Nothing to disconnect from? Being singular plural in an age of machine learning

This essay makes the claim that there is nothing to disconnect from in the digital world, and that the logic of machine learning provides the most obvious empirical case for this. Drawing on Jean-Luc Nancy’s concept of Being as always already a being-with, the argument is made that the inescapability of others does not provide the end-point for a gesture towards disconnection but an opportunity to rethink the ethics of dis/connectivity in a more productive way. I situate these claims in the scholarly discussion on digital disconnection and privacy within media studies with the purpose of contributing a critique of a discourse predominantly concerned with framing disconnection as a form of voluntary and empowered form of media refusal.

for disconnection studies to grapple with Nancy's (2000) assertion that 'we have not even begun to discover what it is to be many'. (p. xiv) Building on work that accentuates the networked conditions of privacy and surveillance (Boyd, 2012) as well as the politics and ethics of machine learning algorithms (Amoore, 2020;Bucher, 2018), the argument is made that the logics of datafication and predictive analytics make it difficult to abstain from and opt out from digital platforms and services. Simply deleting one's Facebook profile, or abstaining from using a digital device for a week does not result in disconnection, or less data production, but more data points. While a week of digital detox may give someone some small peace of mind, to an algorithm, however, this form of absence provides important pieces of information. In other words, refusing to connect or temporarily opt out is a form of connection too. Acts of deletion, absence and silence not only provide information about the person abstaining; they also say a great deal about the people tethered to the refusing agent.
Here, a discussion of digital disconnection and machine learning is combined to explore questions of agency, political practice and ethics, drawing on the writings of Nancy (1991Nancy ( , 2000. For Nancy, existence implies a co-existence. Being-in-common is not what happens after individuals come together, but rather constitutes the ontological starting point. If Nancy's notion of Being is always already being-with, then not beingwith is not an option. The question, then, is how to rethink disconnection in light of being-in-common as the starting point, rather than the end-point?

Discourses of digital disconnection
As life is increasingly tethered to screens and digital devices of all kinds, debates abound on the need for more responsible and mindful media use. If much of the research into the Internet has traditionally been concerned with the conditions of connectivity through the lens of online communities and networks, there is more work emerging on the supposed antidote to ubiquitous connectivity. From discussions around slow media movements (Rauch, 2011;Skågeby, 2011), media refusal (Portwood-Stacer, 2013Ribak and Rosenthal, 2015), digital detox (Sutton, 2017;Syvertsen and Enli, 2019) through to disconnective practices such as leaving social media platforms (Brubaker et al., 2016;Karppi, 2018;Light, 2014), digital disconnection is variously framed and theorised an act of resistance and mindful media use.
In this vein, work on digital disconnection follows a long tradition in media studies of examining non-use and media resistance (Kline, 2003;Krcmar, 2009;Mainwaring et al., 2004;Syvertsen, 2017). There are at least two sides to framing non-use in the literature: involuntary and voluntary. The former is reflected in work on digital divides and social inequalities, often framing non-use as a matter of lack, whether this be a lack of access, skills or means (Hesselberth, 2018). The latter view on non-use is representative of the above cited works that tend to frame disconnection as an active choice of resistance and empowerment.
These studies variously conceptualise disconnection as resistance to surveillance (Light, 2014), as lifestyle politics (Portwood-Stacer, 2013), as a path to liberation and control (Wyatt et al., 2002) and as an everyday phenomenon spanning different motivations and modalities, including distrust in media, personal choice, activism and impression management (Kuntsman and Miyake, 2019;Syvertsen, 2017). While research on digital disconnection differs in the terms used to describe the phenomenon, what they have in common is a conviction that disconnection is not just possible but also meaningful and necessary.

