Sage Journals: Discover world-class research

Abstract

The recent terrorist attacks and ongoing state of emergency in France have brought questions of police surveillance into the public spotlight, making it increasingly important to understand how police attain data from citizens. Since 2005, the French police have been using IBM’s computer program, i2 Analyst’s Notebook, to aggregate information and craft criminal narratives. This technology serves to quickly connect suspects with crimes, looking for as many associations as possible, ranking and visualizing them based on level of importance. Recently, surveillance and state power have been theorized as having shifted to a posthegemonic, order. Drawing from literature on power, surveillance, and identity, this paper considers the various ways that algorithms can impact policing under a state of emergency by comparing the technical protocol of i2 Analyst’s Notebook with the administrative protocol of the French state. Using i2 Analyst’s Notebook as an example, this paper argues that posthegemonic theories of power have their place in determining how algorithms can be used for surveillance, but that they cannot completely explain their use under the state of emergency.

Keywords

Surveillance police algorithm France state of emergency i2 Analyst’s Notebook

This article is a part of special theme on Algorithms in Culture. To see a full list of all articles in this special theme, please click here: http://journals.sagepub.com/page/bds/collections/algorithms-in-culture.

Computational research techniques are not barometers of the social. They produce hieroglyphs: shaped by the tool by which they are carved, requiring of priestly interpretation, they tell powerful but often mythological stories—usually in the service of the gods.—Tarleton Gillespie

Introduction

This paper attempts to outline the role that algorithms play in the French state of emergency by looking at how the policing software, i2 Analyst’s Notebook, organizes and visualizes data. Using police testimony from an investigation by the French national assembly, this paper shows the importance of data extraction as a police tactic under the state of emergency. The technical protocol of i2 Analyst’s Notebook, which allows for data extraction at its current scale, is compared with the administrative protocol of the nation-state. Using Agamben’s work on the state of exception, I argue that the state of emergency is a central function of liberal statehood, as opposed to a temporary lapse in democratic judgment. The algorithm and the data visualization are more than just the mediums of state control and they are part of its material constitution.

This argument brings along with it discussions of power operating as posthegemonic, and the “post-panopticon” era of surveillance, which raise questions of how social life is controlled and by whom. For theories of surveillance and power, this work focuses on Scott Lash’s “Power after hegemony” (2007), and David Lyon and Zygmunt Bauman’s Liquid Surveillance: A Conversation (2013). These authors, along with others, have made the case that power no longer operates from above in a traditional hegemonic fashion, but rather that it operates from below, in horizontal forms of communication. By putting surveillance literature in conversation with that of algorithms, I show how the state of emergency in France validates, but also complicates concepts of posthegemony, particularly regarding the state.

The state of emergency in France creates a unique case study for understanding the role of algorithms in the creation of political identity. For scholars that evoke Foucault, Agamben, and Deleuze to theorize the intersection of state power and surveillance technologies, I hope the example of France today will be a helpful exposé of state violence and identity. I draw heavily from Dan McQuillan’s “Algorithmic state of exception” (2015) and John Cheney-Lippold’s “A new algorithmic identity” (2011) to test these theories against the current state of emergency in France, showing the intersections and limits of current scholarship. Cheney-Lippold uses Deleuze and theories of posthegemony to show how algorithms create a fragmented identity of dividuality. For McQuillan, algorithms create an ungovernable space of exception that allows them to determine identity in this way. I argue that the creation of dividual identity aligns with the technical protocol of i2 Analyst’s Notebook, and the administrative protocol of the French state, but that it is not necessarily indicative of posthegemonic power, nor a result of states of exception created by the algorithm itself. The state of emergency is an example of how the state creates dividual identity, not through exception, but as a rule of its existence. For some, the increasing use of algorithms in governance and policing is seen as inhumane or undemocratic. My goal is not to discourage this conversation, but to take it further by making an ontological connection between how both algorithms and the state operate through identity, control, and exclusion.

State of emergency

In response to the terrorist attacks in Paris on 13 November 2015, France has been under a state of emergency which has drastic effects on law enforcement. Legally, this allows the police¹ significant freedoms that are not limited to conducting house raids and putting people under house arrest without approval from a judge, collecting and monitoring information obtained from potential suspects, and the ability to block websites and prevent public gatherings (Human Rights Watch, 2016a). These freedoms allow police access to vast amounts of data, with over 18 terabytes collected and analyzed since November 2015 (Raimbourg and Poisson, 2016). Since November, over 4000 raids have been executed and more than 500 people have been put under house arrest (Human Rights Watch, 2016b). The state of emergency has now been extended for a fourth time, which will last until July 2017. In an era of policing that relies heavily on computer software to aggregate information, it is important to understand the potential these technologies have for influencing surveillance and policing.

While much of the legal focus in France has been on the changes during the state of emergency, impactful changes to data privacy and security laws took place earlier, in 2014 with the LPM Act (loi de programmation militaire). Along with allowing the deployment of 7000 soldiers during times of crisis, the LPM Act also “provided an opportunity to modify the legal regime applicable to administrative access to communications data, in order to give counter-terrorism services broader access” (CNIL, 2015). The Charlie Hebdo attacks in January 2015 activated the LPM act and brought along with it the Intelligence Act of 24 July 2015.² The Intelligence Act allowed intelligence services to hold communication data of potential suspects for up to five years (previously one year before the LPM), wiretapping and electronic surveillance were extended beyond just the suspect to include family and friends, authorized the use of keyloggers and other software to monitor and record computer use, and introduced IMSI catchers, or fake cell phone towers that capture mobile phone data (CNIL, 2015). These previous legal decisions have affected the way data has been collected and analyzed during the state of emergency.

