: reliant upon each other for distinction. But différance has now been deconstructed in the autoimmunity of British counterterrorism. The identities of the securitised body and the recipients of care have become ambiguous. This phenomenon, as I will argue in the next section, is coterminous with the translation of inductive, calculative modalities from the digital to the non-digital realm of security.

The algorithmic shift in British counter-radicalisation policy

This section will explore the translation of algorithmic calculative modalities into the non-digital realm, coterminous with the moment of autoimmunity. It explores the transition from the racial and religious profile of the potential terrorist to the contemporary refiguration of UK security discourse around mass surveillance and inductive profiling, where every body is rendered in post-human terms: as part material body, part informational flow that contributes to the mapping of radicalisation.

Before the Prevent review of 2011, the knowledge regime underpinning the radicalisation discourse imagined political violence differently. ‘Radicalisation’ was supposedly the result of a ‘vulnerable’ subject position juxtaposed against the proximity of contagious material and actors (like internet chat rooms, jihadist reading material and networks of ‘radicalisers’). ⁶ UK counter-radicalisation policy utilised explicit and implicit profiles of potential terrorist suspects to inform the distribution of policing resources. Its calculation was based upon the linear rendering of a static profile of the risky, racialized suspect and its potential transition towards violence. Indeed, the Prevent strategy began its life as a program of counterinsurgency delivered explicitly through racial demographics. Initial funding from central government for Prevent activities was allocated on the basis of 5% Muslim population density within local authorities (DCLG, 2007: 6). In keeping with the history of British counter-terrorism from the era of the Northern Ireland troubles, racial profiles of suspect communities constituted the counter-terrorism imaginary in the post-9/11 era (Awan, 2012; Baker-Beall et al., 2014; Fussey, 2013; Hillyard, 1993; Kundani, 2009; Pantazis and Pemberton, 2009; Spalek and McDonald, 2010).

Under the Labour Government, the Prevent strategy was reworked to soften this overtly racist tone, repackaging the deployment of suspicion upon minority communities under the rubric of ‘community cohesion’. It became ‘sayable’, in the climate of the ‘death of multiculturalism’ and the assimilatory demands made within policies of ‘community cohesion’, that community integration (specifically, the lack thereof) played a role in fostering the communicability of extremist ideas (for critiques of this trend, see: Heath-Kelly, 2013; Kundnani, 2009). The explicit focus on racial and religious demographics in the first iteration of Prevent was rephrased through an association of terrorism with ‘failed integration’ into British society. Prevent then funded a significant number of activities under community cohesion rubrics to integrate communities, and thus supposedly reduce the risk of terrorism.

However, in the climate of austerity, the 2011 Conservative-Liberal Democrat review of Prevent removed this funding of counter-terrorism through sport, discussion activities and training sessions (DCLG, 2008: 19), arguing that integration and terrorism are separate matters (Home Office, 2011). Surface-level attempts were made by the coalition to distance the association of counter-terrorism with the integration of British Asian communities. For instance, the coalition government responded to criticism of counter-terrorism as racially prejudiced by making tokenistic and half-hearted reference to right wing extremism as a threat of significant interest (Bentley, 2014).

Before the Prevent review, then, politicians attempted to show that Prevent does not target religious or racial communities. However the baseline assumption that underwrote pre-2011 counter-radicalisation is that extremist ideology and social ‘vulnerability’ combine to drive persons towards violent actions. A linear causality was drawn between ‘vulnerability indicators’ and terrorism. Prevent calculated and constituted terrorism as the result of a combination of factors including migration, conversion, displacement, exposure to extremist material in mosques or online, frustrated ambition, and (most recently) background tolerance of extremist rhetoric (David Cameron quoted in Wintour, 2015).

