Keywords

Are we made up entirely and without residue of the data that define us, or is there a disjunction between our data shadows and our embodied selves? How do we come to recognize ourselves, our selves, in the pronouns that interpellate us online, and what is it exactly that we recognize? What does it mean to occupy the semantic and positional space of the pronoun ‘you’? And is there a continuity or a discontinuity between the systems of surveillance and data aggregation that address us and the systems that refer to us? The markers of identity generated by such systems work by both individuating and classifying us; this paper seeks to think about the range of possible relations between that generality and that particularity.

Last year my (very latent) Facebook account was hacked and a figure bearing my name and my visual icon joined me up to a number of plausible sites, and a couple of somewhat less plausible ones, on which it then made pitches announcing special deals on gambling sites and promoting casinos. I’ve now withdrawn completely from Facebook so I can’t tell whether this doppelgänger still represents me there, although given its apparent autonomy it may well have taken on a vampiric life of its own.

That’s a first example of identity theft, one in which it’s still quite clear which is the real and which the false John Frow: my doppelgänger ‘represents’ me and for some people may ‘be’ me, but for people who know me it will, I hope, be obvious that its words in support of gambling are unlikely to be mine. A second, equally banal example: a couple of years ago my wallet was stolen in Barcelona and my bank later traced the trail of my credit card up toll roads to Lyon, where it stopped, the card having been cancelled. My bank believed me, and someone’s representation of themselves as me was labelled a misrepresentation. But consider a third possibility, that of a complete and successful theft of identity. Koopman (2019: 4-5) summarizes the resulting ‘permanent and irreversible erasure of the entirety of [the victim’s] personal information and therefore their entire informational identity’ as follows:

No driver’s license, no passport, no bank account number, no credit report, no college transcripts, no employment contract, no medical insurance card, no health records, and, at the bottom of them all, no registered certificate of birth. The scenario is chilling: everyone around you well attached to their data while you are dataless, informationless, and as a result truly helpless. What would you make of yourself? What could others make of you? What would the bureaucracy be able to do when you petition it with your plight, given the fact that no bureaucracy can address a subject as other than their information? … They would have no way of addressing you from one day to the next, of recording you in their databases, of numbering or naming you, and so no way at all to deal with you on anything approaching a consistent basis. You could not even receive special support through special court orders because, completely unrepresentable as information, you would have no way of being registered into a court, for that would require rendering you into the data from which you have been detached.

The subsumption of personhood into documentary information that is evidenced in the case of identity theft has its beginnings with the governmental systems of early modernity, when a proliferating apparatus of writing begins to integrate verified identities in cadastral and demographic records and to generate both a systematic scrutiny and the rights and recognition accorded by the state to those verified identities. The paradox at the heart of this process is that, while an identity document purports to be ‘a record of uniqueness’, it must at the same time be ‘an element in a classifying series’ (Caplan and Torpey 2001: 8). And the converse of that dichotomy is the further paradox that these mechanisms of construction, verification, scrutiny and control are at the same time lived by me as the confirmation of my uniqueness, my being as a person. Personalization in this sense is a function of generality, ‘a mode of individuation in which entities are precisely specified by way of recursive inclusion in types or classes’ (Lury and Day 2019: 18). My documents identify me not only as ‘the’ person that I am but as ‘a’ person, one of those persons, one of those entities that are classed as persons; and as a member of all those other classes of human being that count and that make me count for this act of classing.

Individual identity is legally underpinned by the documents of state, but, as Ruppert (2011: 218) writes, ‘people are not governed in relation to their individuality, but as members of populations. The embodied individual is of interest to governments in so far as the individual can be identified, categorized and recognized as a member of a population. This is the general problematic of governing, which is to know the nature and then govern and regulate the forces of the collective body’. Populations are known and made legible—indeed, are constructed as populations—by means of devices and practices of identification of individuals as categorized subjects, ‘an element in a classifying series’. The basic identifiers for governmental scrutiny are the biographical data that register a name, a set of kinship relations, a gender, an ethnicity, an address; in addition, biometric data record certain unique physical attributes or habits of bodies (facial structure, fingerprints, iris geometry, height, gait). The biographical data move outwards to the set of social relations within which they are meaningful; the biometric data move centripetally to identify a singular body differentiated from all other bodies. What is of interest to the state may be either the general category (Which classes of people have need of particular social services? What is the desirable distribution of resources across a particular population?) or hypothetical individuals apprehended as members of a category (Which particular neo-Nazis or radicalized Muslims are likely to espouse violence? Which former chemistry teachers are likely to be cooking ice?). In each case the state will aim to build up a coherent picture through the use of statistical evidence, or through a cumulative record of transactions between an individual and the various branches of the state or commercial institutions (driving license, criminal record, medical records, property holdings, and so on). It will identify patterns of group circumstance or patterns of individual conduct, and it will seek to amass and correlate patterns of information across whatever databases are at its disposal.