The legal parameters of the state of emergency have fluctuated since it has been in effect. The period from November 2015 to February 2016 saw the highest number of raids and house arrests. When the state of emergency was continued in February, the constitutional court deemed it unconstitutional to seize data from citizens during house raids. In the period immediately following this decision, from February until July, raids and house arrests were at their lowest. However, after the terrorist attack in Nice during Bastille day celebrations on 14 July 2016, the state of emergency was extended again, with a reevaluation of data seizure laws. For the newly ratified state of emergency, French police can seize property such as cell phones and hard drives, on the condition that a significant connection to a crime or terrorist plot can be proven. For this data to be stored, connections to a crime must be ratified by a judge, which according to the vice-president of the State Counsel, Jean-Marc Sauvé, has a 90% success rate (Jacquin, 2016). Since this ruling, house arrests have increased 50%, and raids are up as well (Pascual, 2016a).

Unsurprisingly, these raids and house arrests have been carried out along highly racialized and religious lines. Due to the strict secularization laws in France, the state does not have official census data on race or religion, making quantitative answers to these questions difficult. Estimations of the Muslim population in France are around 8%, mostly of north African descent, with the most well-known study done by the French Institute for Demographic Studies (INED, 2010). However, with a Muslim population of only 8%, it is estimated that Muslims make up 50–70% of the prison population (Khosrokhavar, 2013). With this massive overrepresentation of Muslims in prison occurring for many years before the state of emergency, certain a priori assumptions about discrimination in the implementation of the state of emergency can be made. It is worth noting here that despite the secular resistance to providing census data for the Muslim population in prison, the French state seems to have no qualms about making its antiterrorism programs explicitly tied to stopping radical Islam. This can be seen clearly in their antiradicalization campaign started in 2014 called #alwaysachoice which provides people with a nationwide hotline to call if they think someone they know is becoming a radical jihadi. One of the many problems with this approach is that it ignores all other forms of radicalization, in favor of only seeing radicalization as a process that happens within Islam. For some, this level of discrimination is explained away by the fact that most severe terrorist attacks in France over the past five years have been associated with ISIS, yet the degree to which the French state selectively uses its secularization laws is evident.

Some of the best evidence available so far regarding how the state of emergency has been implemented by security forces in France comes from a report by the French national assembly, in which different security administrators and police officers testified. One of these officers is Colonel Charles-Antoine Thomas, the head of the gendarmes in Val d’Oise, a region that covers much of the northern suburbs of Paris. When asked how long house raids usually take, Colonel Charles-Antoine Thomas replied: “it depends on the amount of data that’s discovered; that’s what takes us the longest. For us, the best part is the data, because sometimes we can find signs of logistics networks, nobody sleeps with their AK!” (Raimbourg and Poisson, 2016: 108). These sources point to the value of data accumulation in the house raids under the state of emergency. It is software like i2 Analyst's Notebook that facilitates this type of policing, by providing a tool that can sort, order, and present the large amount of data collected through these raids. This approach relies on data visualizations to analyze a high volume of data in such a short amount of time (Seidler and Adderley, 2013).

i2Analyst’s Notebook

In 2005, France purchased IBM’s i2 Analyst’s Notebook (hereafter known as i2AN) in conjunction with the beginnings of criminal analysis programs SALVAC³ and ANACRIM⁴ (Polloni, (2015)). These initiatives were part of a broader plan to bring government data online in a centralized and easily sharable format. They defined how long police could keep personal information about suspects and victims on file, as well as created unified databases for storing this information. ANACRIM was recently updated two years ago to help automate the retrieval of information from all police databases so that it can be more easily integrated into their network analysis software, i2AN (CNIL, 2014). According to IBM’s website, i2AN,

is a visual intelligence analysis environment that can optimize the value of massive amounts of information collected by government agencies and businesses. With an intuitive and contextual design it allows analysts to quickly collate, analyze and visualize data from disparate sources while reducing the time required to discover key information in complex data. (IBM, 2016)

One of i2AN’s most important functions is to import data from multiple datasets and produce visualizations that will help reveal connections that would not be made otherwise. This type of software enables police forces to access large amounts of public data while conserving resources and time (Kennedy et al., 2015). Similar initiatives to digitize and centralize police records have occurred across the world to take advantage of policing software. If the databases are not entirely readable by the data-mining algorithms in the software this can lead to inefficiencies and error (Kantardzic, 2011).