However, the post-2011 shift within counter-radicalisation policy demonstrated a move away from linear calculation and static risk factors. Counter-radicalisation has now been recalibrated to address the general population. The Home Office now conceives of radicalisation as the threat emerging from ideological grooming in all communities. All citizens are now identified as potentially referable to the Channel de-radicalisation program, run by the local police and government, for expressing ‘violent extremist ideology’ in front of their psychiatrist, doctor or nurse – and very little commentary in academia or the public media has noticed the shift imposed upon social care. We can note only mild concern voiced in the pages of The Psychiatrist (Witharana et al., 2012) and the British Medical Journal when considerations were made about balancing patient trust with new obligations to ‘keep watch on patients and try to predict what will happen in the future’ (BMA quoted in Dyer, 2011; see also English, 2011). The discomfort felt by the medical profession is evidently being kept on the back burner while they, at the time of writing, fight larger public battles against suspected moves to privatise more of the healthcare system

The post-2011 changes to the Prevent strategy explicitly demonstrate the translation of big data methodology to non-digital realms of counterterrorism policy. Algorithmic security operates according to an ontology of big data analysis. In this ontology, the sheer scale of such datasets is framed as an asset, rather than an impenetrable complexity, because the non-linear, algorithmic discourse of knowledge production asserts that more complexity and more data result in the production of more visibility. Scale, in this discourse, enables the invisible to become visible through analytic techniques of partitioning and reassembly, whereby correlations between previously unrelated data scraps become apparent (Amoore, 2011; Amoore and Piotukh, 2015; Comfort et al., 2010; Leese, 2014).

This inductive calculative frame (but not the use of algorithms themselves) has transitioned into the non-digital realm of security. For example, the incorporation of the NHS into counter-terrorism provision occurs because healthcare workers have extensive contact with the public. This, in the framing of the Prevent review of 2011, qualifies the health service as the preeminent site for the deployment of counter-radicalisation. We can see this fetishization of scale, and evidence of the inductive calculation agenda, in the strategy’s bold rationale for the incorporation of the NHS into the fight against ‘radicalisation’:

1.3 million NHS workers have contact with over 315,000 patients daily and some 700,000 workers in private and voluntary healthcare organisations see many thousands more … Given the very high numbers of people who come into contact with health professionals in this country, the sector is a critical partner in Prevent. There are clearly many opportunities for doctors, nurses and other staff to help protect people from radicalisation. The key challenge is to ensure that healthcare workers can identify the signs that someone is vulnerable to radicalisation, interpret those signs correctly and access the relevant support (Home Office, 2011: 83–85).

The government evaluates the importance of the NHS to counter-terrorism in relation to their unparalleled contact with the public, not because the NHS is a site of insecurity where terrorist plots are hatched. Whereas religious and racial characteristics were previously used to generate a profile of possible militant communities, thus streamlining the distribution of security resources, the relationship between the profile, the population and security has been reinvented. Now mass surveillance of a population sized ‘N’ is construed as the efficacious route for counter-terrorism. The dementia patient, the short-sighted and the population at large are now constituted as equally relevant subjects for counter-radicalisation surveillance, because all bodies have been reframed as informational flows that allow radicalisation to be detected more efficiently.

Instead of using a profile to limit the size of the ‘N’ when hunting terrorists, the Prevent review explicitly endorsed the big data methodology of obtaining the largest possible ‘N’, because increased sample size in modelling, in the logic of algorithmic security, is understood to have a positive impact on calculative efficacy (Amoore, 2011). Big data analytics privilege large sample sizes because they are understood to reveal patterns of correlation, invisible to the naked eye, from the digital evaluation of unrelated data scraps (Comfort et al., 2010). This algorithmic logic now appears in UK counter-radicalisation policy, where the NHS is specifically selected as a counter-terrorism partner due to its preeminent contact with the public. The policy documentation considers mass ideological surveillance as a large-N procedure that can produce to the terrorist profile through massive amounts of patient contact and tasking NHS professionals with reporting potential extremists. The policy, however, provides minimal training in recognising potential extremism.

It is also important to note that NHS training modules on the new counter-radicalisation duties leave the characteristics of the potential extremist largely unspecified. The training provided to NHS staff, especially the ambiguous profiles of risky subjects, demonstrates that they are not being instructed to recognise a static profile. Rather, the training attempts to operationalise inductive profiling, whereby NHS staff constitute the extremist profile as a kind of ‘data derivative’ – from, it is surmised, the intuition and ‘professional judgement’ (NHS England, 2015) gained from exposure to massive amounts of ‘data’ in the form of prolific patient contacts. The guidelines surrounding the Prevent competencies for NHS staff explicitly indicate that risk factors for radicalisation are not static but continuously evolve, therefore staff are required to operationalise the ambiguity of non-linear profiling beyond fixed indicators. They must:

Understand there is no single checklist or profile of a terrorist, and that health staff are a key group and must use their professional judgement in assessing behaviours and risks. (NHS England 2015)

The reframing of surveillance around intuition, rather than training staff to recognise fixed indicators, is also apparent in the ambiguous content of counter-radicalisation training. Figures 1, 2 and 3 show examples from the online training modules that accompany the 2011 Department of Health Guidance on the Prevent strategy. These predominantly situate ‘radicalisation’ within broader types of preventable abuse (including domestic abuse, child abuse and elder abuse). When accessing the training module, NHS staff are required to read forty-one PowerPoint slides on their duties of care towards vulnerable children and adults, occasionally completing tests on the correct identification of abuse and the required reporting response to line managers. The guidance turns to radicalisation as an issue that staff are duty bound to report, and which can be identified through the following indicators:

Figure 1.

But it is important to consider how one would distinguish ‘extremist material’ or ‘hate terms’ that fell within the remit of Prevent. What constitutes graffiti symbols, opinions and online material as terrorist-linked? How are NHS professionals to recognise specifically relevant indicators from the benign? The training offers no guidance on this, and serves to induce an organisational culture of reporting rather than expertise. Indeed, at levels 1 and 2 of Prevent Safeguarding competency training (the lowest levels), staff training is only intended to explain the objectives of the Prevent strategy, discuss some ‘vulnerability’ factors in radicalisation, and impart upon staff their new responsibilities to report radicalisation-related-concerns (NHS England, 2015).

I will argue that the characteristics of the potential terrorist are left blank because the inductive calculative logic of the new Prevent strategy requires them to be so. Surveillance is no longer understood as the result of a pre-existing profile of characteristics; it is now framed as producing the profile through large-N induction. The characteristics of the potential-terrorist must be left blank, so that a variety of patient-contacts can be reported and the new data derivative can come into existence. In accordance with the algorithmic logic of calculation, a big data methodology is applied to terrorist detection – and the characteristics of the potential militant must be left vague to enable the mechanism to produce a new profile through its access to massive amounts of population data.

The purpose of the training is to induce a culture of reporting, not to describe a fixed profile. Subsequent pages of the training module detail the specific contacts available for referrals, familiarising staff with the names and agencies they can find on the hard copies of documentation in the workplace. The module then directs staff to undertake the end-of-module quiz, to successfully complete their training. Question six tests retained knowledge of radicalisation indicators using a multiple choice format. From the outlandish inclusion of ‘joining the local art group’ as a possible answer, one can begin to question the goals of the training program:

Figure 2.

Through this provision of such an obvious wrong answer within the test, it can presumed that the training module is not intended to convey any expertise about ‘radicalisation’ – otherwise the test might direct the health professional to consider something more serious than the ‘local art group’. One could imagine a module which instead offered options regarding religious appearance or political dress, to generate a comparison between ‘indicators of radicalisation’ and uninformed speculation. But the module does not function as such; it instead offers easily identifiable – and slightly humorous – wrong answers.

We can thus assume that the test functions not to generate expertise of ‘radicalisation’, but rather to assert staff members’ personal responsibility to report. This would suggest the opposite intention to generating expertise through training, by encouraging staff not to question whether they are witnessing ‘radicalisation’ in a patient but to simply act, for the sake of contributing data for later calculation. In this algorithmic paradigm, reporting must be maximised because irrelevant data is no longer strictly considered irrelevant; rather, it is imagined within the terms of a large-N methodology as contributing to a richer picture from which patterns can be distinguished. For example, question seven tests the staff member’s recollection of reporting procedures – providing obvious wrong answers (‘forget about it’, ‘social media’) to consolidate the take-home message of reporting, and deferring, the situation to a manager:

Figure 3.

The training does not impart any detailed consideration of ‘radicalisation’, or any expertise. Instead it reproduces the already existing safeguarding expertise present within the NHS regarding the identification of patients suffering physical and sexual abuse, and redeploys it in the context of security. ‘Ideological’ grooming is slotted into a pre-existing apparatus for safeguarding vulnerable patients against physical, sexual, economic and emotional abuse in their wider lives. It is an extension of the everyday mechanisation of care – whereby symptoms of abuse become indicators which stimulate reporting and the subsequent involvement of other agencies.