Those databases are now for the most part digital, and thus susceptible to algorithmic operations of search and analysis which build ‘what is uncertain and unknown into the identity calculation itself’ (Amoore 2008: 25). In the case of national security systems, such speculative uses of data, monitoring whole populations in quest of individual anomalies, are intended to generate as-yet unsuspected patterns rather than to find evidence to support known possibilities (Raley 2014: 123). They bring into being what Amoore (2011: 27) calls a form of data derivative, meaning ‘a specific form of abstraction that distinctively correlates more conventional state collection of data with emergent and unfolding futures’; the data derivative comes into being from ‘an amalgam of disaggregated data’, sorted by way of recursively refined algorithmic association rules and given visual form as ‘risk map, score or colour-coded flag’ (Amoore 2011: 27). It might, for example, be derived from an associative matrix connecting a flight destination, fare payment by a third party at short notice, a dietary choice, and a history of attendance at a religious institution. The knowledge formed here is ‘actuarial’ (Andrejevic 2012: 95), converting ‘the databased residue of daily life’ (Amoore 2009: 52) into an actionable probability. A risk value is assigned to an individual, and this pre-emptive identification allows the analyst either to read outwards to the ‘nodes of connections between data’ (Bauman et al. 2014: 125) (i.e., the network of a suspect’s personal connections) or to act to avert an immediate threat. The data derivative is ‘indifferent to the contingent biographies that actually make up the underlying data’; it ‘is not centred on who we are, nor even on what our data says about us, but on what can be imagined and inferred about who we might be – on our very proclivities and potentialities’ (Amoore 2011: 28). To my singular body it attaches a virtual state, my data shadow, which then defines me.

Edward Snowden’s revelations about the US National Security Agency identified a range of forms of Internet surveillance, among the most prominent of which are, on the one hand, the PRISM programme, which, through its XKEYSCORE software, allows analysts to read from the servers of Internet service providers every keystroke of every person’s online activity; and, on the other, the direct upstream harvesting of data from private-sector Internet infrastructure—‘the switches and routers that shunt Internet traffic worldwide, via the satellites in orbit and the high-capacity fiber-optic cables that run under the ocean’ (Snowden 2019: 122). This is genuinely ubiquitous surveillance. Yet note that it’s difficult to draw a clean line between such state surveillance and the data-harvesting capabilities of corporate information harvesters and brokers, with some of which—Snowden mentions ‘Microsoft, Yahoo!, Google, Facebook, Paltalk, YouTube, Skype, AOL, and Apple’ (Snowden 2019: 122)—the NSA has a closely symbiotic connection.

A number of different kinds of commercial corporation harvest data on Internet users: retail corporations, search engine operators, social media companies, data brokers, data analytics providers, and so on. Apart from data-broking and data-analysis firms, which sell information directly to their clients, the business model on which these corporations run is personalized advertising based in the interactive capacity of digital media, with harvested data either used directly, in the case of retail firms, or auctioned off to advertisers in the case of search engines and social media companies. The pioneers in the field were probably the giant consumer credit bureaus like Experian, Equifax, and TransUnion (cf. Lauer 2017), and retail corporations like Walmart and Amazon in the US and Tesco and Marks and Spencer in the UK, which hold and monitor massive amounts of data on the contact details, purchasing history, and lifestyle preferences of their customers, along with all the ancillary information that flows from it (financial status, sexuality, mobility, physical fitness, cultural tastes, dietary and pharmaceutical choices, and so on). But the field of personalized advertising is now dominated by the two biggest players, Google and Facebook, with the other three technology giants, Microsoft, Apple, and Amazon, integrated into the field in somewhat different ways.