Azzam et al. (2013: 8) define data visualization as “a process that (a) is based on qualitative or quantitative data and (b) results in an image that is representative of the raw data, which is (c) readable by viewers and supports exploration, examination, and communication of the data.” This definition helps clarify the purpose of data visualization, which is to speak and interpret on behalf of information that might not otherwise have a voice. For the French police, i2AN’s visualization tool is most valuable in analyzing criminal networks. This role is explained in the police journal National Gendarme Review,

Intelligence network analysis is an essential step in the adoption of action and a prerequisite to decision making. i2 Analyst’s Notebook permits analysts to quickly class, analyze and visualize data from multiple sources. It reduces the time necessary to recognize necessary information to craft a solid and timely network which allows the identification, prediction, prevention and interruption of terrorist, criminal, and fraudulent activities. (Corrieu and Sajhau, 2014: 50)

The use of the visual not only allows for a unique view of the data, but it also represents the speed and processing capacity of the software. In ideal circumstances, data visualization promotes exploration and examination like the above definition denotes, but sometimes the opposite is true. In certain situations, it is possible for the visualization themselves to become the answers instead of an interpretive tool for decision makers.

Despite the inherent notion of objectivity in the word data, our social world and technologies shape and rationalize our understanding of facts. This is especially true for police work. Dealing with environments that are constantly in flux, there is never a situation where officers will have all the available information on a subject or crime. Widespread use of the internet, mobile phones, and other geolocatable technologies (bank cards, transit pass, etc.) have vastly expanded the data available to police departments (Alison, 2014), which have responded by introducing software that aggregates information from multiple databases in real time. With so much information being processed at once, data visualization becomes a vital component in the synthesis and comprehensibility of data. To make the information intelligible to analysts, visualizations create cognitive shortcuts for faster and more efficient information processing (Kantardzic, 2011).

i2AN is not a predictive policing software like the more popular software PredPol. For PredPol, algorithms play the role of the analyst once the data has been collected (usually a process done by other algorithms). What is usually meant by predictive policing is algorithms making predictions for where and when crime is going to occur. For i2AN, algorithms data-mine and provide visualization readouts of that information, but a human analyst is still required to decide about a target or action. The level to which this software constitutes predictive policing is based on several factors, the most well-known being algorithms that attempt to replace the role of the analyst in evaluating the results of the data-mining process. However, the presence of a human analyst does not preclude policing from being predictive to some degree, even if it doesn’t fall under the traditional definition of predictive policing. The processes of data-mining and visualization are correlative in nature (boyd and Crawford, 2012), and many of these policing software are based on theories of crime that value the correlations of past crimes to predict future ones (Benbouzid, 2015). The human analyst is still making a prediction of the future based on past events, so the major difference between our “predictive” and nonpredictive policing software is determined not by the nature of prediction, but who/what is doing the predicting.

Clarifying the similarities is important, since much of PredPol’s popularity has come in the form of critiques about its predictive capacities, without a deeper evaluation of the underlying connection between search algorithms and the state (Brakel and Rosamunde, 2016; Kutnowski, 2017). New algorithmic categorization has changed how power is distributed through the coevolution of technologies and governance (Just and Lazer, 2016). Police software such as i2AN can take the form of a list-plus-algorithm, with the list of inputs serving as the background on which the algorithm computes.Fleur Johns (2016: 128) traces this problem to the protocol of list-plus-algorithm software like i2AN.

Lists-plus-algorithms are envisaged doing many useful things: prompting redirection of resources towards areas of greatest need, for instance. Yet, for many, lists-plus-algorithms dissemble as much as they disclose, perturb as much as they excite. Even as they are identified with enhanced capacities of prediction and preemption, these practices also evoke a sense of diminished capacity[…]The use of list-plus-algorithm for global governance seems, at once, an apotheosis of rationalism and a condition of reason’s demise.

For i2AN, and other similar policing software, information about a subject or object in question is mined from multiple private and public databases, depending on the level of access granted to the agency in question. The data then appear in visualization for an analyst, or another algorithm, to evaluate. When the simple mining of databases is not enough, the information is extracted through more forceful measures such as a house or business raid in which cell phones and hard drives are confiscated. The very real threat of terrorism in France today has exacerbated the need for the French state to go to these extreme measures, underlining the diminished capacity that Johns mentions of these algorithms to accurately identify terrorist threats.

These lists-plus-algorithms must briefly be put in their historical context, as their development is part and parcel to that of capitalism (Mager, 2012) and modern warfare (Amoore, 2009). Capitalism, the liberal democratic state’s counterpart, relies on the commodity form to organize and distribute goods. Search algorithms are built to take a commodity and divide it into its material inputs for supply chain organization or into its constituent cultural value so that it can best be advertised to consumers. i2AN is a direct descendent of the decision-tree algorithms that, originally used for retail, have now been adopted by militaries for counter-terrorism and cybersecurity. The primary role of these software is to “translate probable associations between people or objects into actionable security decisions” (Amoore, 2009: 55). To accomplish this, a unified system of addressability had to be created. Amoore traces how the development of these technologies went from addressability to locatability. Addressability allows one’s location to be searched, while locatability allows one to be tracked. Individuals are broken down into their constituent parts and repackaged to craft a network that is deemed the most lucrative for identifying potential suspects. The ability to quickly leverage algorithmic decisions about potential terrorist targets through house and business raids increases the effectiveness of this apparatus, which is historically linked to both commodification and warfare. House arrest is a highly accurate form of locatability, where those arrested are required to check in to a police station up to three times a day. However, this method is costly and doesn’t cast a wide enough net. Going after a hard drive or cell phone reduces the need to use other, traditional methods of obtaining information such as interrogation and torture, and can provide a wealth of data on criminal networks that can then be analyzed with i2AN. The use of the state of emergency to accomplish these goals can be followed to the protocol of both i2AN and the state.