The manipulation of these safeguarding structures for the policing of political thought and expression is, on reflection, an indication of contemporary securitisation in the UK. Healthcare professionals have been responsibilised for reporting ideological deviance without significant outcry from the public or media. Instead, the biopolitical constitution of population around an immunological boundary between self and other has collapsed into indistinction, whereby all bodies are securitised and surveilled.

Conclusion

Safeguarding against politics: Autoimmunity writ large

In this concluding section, the article considers the autoimmunity resonant within the conception of counter-terrorist ‘safeguarding’, whereby the terrorist is the product of preventable abuse. This collapsing of the discursive separation between health and security is the culmination of biopolitical logics, which deploy both risk and pity in their imagination of insecurity (Aradau, 2004; Heath-Kelly, 2013).

What does it mean to safeguard against terrorism? In the pre-2011 immunological era of British counter-terrorism, this would have meant the protection of the population from terrorist violence, but the autoimmune developments in post-2011 counter-radicalisation policy demonstrate a change in that meaning. Safeguarding against terrorism now refers to the identification of ideological abuse which prefigures terrorism. ‘Radicalisation’, in the Prevent review, has been reframed as a form of abuse that vulnerable children and adults can suffer, much like the domestic and sexual abuse which healthcare professionals are already required to notice and act upon within their ‘safeguarding’ protocols.

This is a remarkable depoliticization. Building upon previous policy articulations of ‘radicalisation’ as a process driven by risk and vulnerability factors (Heath-Kelly, 2013), political violence has now been further depoliticized such that it is categorised as a pathological behaviour that results from ‘ideological abuse’. For example:

Preventing someone from becoming a terrorist or from supporting terrorism is substantially comparable to safeguarding in other areas, including child abuse or domestic violence … The Department of Health has also supported the review of the ‘No Secrets’ guidance on safeguarding [vulnerable] adults. This will embed the principles of Prevent within existing processes for safeguarding vulnerable adults and enable healthcare workers across the country to understand the parallels between Prevent and existing support and intervention processes. (Home Office, 2011: 83–84)

This framing of political violence as a health and safeguarding issue is developed further in the Department of Health’s guidance document for healthcare professionals vis-à-vis the Prevent strategy, Building Partnerships; Staying Safe. Here they outline the compatibility of ‘safeguarding’ patients from being drawn into terrorism with existing procedures for dealing with vulnerable adults, framing both as activities of care:

Public sector agencies, charitable organisations and private sector bodies including health services need to work together to protect vulnerable individuals from being drawn into terrorism. Collaborative working also adds real value to patient care … It is important that we all share a responsibility for safeguarding and promoting the welfare of vulnerable individuals. In your work you may notice unusual changes in the behaviour of patients and/or colleagues which are sufficient to cause concern. Contracts of employment, professional codes of conduct and safeguarding frameworks such as No Secrets and Safeguarding Adults: The role of health services require all healthcare workers to exercise a duty of care to patients and, where necessary, to take action for safeguarding and crime prevention purposes. Through Prevent this will include taking preventive action and supporting those individuals who may be at risk of, or are being drawn into, terrorist related activity. (UK Department of Health, 2011: 4–5)

How has it become sensible to imagine terrorism as the result of ideological abuse? The depoliticisation of political violence is common within British history, with Margaret Thatcher provoking the Hunger Strikes of the 1980s in Northern Ireland’s jails by refusing to recognise the political status of Irish nationalist detainees, and Tony Blair participating in the post-2001 framing of Al-Qaeda as apocalyptic and evil rather than as politically motivated. But never before has the British state labelled convicted terrorists victims of a process of exploitation.

The Prevent review represents an important moment in the culmination of biopolitical logics in autoimmune deconstruction: their depiction of the figure of the terrorist has become completely ambiguous. It is both the vulnerable subject of social care and the hated securitised enemy. Claudia Aradau (2004) once identified the perverse biopolitical inclination to securitise victims of human trafficking against the future risk posed through their trauma, demonstrating the connections between logics of risk and pity in the constitution of population. British counter-terrorism has, however, taken this intersection to an autoimmune conclusion. The limit point has been reached when securitisation is deployed as medicalised safeguarding, and political resistance is framed as pathological symptom.