Let me take Google as the key example here. In Shoshana Zuboff’s comprehensive account, the dot.com crisis of 2000 provided the occasion for Google’s founders to abandon their initial strong opposition to advertising. Two conditions allowed the company to transform online advertising, particularly after its purchase of DoubleClick in 2008 (Cheney-Lippold 2017: 20) and as it came to realize that it was indexing not only, at an aggregate, topological level, the entire network but also a second world, that of individual users, which it then overlaid on the first in order ‘to deliver relevant search results to the users, and to deliver relevant users to advertisers’ (Stalder 2010: np). The first condition it satisfied was that its computational tools and infrastructure enabled it to create user profile information from analysis of search patterns, keystroke by keystroke, and to match advertisements to the user’s interests as they are deduced from these traces of online behaviour. The second condition was its development of an options-based pay-per-click revenue system which it carried to the contextual advertising system it established for its search engine and for Gmail, such that when an advertiser bids for a keyword the system tracks those of the user’s searches that are contextually relevant for it (e.g., a search for online clothing), matches the user to a product range, serves the advertisement, and, if the visitor clicks on it, invoices the advertiser for the price negotiated for that particular user profile—all in real time (Turow and Draper 2012: 135).

Unlike older business models targeted only to keywords or content, Google was thus able to tailor advertisements to the interests, the social connections, and the physical and online locations of a particular user, and it did so by collecting stores of what Zuboff calls ‘behavioral surplus’, which embrace ‘everything in the online milieu: searches, e-mails, texts, photos, songs, messages, videos, locations, communication patterns, attitudes, preferences, interests, faces, emotions, illnesses, social networks, purchases, and so on’ (Zuboff 2019: 128). Hence the expansion of Google, Facebook, Amazon, and Apple into the Internet of things: a world of information-gathering devices, each of which is ‘a slightly different configuration of hardware, software, algorithms, sensors, and connectivity designed to mimic a car, shirt, cell phone, book, video, robot, chip, drone, camera, cornea, tree, television, watch, nanobot, intestinal flora, or any online service’ (Zuboff 2019: 129).

We all have a sense of the sheer scale of the enterprise: in 2020 there were 6.9 billion Google searches a day and the company generated $116 billion, 97% of its total revenue, from advertising sales.Footnote 1 Its subsidiary company YouTube had 2 billion monthly users in February 2019, Chrome had 62% of the browser market globally, Android had 2 billion users in mid-2017, and Google Maps and Gmail each had well over a billion users. The services that Google and its parent company Alphabet offer are multifarious, but their aim is singular: to collect behavioural data about individuals that can be monetized as advertising revenue. As Zuboff puts it: ‘With click-through rates as the measure of relevance accomplished, behavioral surplus was institutionalized as the cornerstone of a new kind of commerce that depended upon online surveillance at scale’ (Zuboff 2019: 83).

The collection processes employed by Google and other commercial entities are structurally homologous with those of state intelligence services. ‘Online surveillance at scale’ harvests information that had never previously been captured at scale—‘about people’s time-space paths through the course of the day, the details of when and where they chat with friends, even the random queries that drift through their minds (to the extent that these are transformed into Google searches)’ (Andrejevic 2012: 93)—and it does so by making use of algorithmic procedures, such as mathematical association rules, which move between the commercial sphere where they were initially developed and that of national security apparatuses (Amoore 2008: 26).

But the ubiquity of surveillance doesn’t mean that we live in a world of totalized panoptic control. Mark Andrejevic has suggested the alternative metaphor of a world made up of a series of distinct but sometimes overlapping digital enclosures, meaning the coverage range created by the interactive and data storage capabilities of any digital surveillance technology—a world characterized, then, ‘by a proliferation of different monitoring networks with varying capabilities for information capture under the control of different entities’ (Andrejevic 2012: 93). Under certain conditions (e.g., a totalitarian government such as that of China with a tight hand on the public domain) data from a number of different enclosures might be aggregated; and security agencies such as the NSA do in practice make use of commercially gathered data, either by stealing it or by exploiting software vulnerabilities or merely by requesting access to it. This is an area in which tech companies in the West are, or want to be seen to be, pushing back, but probably the most we can say about this is that the balance between privacy and omnivorous data collection is precarious and in a state of considerable flux. Further, the coexistence of digital enclosures within an overarching assemblage means that information collected for one purpose—the mapping of the built environment by Google Earth, for example, or the monitoring of the flow of water or electricity or traffic—might be migrated ‘across a range of other, sometimes unanticipated functions’ (Andrejevic 2012: 93; cf. Lyon 2014: 5-6, 8). The trade goes both ways, with technologies and software developed for military or security purposes finding their way into the surveillance activities of business—or, more precisely perhaps, with an increasing lack of differentiation between these spheres.