Protocol

The concept of protocol is used in many ways, with two of the more common uses being the operations of a technical system, and a set of administrative rules. Both definitions are valuable in understanding how the technical protocol of software such as i2AN reflects administrative protocol of the French state of emergency. Algorithms play a large role in surveillance and search, and it is their technical protocol that determines how information is stored and accessed (Galloway, 2004). Galloway defines protocol as “against interpretation. This is to say that protocol does little to transcode the meaning of the semantic units of value that pass in and out of its purview” (p. 52). Simply put, protocol is the potential for existence within a system or software. “The limits of a protocological system and the limits of possibility within that system are synonymous” (p. 52). Protocol is not descriptive of data pathways within software, it is those data pathways. Understanding protocol allows for actions by the police to be viewed through the possibilities of their software. Rather than a deterministic view of technology’s influence on policing, protocol provides just one factor out of the many technical, political, religious, and economic factors that shape police action in the state of emergency.

When looking at the technical protocol of i2AN, one can start with its GUI-WIMP interface (Graphical User Interface-Windows, Icons, Menus, and Pointers) that is accessed predominantly through a desktop computer. This means that the i2AN interface responds to user’s actions with the mouse and keyboard. There are certain affordances, imagined and material, that are present while working with i2AN on a desktop that work together with technical and administrative protocol to construct possible action, (Nagy and Neff, 2015). A 2D social network analysis, in the form of a node and link graph, is laid on top of a GIS map, where suspects can be dragged around the screen, enlarged with a double-click, and ordered based on their importance. These are all components of how criminality is visualized and understood in i2AN.

The default screen for working on a new project in i2AN is a blank page surrounded on the top and right-hand sides with toolbars. To begin working the user must import a dataset that has values recognizable to i2AN. These datasets have been compiled by data-mining algorithms and generally have not yet been interpreted by a human analyst. The program has multiple options for data visualization and gives its user multiple options for how they want the clustering and centrality of the nodes to be weighed and visually displayed, which augments the focus of the graph. These options can be checked or unchecked, leading to a large variety of outcomes based on the user’s preference. i2AN also gives users the option to customize the weight of each link in the network to emphasize different degrees of connection. The magnitude of these options is only a sampling of the various ways a single dataset can be interpreted and is an instructive glimpse at the possibilities of i2AN.

i2AN’s use of node and link graphs for visualizing social network analysis brings up issues of accuracy, particularly under the state of emergency, where the goal is to find out about future terrorist plots, or terrorists, from the inside out. Node and link graphs are not known for being a good visualization for this type of risk (Hall et al., 2015). For these visualizations to adequately represent the relationships between nodes, there needs to be a clear identity distinction between the insiders and outsiders of a network. Commenting on node and link visualizations, Hall et al. (2015) note that “visualization presupposes fixed behaviour types: insiders are either loyal or malicious. Such a distinction complies with militaristic approaches of the past, but in cloud computing the distinction between insiders and outsiders is not easy to make” (p. 95). It is important to note that the theoretical background for terrorist network analysis is still developing (Lindelauf et al., 2011). Despite adopting theories from other fields that use social network analysis, “the evidence concerning terrorist network structures however is often anecdotal” (p. 63). While identity descriptors such as name, race, gender, might be known, the meaning of that identity in terms of its threat may not be. As this article will attempt to show, distinguishing insiders from outsiders is a task that algorithms attempt to accomplish. This task is central not only to the state of emergency but to the state itself.

Administrative protocol also represents limits of possibility, but instead of a technical structure, it applies to a social structure. The nature of the sovereign and the state of exception reveal the administrative protocol of the nation-state. In allowing a sovereign body to govern and make laws that all must follow, it becomes quite difficult to prevent that body from breaking its own rules. This rule of exception is a feature, not a bug of sovereign statehood.Early philosophical investigations into this dual quality of the state begin with Schmitt C (2005), in his essay Political Theology, where the sovereign is that which defines the nature of exception. Further work by Ernst Kantorowicz (1957) in, The King’s Two Bodies, looked at the dual nature of governance in medieval Europe, coining the popular phrase “body politic.” To reference the above quote by Fleur Johns (2016: 128) that the state is the only one to both set the rules and call time out is both “an apotheosis of rationalism and a condition of reason’s demise.”

The contemporary philosophical work on the state of exception is centered around Giorgio Agamben’s (1995), Homo Sacer series, which draws heavily from both Schmitt and Kantorowicz. At the surface, it seems easy to separate exception and rule into two different legal categories, but Agamben stresses that exceptional legal circumstances in fact define the rule. “Only because its validity is suspended in the state of exception can positive law define the normal case as the realm of its own validity” (p. 17). The ontologically inseparable connection between rule and exception demonstrates the administrative protocol of sovereign statehood. The state of emergency is thus a defining feature of the French state, used to redefine who it considers as citizens internal to its rule, and who it considers as external threats. How algorithms affect this distinction between inside and out is discussed further in the section on identity and exception. Both administrative and technological protocol shape policing and the use of algorithms, and it is through this lens that the state of emergency can be used to question theories of power and posthegemony.