And it is the algorithmic paradigm of inductive calculation which made this final step in depoliticization possible. The mode of calculation which utilises a large ‘N’ to inductively partition and reassemble complex datasets enabled the move away from profiling suspect communities with identifiable characteristics. This modality enabled the reframing of transitions to terrorism as population-wide pathological processes to which all are vulnerable, and against which all should be safeguarded through the NHS. The effects of the big data paradigm have translated into the non-digital realm of counter-terrorism practice, enabling the culmination of autoimmunity.

Here social security and national security blur into indistinguishable processes. And while Foucault and Howell have pointed to their shared biopolitical logics over history, we do a disservice to the novel features of the Prevent review if we ignore the autoimmunity symptomatic within this collapsed distinction. The sick and the well, the dangerous and the protected bodies of security, have become ambiguous in the era of algorithmic inductive calculation. It is only when healthy cells can no longer be distinguished from threatening ones, in the autoimmune metaphor, that political violence can be pathologised as the result of abuse.

In conclusion, then, it is the calculative logic of big data which is responsible for the autoimmune implosion of British counter-terrorism, where the figures of terrorist and victim become interchangeable. This article has shown that the NHS has been selected as a counterterrorism partner of the Home Office solely because of the large-N potential evident within its public-facing operations. The decision to incorporate the NHS into counterterrorism was made so that counter-terrorism processes could have access to a bigger N. Ambivalent to the discursive separation of the health and security sectors in public discourse and liberal thought, the maximisation of sample size was prioritised above all else – demonstrating the commitment of governmental actors to big data ontologies of complexity and inductive analysis, such that their translation to the non-digital arena was not considered problematic.

This inductive calculative frame appeared again in the training provided to NHS staff to identify ‘radicalisation’. NHS staff, it seems, are not supposed to be looking for terrorist profiles or indicators (otherwise their training would be far more intensive). Instead, the calculative modality of big data analysis has spilled over into the non-digital policy world, and healthcare professionals are trained to refer patients to their institutional safeguarding boards without comprehensive expertise. Increasing the amount of data forwarded to the security services is the priority, not training NHS staff in fixed indicators of radicalisation, because algorithmic calculative paradigms prioritise the scale of data from which the derivative judgement of security appears. The historical logic of British counter-terrorism has been augmented so that, within the discourse of safeguarding at least, racialized logics of suspicion no longer determine the deployment of surveillance. Rather, the algorithmic paradigm has made it sensible to surveille the entire population so that a data derivative emerges from the array of disparate facts.

The question of the ethics of this shift becomes increasingly important, but beyond the scope of this article. Is it ethically preferable for counter-terrorism surveillance to be performed on an ‘egalitarian’, population-wide basis, or was it more ethical to deploy surveillance on the basis of reductionist profiles of racialized characteristics, given that it affected fewer people? Whatever the answer to this question, it is clear that the Prevent review reveals an important shift in the mechanics of suspicion.

To conclude, security and health have historically been intertwined in the constitution of population, but the autoimmune response evident in the Prevent review is the result of algorithmic regimes of calculation and the translation of inductive modalities from the digital to the non-digital realm. Thinking security efficacy through big data logics has led to the incorporation of social care sectors into the practice of counter-terrorism, effectively blurring the subjectivities of the vulnerable body and the dangerous body beyond distinction. The Prevent review has significantly altered the nature of profiling in British counter-terrorism, as a result of the algorithmic reorganisation of the human body as post-human assemblage of materiality and data. We now witness an era of algorithmic autoimmunity, where the NHS and the security services are merged and Foucault’s identification of shared biopolitical logics between the clinic and national security has come to full, self-destructive fruition. Mirroring Hayles’ reading of the post-human episteme (1999), the Prevent review dictates that all bodies must now be read for their implicit data; and all such readings enact the ‘self-learning’ of the inductive analytic, whereby NHS professionals gain their supposed ‘intuition’ for ‘radicalisation’. Patient-bodies are figured as part material entity and part information flow; and NHS bodies are figured as part material entity and part intuitive analytic.