The identifying and personalizing data that we yield through digital interactions may be given voluntarily or involuntarily. Involuntary generation of data takes place by means of cookies or other tracking devices which introduce memory, or statefulness, into a stateless system such as the basic Internet protocols (Sipior et al. 2011: 3), and they are thus deictically charged: localized in time and space to a particular Internet subject. Alternatively, the involuntary generation of data takes place by way of the network of automated sensors (facial recognition systems, RFID tags, location tracking, the plethora of sensors on any smartphone, and so on) that cover our world, directly registering traces of our bodily presence in space and time (Kang and Cuff 2005: 94). In some instances we may give permission for our data to be collected; but since the alternative is not to use the interface at all, and since privacy agreements tend to be unreadably lengthy and legalistic, the permission can only technically be said to be voluntary: it is in effect a function of ‘a regime of compulsory self-disclosure’ (Andrejevic and Gates 2014: 191).

The prevalence, on the other hand, of voluntary disclosure of personal data seems, given its value for commercial exploitation or for scrutiny by the state, to require some explanation. ‘We need to understand’, writes Koopman (2019: viii), ‘why we do not question, and why we even eagerly participate in, projects of government data harvesting and corporate data collection, and a raft of programs designed to store and analyze every flake of data dandruff we cannot help but leave behind in nearly everything we do’. The explanation—beyond the sheer usefulness of centralized medical or administrative databases—surely has to do with what it is that digital interaction offers: a mode of sociality, the affirmation of a sovereign self, pleasure in the construction and display of a public self, and the promise of ‘genuine individuation’, such that ‘disidentification will no longer be necessary as a way of maintaining individuality in a scene of falsely personalized address’ (Cohen 2019: 174). In a context of unfathomably complex communications, the ‘practical consciousness’ of digital subjects works as though communication were unproblematically immediate and intimate and is built on an imaginary of ‘sovereign control, a sovereignty of self-hood’ manifested through willing personal disclosure (Bauman et al. 2014: 138). Such disclosure is part of ‘a sharing practice involving mutuality and reciprocity rather than a one-way flow of information’ (Raley 2014: 133); the gift of free labour to websites forms a community, a set of social relations, a commons. Constructing a profile and engaging in Facebook’s ‘Like’ economy, for example, ‘transforms users’ affective, positive, spontaneous responses to web content into connections between users and web objects and quanta of numbers on the Like counter’ (Gerlitz and Helmond 2013: 1358).

Digital disclosure, in generating value, is formative of social relations; this is that double movement by which the Internet takes the form of being ‘always and simultaneously a gift economy and an advanced capitalist economy’ (Terranova 2000: 51). The social media profile and timeline and accumulated posts and the acts of friending and liking and rating make up ‘a presentation of persons’ (Koopman 2019: 7): a composition of the self that persists across time and across digital space. Made entirely out of data—out of stories, images, affects, arguments, observations—it corresponds to that other presentation of persons that is formed without my willing it from the algorithmic compilation and analysis of tracked online data and that may convert it into value. My profile, freely offered to the world, is one of the dual sources of the profiling, the ‘reputation’, that sells me to advertisers or that defines me for the state.Footnote 2

Taken together, the regimes of voluntary and involuntary disclosure thus construct what Goriunova (2019: 126) calls the digital subject, a concept that includes ‘a subject of a data profile or of a Facebook stream, a history of browsing or search engine queries, mobile phone positioning records, bank transactions, sensor data, facial recognition data, biometric movement recognition data, or email inboxes, among other things. The digital subject thus moves between captured, unique, and persistent biological characteristics and premeditated forms of symbolic expression, judicially inferred subjects of actions, and performed identities’.