Power and surveillance

The implications of algorithmic police surveillance in France must be evaluated in the context of theories of power and hegemony. Along with developments of postmodernism in the second half of the 20th century, critiques of modernist analyses of power began to emerge from thinkers such as Michel Foucault and Jean Baudrillard. This line of thinking has involved the term “posthegemony,” which replaces structuralist narratives of power from above with power from within or below. The modern usage of hegemony is commonly associated with Antonio Gramsci, to refer to a source of totally dominating power, be it the state, capital, colonialism, etc. Many others have taken up this term in different disciplines, with a notable example being Stuart Hall in cultural studies. It is through this conception of hegemony within cultural studies that Scott Lash roots the move to posthegemony. One of the tenants of this theory relies on how we perceive and communicate data, information, and fact. Lash points to the algorithm being an integral medium for communicating power and domination in a posthegemonic era (Lash, 2007).

For Lash, the difference between hegemony and posthegemony relies on where power is located. “The hegemon is above. It is outside and over. In the post-hegemonic order, power comes to act from below: it no longer stays outside that which it ‘effects’. It becomes instead immanent in its object and its processes” (Lash, 2007: 61). One of the mediations of this power from below is in “a change from an extensive (and hegemonic) regime of representation to an intensive regime of communication” (Lash, 2007: 56). Following theories of networks and the multitude from Hardt and Negri's Empire, Lash sees posthegemonic control in the communication networks made constant by ubiquitous media and the protocol through which the media technologies are constructed. Since many agree we are in the era of ubiquitous media and computing, the potential for algorithms to produce posthegemonic power is vast (Kitchin, 2017; Lash, 2007). In this sense, the algorithm does embody posthegemonic power from below since it is embedded in the architecture of mediated communication. Data becomes a higher priority for police because mediated, and thus searchable, recordable, and organizable, actions encompass a greater portion of total human actions. When viewing power at the level of the algorithm, a shift to posthegemony seems plausible. However, forms of data extraction seen in the state of emergency in France place the role of the algorithm within a broader structure of state power, demonstrating the limits of posthegemony.

If the necessary power to control society was produced only from below, through our networked communication and with the help of algorithms, there would be no need for the state of emergency, or perhaps even the state itself. The state of exception demonstrates the necessary separation of inside and outside, above and below, that is constitutive of the state. Algorithms as material lines of code reflect this separation as well. They interpret information from the inside, to produce a result from the outside. It is the act of interpreting, through code and visualization, that defines this separation in the use of algorithms. When interpreting the use of data and information under posthegemony Lash writes: “in the post-hegemonic age the concept is dis-intermediated and we come up against facts as raw sense data. In their rawness, facts come very much alive” (p. 64). It may be the case that we as subjects encounter this raw sense data, but the goal of i2AN for the French police, however successful, is to remove this rawness and make the data comprehensible so that it can be acted upon. The state of emergency exists as an admittance of the failure of posthegemonic forms of power to guarantee social peace. i2AN may help govern from below in some posthegemonic fashion, but the data it receives, extracted from house and business raids, demonstrates the hegemonic form of domination that still persists.

Following the move to posthegemony, the term postpanopticon has been used in the field of surveillance studies by Zygmunt Bauman, and David Lyon, in tandem with the phrase liquid modernity. The concept of liquid modernity is analogous to posthegemony in that it highlights the flexible nature of control and power. Postpanopticon describes a mode of surveillance that is not determined by direct state observation, as was the case with Jeremy Bentham’s panopticon, but by indirect monitoring, prediction, and self-surveillance (Bauman and Lyon, 2013). Self-surveillance to this degree is made possible by ubiquitous media use and algorithms. Postpanoptical prediction creates what Lyon calls a data double. “The information that proxies for the person is made up of ‘personal data’ only in the sense that it originated with a person’s body and may affect their life chances and choices” (p. 8). This is how i2AN operates, by creating a conglomerate of data points about an individual and trying to connect them to a network of other potential suspects. Finally, Bauman sees the traditional era of panopticon surveillance still in operation today, but only for certain people. In the era of liquid modernity, more hegemonic forms of panopticon surveillance operate at the social margins, in prisons, at borders, and the outskirts, while those with more social value and wealth are surveilled at a postpanoptical distance (Bauman and Lyon, 2013). What is of interest here is what these theories of surveillance can tell us about the current state of emergency in France and what role the algorithm plays.

As mentioned above, the state of emergency exposes the state as a traditional, hegemonic form of power, despite methods of control that might be deemed posthegemonic. The state of emergency is the enclosure and reduction of the free space that was previously purchased by social classes willing, and able, to surveil themselves. In the case of house arrests, the social margins creep into everyday life, making the home a prison, and the neighborhood a border. One of the central functions of the house raid is to help police forces redraw these boundaries by gathering data of criminal networks. This is demonstrated in a conversation between the minister of justice, Jean-Jacques Urvoas, and the departmental director of public security in Loiret, Pascal Belin.

Urvoas: In your opinion, is intelligence gathering the main benefit of the state of emergency?