One important way in which commercial differs from state surveillance is that in most instances the state works in the third person—it talks to itself about its subjects—whereas commercial surveillance converts its descriptive data into second-person address. Online advertising forms a vocative self: a self substantiated by the nameless and invisible voice that addresses me. Here’s how it speaks to me—a few sentences taken at random from websites I’ve recently visited:Verse

Verse Discover the 7 steps to harness your ambition and rescue your dormant business Lego Marvel Avengers: Create the Ultimate Quinjet. Shop Now Try Prime Video Free Go Now, Go There, Go Anywhere

In each of these examples the pronoun ‘you’ is silently embedded in an imperative that works ambiguously as both an order and an invitation: an anonymous speaker addresses me as a subject who is invited or ordered to attend to an injunction. The speech lacks authority, since I don’t know who is speaking, and although these sentences are in the imperative mode they have no power to compel other than by awakening my interest—a remote chance, since I don’t run a business, dormant or otherwise, or want to know what a Quinjet is or to accept an offer that I understand is meant to hook me into a subscription I don’t need, or to buy a new car. And this ‘you’ that is addressed to me is at once specific and indistinct, neither singular nor plural but somehow both at once, a generalized addressee who is nevertheless me alone, the sole receiver of these words.

When the crooner asks, ‘Who … stole my heart away?’ and answers, ‘No one but you’, we don’t know who this ‘you’ is, other than that he or she has the attribute of having stolen the singer’s heart away; we don’t know which person might fill this empty slot. We do know that it’s not us, the persons listening to the song and in some, perhaps indirect sense being addressed by it: this ‘you’ passes to the side of us. But the identity of the ‘you’ can be subsequently specified, either within the song—by being given a name, for example—or by the adducing of external information, perhaps about the singer’s or songwriter’s biography. In conversation between two people the contextual specification of the pronominal shifter is in the first instance total: ‘you’ is the other partner to the dialogue. Given the citational capacity of all speech, however—our tendency to weave the speech of others into our own—the specification may be more complex. When more than two people are present in a conversation the reference of the second-person pronoun may require disambiguation; in this case the referent of the pronoun will be the most likely or most salient candidate in the contextual field. In the case of written texts, the reference of second-person pronouns is always in a sense an act of simulation, a pretence that the openness of reference has always already been filled, that the nameless reader was always the one intended to receive this word. Seeming to single me out, personalized online address has the off-key familiarity of intimate words spoken by a stranger.

The uncertainty of deictic reference is at the heart of the process of contextual specification that we know as the interpellation effect: a process of conversion of a non-specific into a specific but uncertain designation of the pronoun ‘you’ (Chun 2017: 3 and passim). In Charles Fillmore’s example, you are a young woman who has been wolf-whistled in the street. You want to reprimand the whistler but it’s not clear whether you are the intended target (it might have been some other young woman in the street), and ‘to turn around and scowl is to acknowledge that you believe the message was intended for you, and that may be taken as presumptuous’ (Fillmore 1997: 59). The uncertainty of reference applies both to the person who whistled and to the young woman, and in both cases, we know the kind of person who fits the description, but not the particular instances that would fill those generic slots. In Althusser’s version of this, a policeman calls out ‘Hey, you there’ in the street and I turn around, assuming he means me (Althusser 2001: 118). Both cases represent a situation of ambiguity in which I respond to the message by appropriating it to myself: I fill it with my desire to be the one hailed or whistled at, (mis)recognizing myself in the pronoun uttered by the other as though it were personally addressed to me. Although Althusser’s account is problematic, based as it is in a model of subjection to and by a sovereign power enacted through my response to the Absolute Subject—at once the State, the Father, and God—it nevertheless gets nicely at the mechanisms of imaginary singularization and personalization through which recommendation systems and targeted online advertising operate.

In algorithmic recommendation systems such as those used on music streaming platforms, the particularized musical identity of the addressee is constructed from the continuous collection and aggregation of contextual data points. Content filtering systems like Pandora organize music by analysis of its structural features and continuously revise their weightings as they match them with feedback from listeners, without regard to genre labels, cultural mapping, or demographic position. A collaborative filtering system like Spotify, by way of its The Echo Nest subsidiary, takes this a step further by combining the outcomes generated by acoustic analysis software with ‘semantic analysis of online conversations about music that take place every day, all over the world—millions of blog posts, music reviews, tweets and social media discussions’ (Prey 2018: 1090-1). Overlaying on these analyses a preference analytics that capturesFootnote 3 and records in real time a listener’s musical behaviour and preferences,Footnote 4 Spotify treats the cultural mapping of music as a further insight, differentiating music that is structurally similar in accordance with the highly differentiated taste cultures of the digital world. The ‘paradigmatic claim’ of such algorithms is ‘to specify the individual in the complex conjugated personalized address: ‘People like you like things like this’ (Lury and Day 2019: 24). What these and other recommendations systems, like those of Netflix or Amazon, have in common is their generation of a ‘you’ that is not based on fixed markers of identity, either demographic (class, gender, age, ethnicity, and so on) or generic (jazz fan, Christian rock fan), the properties of which are presumed to be known in advance. Rather, they generate personalized recommendations from categories that emerge from a process of recursive revision.

If in a formal sense the profiles and ‘reputations’ constructed for us and addressed to us by advertising and recommendation engines have no content other than the acts of recognition or misrecognition—of imaginary personhood—that transiently fill them, it is nevertheless the case that these shifters are constantly being specified contextually through acts of rigid designation that seek to tie them to a name and a legally established identity (and that are just as constantly resisted by acts of counter-naming or heteronymy or masking). In practice this means the construction of data-shaped personal selves—data shadows or data doubles—across online databases, where information freely offered on social media as a referential truth (self-expression, life-writing, autobiographical timelines, and so on, however fictive these truths might be), or volunteered to state or corporate databases, or captured from phone usage and location tracking or from facial recognition systems, is fastened to a persistent identity by a kind of point de capiton pinning my online transactions and pathways to the official documents that are the baseline of my composite existence. The development of cookies and of even more persistent tracking IDs such as Flash cookies or web beacons has been at the heart of the ability of the state and corporations to silently monitor my activity in this more or less integrated way across convergent sites and devices. Although for many purposes on the Web it doesn’t matter whether anyone knows you’re a dog, and much information harvesting and analysis is concerned not with ‘the personal identity of the embodied individual but rather the actuarial or categorical profile of the collective’ (Hier 2003, cited in Cheney-Lippold 2011: 177), these tracking and fastening devices nevertheless allow in principle for the attachment of a corpus of data to my name, and of my name, together with the descriptive attributes it brings with it, to my embodied self. Stalder (2010: np) distinguishes three types of profiles, which together create a comprehensive profile of each user: ‘First, users are tracked as “knowledge beings” and, in a second step, as “social beings,” exploiting their real identities, contacts, and interaction with those contacts. A final data set captures the users as physical beings in real space’. We might think by analogy of the legal identification of criminal culpability by means of documentary evidence such as eyewitness testimony or forensic analysis, where the indexical tie to the body of the culprit is established by a witness’s sworn statement that they have seen this person’s body or by the traces the body leaves at the scene of the crime: here too there is nothing but data, nothing that we can call simply a truth; but the verdict, the truth-saying, of culpability is given by the accumulation of those traces of information.

The person addressed by the second-person pronoun or implied by deictic markers or captured by stateful trackers is not a substantial particular, a self-identical presence, but the occupant of a semantic place; the space and time for which he or she serves as a reference point are constructed in dense networks of metaphor, and the body that orients that person in space and time is imagined and positioned through these networks. Since the place of the shifter may be occupied by anyone who is addressed by it, the ‘you’ is structurally riven, positional and alienable yet embodied, a reference point in time and space and yet movable from discursive point to point, a figure in a statement (Frow 2014: 164). Yet is this shifting ‘you’ not grounded in a material and experiential reality that occupies the pronoun and that we experience as the solidity of a selfhood? In one sense, it quite clearly is. My body can be arrested and thrown in prison; it can be tortured or killed. I can be enticed to spend money that I have earned by virtue of real physical or intellectual labour; I can go into debt and undergo material hardship when I lose my credit rating or my right to work. We could, therefore, posit the self of experience, as William James or Alfred Schutz do, as taking place in a field of deictic reference encircled by the body and mobilized in my face-to-face interactions with others, and then grant that everyday experiential self priority over the more remote modes of selfhood engaged in the worlds of secondary representations. We could, that is to say, posit a necessary gap between my embodied everyday self and my data self, my data double, my data shadow; between the ‘I’ that I live from the inside and the ‘you’ that is directed to me from without or the ‘he’ or ‘she’ that describes me in a database.