Belin: I would say the interest in house raids has diminished really quickly. For example, with this group of Chechens, the first house arrest and raid focused on the treasure of the group; the examination of their computers - with documents in cyrillic … - that was very interesting. But the rest of the group quickly went underground. The house raids were no longer valuable: we found wiped telephones, no call records, no text messages, no contacts … just like the computers, completely empty. (Raimbourg and Poisson, 2016: 121)

While arresting criminals and ending the threat of terrorism are clearly the overall goals of the state of emergency, the more immediate and specific goal of the house arrest and raid, as tactic, is access to data that would not be available under normal legal circumstances. Those that previously existed at a postpanoptical distance, now feel the immediacy of hegemonic state power. With this information, one’s data double can be used to discover criminal networks and lead to future arrests.

Identity and exception

Surveillance tools that function using algorithms bring up questions of identity and formation of subjectivity under the state. Cheney-Lippold’s “A new algorithmic identity” (2013) explores this topic extensively and offers a base on which identity in the state of emergency can be analyzed. Cheney-Lippold uses Deleuze’s theory of the dividual and Foucault’s work on biopolitics and power to argue that algorithmic search and surveillance create floating, adaptable identities, analogous to Bauman and Lyon’s data double. This allows for soft forms of state control, as seen in the posthegemonic era, to replace the hard power of previous state discipline. “Many mechanisms of control have shifted to practices external to the body, where discipline has often become more or less unnecessary if control can be enacted through a series of guiding, determining, and persuasive mechanisms of power” (pp. 168–169). This form of control creates dividuality, an idea originating with Deleuze, which refers to individuality that is constantly subdivided into constituent parts, or dividuals. “These dividuals become the axiom of control, the recipients through which power flows as subjectivity takes a deconstructed dive into the digital era” (p. 169). Control in this sense is the form of power that modulates conditions of possibility without direct, disciplinary uses of force. It seems that while softer forms of control certainly fit with some of the technical protocol of search and surveillance algorithms, this does not exclude those algorithms from also serving hard discipline. The example of France during the state of emergency shows that those algorithms can be used for disciplinary roles, while still treating the citizen as a dividual. After all, the goal of the state of emergency in creating dividuality through i2AN is to capture terrorists, the hyper-targeted individual.

The raid and house arrest are tactics that perform a particular type of surveillance. They are conducted on individuals that could be dangerous to the state, but who do not have enough evidence against them to be arrested. Their dividuality, or data double, has not solidified into a credible identity for the state, and thus requires further investigation. This identity monitoring during the state of emergency is explained again by Pascal Belin during his testimony to the French National Assembly.

In three fourths of the cases, the house raids allowed us to check up on people who had been signaled to us as having a troubling change in behavior or talked about supporting Daesh: we could see, at these people’s own homes, if these signs were just associated with a general unhappiness, or if it was something more serious. (Raimbourg and Poisson, 2016: 121).

This type of verification is another instance of an enclosure of borders coming from above, when the tactic of surveillance from below is insufficient. The state of emergency is the admittance of an increased threat level that requires a reconfiguration of data doubles and dividualities, to more precisely surveil certain populations. Where Bauman and Lyon urge us to think of a liquid modernity with data flows and networks as sources of power, the house raid is a reminder of the state’s desire to monitor and appropriate them.

What is at play here is not only the question of how to coherently make sense of the massive amounts of data that get plugged into these lists-plus-algorithms, but how this process of mining, visualization, and analysis changes the nature of the subjects that the data represent. The algorithm calculates correlations from the list that is definite, ordered, and yet endlessly alterable; this combination of protocols is what Johns (2016: 128) calls hybridization.To go further into algorithmic decision making, Diakopoulos (2016) divides it into four different parts: prioritizing, classification, association, and filtering. These tasks are shared by both list and algorithm in a way that makes transparency and regulation difficult. In theory, it is the judicial system of liberal democracies that is supposed to defend the civil liberties of individuals against unlawful search and identification from these algorithms. In many cases this means restricting these algorithms from working at their full potential. Only certain information about subjects can be used at a certain time or only with permission from a judge. In France, along with many other countries, the collection and sorting of data for commercial and judicial purposes is common place. The state of emergency allows for these practices of information gathering by the police to take place without the normal judicial restraint.

Deferred decision making from state to algorithm is elaborated by Dan McQuillan in “Algorithmic states of exception.” Drawing from Agamben’s notion of homo sacer, McQuillan urges us “to look at the nature of the material-political apparatus that connects data to decision-making and governance” (p. 565). For Agamben’s thesis that Western law has placed nation-states in a permanent state of exception, there is perhaps no better example than France’s ever-extending state of emergency. Ontologically this state of exception puts the state in a position of “being-outside and yet belonging” which McQuillan correlates to the nature of algorithms in governance. The tension of power from above versus below, and inside versus outside, are brought up once again. Since the sovereign nation-state can both make laws and exempt itself from them, it is both outside the purview of its own laws, while being fundamentally bound to their creation. As terrorism threatens to undermine the foundations of the western nation-state, the state must reciprocally circumvent its own legal system to defend itself.