Yet thinking about the relation between real and algorithmic personhood in this dichotomous way is, I think, both conceptually and politically unhelpful. It posits, in the first instance, a body that is distinct from the information that shapes it. As Irma van der Ploeg has argued across a series of papers, the increasingly prevalent translation of aspects of our bodies into digital code is not a matter of changed representations, with ‘the thing itself’ remaining the same (van der Ploeg 2012: 178), but a fundamental change at the level of ontology, since ‘there is no clear point where bodily matter first becomes information’ (van der Ploeg 2003: 70). The body through which we apprehend ourselves and others may look like a ground truth, but that body is not a pre-discursive matter. It is information in its substance and its processes—in the DNA that composes it, in its homeostatic regulation of the endocrine, immune, and autonomic nervous systems—and in all of the systems (the regime of state security, the state welfare and schooling apparatus, the insurance industry, the taxation system, medical databases, regimes of visual representation) in which it is inscribed. Likewise, for the person who is and who understands themselves through their body, this is nonetheless a body experienced through a bodily imaginary, the effect of a ‘system of exchange, identification and mimesis’ in which I shape my sense of myself by way of a recognition and incorporation of the bodies of others (Gatens 1996: 31). It is through this imaginary body that my fundamental fantasies about who I am and how I engage with others are shaped.

In the same way, to think in terms of a dichotomy of digital and real personhood is to posit too stark a disjunction between representations of the self and an offline actual self. The digital subject—the vocative self of advertising, the data double of surveillance systems—is not an external representation but the constantly mutating effect of ‘the practices through which one becomes data through interactions with numerous other actors and actants’ (Ruppert 2011: 225). Taking issue with the concept of the data double, Koopman (2019: 170) argues that ‘data has become a crucial part of the very terms by which we can conduct ourselves. We are our data. Therefore we are precisely not doubled by it’. Interaction with data, whether voluntary or involuntary, witting or unwitting, is integral to the actuality of our selfhood. This is a matter of a pragmatic and contingent formation of digital personhood; the algorithmic subject is sustained by the interplay of systems of ubiquitous surveillance but also by the ‘unique combinations of distributed transactional metrics that reveal who they are’ (Ruppert 2012: 124). I become who I am through my engagements with the real and the digital worlds, and the difference between those realms is increasingly tenuous. This data self that I ‘am’, however, is never singular: we can’t speak of digital selfhood as a consistent aggregated identity formed across databases, since different databases measure and construct different realities: a reality of consumer desire, a reality of cultural preferences, a reality of political convictions and actions, a reality of economic capacity, and so on. Against the impetus of the state, and perhaps our own habitual impulse, to reduce identity to a single point of reference, we can at most speak of a plurality of more or less convergent, more or less heterogeneous forms of personhood, none of which is an absolute ground.

These distinctions may look scholastic, but I think there are strong reasons for teasing them out. In the case of my initial example of identity theft I tried to demonstrate the futility of any appeal to a real and embodied self. What counts in practical—that is, in administrative—terms is not the body that I am but the forms of documentation that make me up and the way they fit together. Similarly, resistance to state or corporate surveillance can’t be grounded in appeal to the fundamental and singular reality of my person but only in alternative ways of figuring myself and of challenging or ignoring the specific forms of figuring and naming that construct and address me. ‘Figuring’ here means both calculating and performing the form of the person, and it comprises the acts of recognition that construct me as ‘an element in a classifying series’ and thus as a governable subject. If we are to understand the new modes of personhood of an evolving world of information technologies and self-educating machines, we must understand the complexities of new systems of construction and extraction of value, the extending universe of ubiquitous surveillance, and the changing forms of address that situate me in this universe.

‘We’, I said: ‘We must understand’. But what ‘we’ must understand is the slipperiness of these pronominal shifters and the way they construct communities of understanding which are far from self-evident: who is this ‘we’, who is the ‘me’, and what’s the status of the slide between them that I performed a few sentences ago; who is the ‘you’ I’m addressing in the now of this room, and in the non-time of this writing: you present and absent, ‘you’ singular and ‘you’ plural, and what kind of plurality does that singular become? ‘We’, whoever we are, must above all learn to be distrustful of the communities we invoke and of the ‘you’ that invokes us and with which, in this time of speaking and this non-time of writing, I address and invoke you now.