For McQuillan, the fear is that “the new algorithmic apparatus can be characterized by the production of states of exception” (p. 578) indicating not just a lack of legal protection, but a fundamental inability of law to deal with these issues since they are themselves legal. This means that the somewhat autonomous nature of these algorithmic apparatuses has the potential to be as ungovernable as the state itself. Expanding on this definition of algorithmic states of exception, the example of i2AN highlights ways that algorithms can alter and exacerbate states of exception that are not their own. It is not that i2AN has created its own state of exception, but rather that it has been embedded into that of the state, with the same extrajudicial predicament as the original homo sacer in Ancient Greece. i2AN creates the visual and geographic boundaries for a state that treats its subjects as data-mining opportunities. Above, it was the state itself that was treated as “being-outside yet belonging,” but it is through this relation that Agamben points to the state’s ability to use measures such as a state of emergency to render its own subjects as also “being-outside yet belonging.” Being the subject of extrajudicial house raids and arrests situates citizens belonging to the state at the same time as it removes them from the regular judicial system, the being-outside the law. The ability to split the subject in this way is an example of algorithmic identity dividing the individual into multiple dividual parts. Algorithmic identity is blurred at the level of the dividual, being outside of specific categories, yet each subject has the potential to be the homo sacer, or the terrorist. It is in this sense that the police raid and house arrest, as tactic of the state of emergency, creates the homo sacer. The criminality is undetermined at first, with the raid serving as an identity adjustment period through data extraction and analysis. For some this adjustment comes quickly, but for others, this liminal period of house arrest can last over a year (Pascual, 2016b).

Conclusion

The role that algorithms play in policing during the state of emergency in France is complex and subject to constant change. From a theoretical standpoint, the operations of i2AN’s lists-plus-algorithm and its visual component challenge ontological conceptions of individual state subjects. I have attempted to show how the creation of dividual state subjects and data doubles, through algorithms can be used for hard state power, or discipline, as well as soft biopolitics or surveillance. Not only are the data-driven police raids a clear form of hard, disciplinary power, but they also expose the basis through which posthegemonic soft power regimes of control are created. Just as the carrot rarely works without the threat of the stick, it is also the stick that makes one interested in the carrot in the first place.

This analysis of i2AN is only a glimpse of the different roles played by algorithm and the state. I argue that the data sorting and mining capabilities of i2AN allow data collection to be a feasible means of policing, which has taken the form of raids and house arrests during the state of emergency. The form of data extraction reflects the technical protocol of i2AN, just as the state of emergency reflects the administrative protocol of the liberal nation-state. A task for further research is an in-depth look at how police analysts directly interact with i2AN and what effect that has had on police operations. Considering the ways in which social conditions influence relationships between decision making and technology is incredibly valuable and can add much to the theoretical grounding presented in this paper.

Raids and house arrests provide the French police with a massive amount of information about a select population that they would not have without the legal exemptions of the state of emergency. With a justice system that is both racially and religiously discriminatory, the data extraction taking place exacerbates existing inequalities and has the potential to do so even after the state of emergency comes to an end. Theories of postpanopticon surveillance or posthegemonic power get us part of the way in understanding how control is mediated through algorithms but they must not neglect theories of hegemony that came before them. To develop a framework for how algorithms influence power, surveillance, and identity, we must continue to refine and develop theory that reflects contemporary examples of policing such as France’s state of emergency.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Agamben

(1995) Homo Sacer: Sovereign Power and Bare Life, Stanford, CA: Stanford University Press.

Alison P (2014) ‘Datafication';, Transparency, and Good Governance of the Data City. In: Digital Enlightenment Forum Yearbook, IOS Press, pp. 215–224.

Amoore

(2009) Algorithmic war: Everyday geographies of the war on terror. Antipode 41(1): 49–69.

Azzam

Evergreen

Germuth

et al. (2013) Data visualization and evaluation. New Directions for Evaluation. 2013(139): 7–32.

Bauman Z and Lyon D (2013) Liquid surveillance: a conversation. Cambridge, UK; Malden, MA: Polity Press.

boyd

Crawford

(2012) Critical questions for big data. Information, Communication, and Society 15(5): 662–679.

Benbouzid B (2015) From situational crime prevention to predictive policing. Champ pénal/Penal field (Vol. XII). Available from: https://champpenal.revues.org/9066 (accessed 26 September 2016).

Brakel V and Rosamunde (2016) Pre-emptive Big Data Surveillance and its (Dis)Empowering Consequences: The Case of Predictive Policing. SSRN Scholarly Paper. Rochester, NY: Social Science Research Network. Available at: https://papers.ssrn.com/abstract=2772469 (accessed 15 June 2017).

Cheney-Lippold

(2011) A new algorithmic identity soft biopolitics and the modulation of control. Theory, Culture and Society 28(6): 164–181.

10.

CNIL (2014) ANACRIM: Logiciels de rapprochement judiciaire à des fins d’analyse criminelle. Available at: https://www.cnil.fr/fr/anacrim-logiciels-de-rapprochement-judiciaire-des-fins-danalyse-criminelle (accessed 14 March 2016).

11.

CNIL (2015) 36th activity report. Available at: https://www.cnil.fr/sites/default/files/atoms/files/cnil_rapport_2015_gb.pdf (accessed 3 January 2017).

12.

Corrieu

J-M

Sajhau

(2014) Sécurité et protection des citoyens: Accroître la prévention et l’action grâce au numérique. Revue de la Gendarmerie 249: 45–49.

13.

Diakopoulos

(2016) Accountability in algorithmic decision making. Communications of the ACM 59(2): 56–62.

14.

Galloway

(2004) Protocol: How Control Exists After Decentralization, Cambridge: MIT Press.

15.

Hall

Heath

Coles-Kemp

(2015) Critical visualization: A case for rethinking how we visualize risk and security. Journal of Cybersecurity 1(1): 93–108.

16.

Human Rights Watch (2016a) France: Abuses under state of emergency. Human Rights Watch. Available at: https://www.hrw.org/news/2016/02/03/france-abuses-under-state-emergency (accessed 29 February 2016).

17.

Human Rights Watch (2016b) France: Prolonged emergency state threatens rights. Human Rights Watch. Available at: https://www.hrw.org/news/2016/07/22/france-prolonged-emergency-state-threatens-rights (accessed 12 November 2016).

18.

IBM – Data analysis – i2 Analyst’s Notebook (2016) CT253. Available at: http://www-03.ibm.com/software/products/en/analysts-notebook (accessed 14 March 2016).

19.

INED (2010) Trajectoires et Origins: Enquête sur la Diversité des Populations en France. Institute National D’Études Démographiques, Paris, France.

20.

Jacquin J-B (2016) Jean-Marc Sauvé : “L’état d’urgence ne peut être renouvelé indéfiniment.” Le Monde.fr, 18 November. Available at: http://www.lemonde.fr/police-justice/article/2016/11/18/jean-marc-sauve-l-etat-d-urgence-est-un-etat-de-crise-qui-ne-peut-etre-renouvele-indefiniment_5033308_1653578.html (accessed 2 April 2017).

21.

Johns F (2016) Global governance through the pairing of list and algorithm. Environment and Planning D: Society and Space 34(1): 126–149.

22.

Just N and Latzer M (2016) Governance by algorithms: reality construction by algorithmic selection on the Internet. Media, Culture & Society 0163443716643157.

23.

Kantardzic M (2011) Data mining: Concepts, Models, Methods, and Algorithms. 2nd ed. Piscataway, NJ: IEEE Press; Hoboken, NJ: John Wiley. Available at: http://site.ebrary.com/id/10494656 (accessed 22 September 2016).

24.

Kantorowicz E (1957). The King’s Two Bodies: A Study in Mediaeval Political Theology. Princeton, NJ: Princeton University Press.

25.

Kennedy

Moss

Birchall

et al. (2015) Balancing the potential and problems of digital methods through action research: Methodological reflections. Information, Communication and Society 18(2): 172–186.

26.

Khosrokhavar

(2013) Radicalization in prison: The French case. Politics, Religion and Ideology 14(2): 284–306.

27.

Kitchin

(2017) Thinking critically about and researching algorithms. Information, Communication and Society 20(1): 14–29.

28.

Kutnowski

(2017) The ethical dangers and merits of predictive policing. Journal of Community Safety and Well-Being 2(1): 13–17.

29.

Lash

(2007) Power after hegemony: Cultural studies in mutation? Theory, Culture and Society 24(3): 55–78.

30.

Lindelauf R, Borm P and Hamers H (2011) Understanding terrorist network topologies and their resilience against disruption. In: Wiil UK (ed) Counterterrorism and Open Source Intelligence. Lecture Notes in Social Networks. Vienna: Springer, pp.61–72. Available at: http://link.springer.com.offcampus.lib.washington.edu/chapter/10.1007/978-3-7091-0388-3_5 (accessed 27 September 2016).

31.

Lyon

Bauman

(2013) Liquid Surveillance: A Conversation, Cambridge; Malden, MA: Polity Press.

32.

Mager A (2012) Algorithmic Ideology. Information, Communication & Society 15(5): 769–787.

33.

McQuillan

(2015) Algorithmic states of exception. European Journal of Cultural Studies 18(4–5): 564–576.

34.

Nagy P and Neff G (2015) Imagined Affordance: Reconstructing a Keyword for Communication Theory. Social Media + Society 1(2): 2056305115603385.

35.

Pascual J (2016a) Etat d’urgence: Hausse de 50 % des assignations à résidence en trois mois. Le Monde.fr, 19 September. Available at: http://www.lemonde.fr/police-justice/article/2016/09/19/etat-d-urgence-hausse-de-50-des-assignations-a-residence-en-trois-mois_4999975_1653578.html (accessed 20 September 2016).

36.

Pascual J (2016b) Assignés à résidence depuis un an. Le Monde.fr, 13 November. Available at: http://www.lemonde.fr/attaques-a-paris/article/2016/11/13/assignes-a-residence-depuis-un-an_5030334_4809495.html (accessed 7 April 2017).

37.

Polloni C (2015) Police prédictive: la tentation de “dire quel sera le crime de demain” – Rue89 – L’Obs. Rue89. Available at: http://rue89.nouvelobs.com/2015/05/27/police-predictive-tentation-dire-quel-sera-crime-demain-259384 (accessed 14 March 2016).

38.

Raimbourg and Poisson (2016) Rapport assemblée nationale N. 3784. Available at: http://www.assemblee-nationale.fr/14/pdf/rapports/r3784.pdf (accessed 21 October 2016).

39.

Schmitt C (2005) Political Theology: Four Chapters on the Concept of Sovereignty (University of Chicago Press ed). Chicago, IL: University of Chicago Press.

40.

Seidler

Adderley

(2013) Criminal network analysis inside law enforcement agencies: A data-mining system approach under the national intelligence model. International Journal of Police Science and Management 15(4): 323–337.