Technologically tethered selves
The desire to disconnect may be very real indeed, but it can be difficult, if not impossible, to accomplish and sustain in a technology-based society (Karppi, 2011;Kaun and Treré, 2018;Van Dijck, 2013). There are different dimensions to this. On an everyday level, it can be burdensome to not own a smartphone or to not use Facebook. For example, as one Internet scholar recently remarked on Twitter, never had it felt more pressing to be on Facebook as when trying to find daycare, because all new places were advertised on Facebook mom groups only. More fundamentally, as Hesselberth (2018) notes, discourses of disconnectivity are often very far removed from the lives of those for whom opting out, detox and abstaining from media may not be as easily available as they are for the privileged and resourceful few.
However, there is also a very real technological aspect to dis/connection that may complicate or even make opting out impossible, if by disconnection we mean something akin to not having an account on a social media platform or not leaving a trace in a digital database. Too often, discourses on digital disconnection posit it as an individual's active choice. Yet the operational logics of networked spaces complicate the notion of individual agency in sometimes quite fundamental and unprecedented ways. As Danah Boyd (2012) has argued with regard to privacy in networked spaces, 'any model of privacy that focuses on the control of information will fail' (p. 349). Because 'our data -and with it, our privacy -is increasingly networked' (Boyd, 2012: 348), thinking that individuals are in charge of their own data and information flows seems problematic. Even for the person who knows exactly how to control and manage their privacy settings, there is no guarantee that others will, or know how to. As Marwick and Boyd (2014) write, 'When Taylor McCormick, a student at the University of Texas, joined the campus's Queer Chorus Facebook group, his participation was broadcast to everyone in his network, effectively outing him to his parents'. (p. 1062) Despite having made sure his privacy settings would not allow sharing sensitive information like this, it turned out that the group settings did.
If social media is fraught with flawed privacy settings and leaky information flows (Agostinho and Thylstrup, 2019), the data-driven and algorithmic infrastructure of machine learning complicates disconnection on a different level. In the digital disconnection literature, people are described as avoiding digital devices such as smartphones, limiting their screen times or abstaining from using specific platforms. In addition to framing it as an individual choice, disconnection is often also portrayed as a selective and strategic form of refusal. In many cases, this means resistance to a specific platform (e.g. Facebook) or the choice of temporarily opting out (e.g. detox camps). Yet the underlying networked condition of being algorithmically processed by calculative devices, no matter what platforms or timeframe are being renounced, is rarely part of that same discourse (see Karppi, 2018 for a notable exception).
We are inherently digitally connected, whether we like it or not. We might choose not to use Facebook, but data traces and connections are nonetheless forged via our friends who do. We might take a few online breaks and detox once in a while, but then we return. We might use ad blockers, but consent to cookies all the time. We may not use Google search all that much, but we use our Google accounts to log in to every website asking for identification. The obvious point is that for all the conscious, critical and careful considerations, data about us are used every step of the way to generate more sustained profiles.
More fundamentally, algorithms stipulate how the world appears and takes shape. As Amoore (2020) describes in her latest book Cloud Ethics, algorithms decide 'what matters in the world, what or who can be recognised, what can be protested, and which claims can be brought' (p. 10). In a compelling example of what this means in practice, Amoore recounts the tragic death of Freddy Grey at the hand of the Baltimore police in April 2015 and the deep learning techniques used by the police to control the public protests that subsequently erupted. During the days of the protest, 'terabytes of images, video, audio, text, and biometric and geospatial data from the protests of the people of Baltimore were rendered as inputs to the deep learning algorithms', which 'learned how to recognise what a protest is, what a gathering of people in the city might mean' (Amoore, 2020: pp. 3−4). High school students were prevented from boarding buses to join the protest, people were arrested for their social media content and groups were apprehended on the basis of image recognition. These people were not necessarily targeted based on something they had done, but because an algorithm had learned via proxies to (mis)recognise their attributes and to infer intent.
Such is the composite nature of machine learning and its associative life. The attributes of dispersed people are correlated to supply the conditions for future arbitrary actions against unknown others (Amoore, 2020: 7). Machine learning algorithms parasitically feed on the data and actions of users. The algorithm does not care whether those actions indicate use or non-use (e.g. lurking, listening, deletion, taking a break, leaving); it is all data anyway. In other words, obfuscation or absence does not hide users' traces but rather exposes them. Moreover, the actions animating machine learning models need not belong to the targeted person. This means that someone's disconnective practice might be turned into someone else's connection and vice versa. As Amoore (2020) suggests, this posits 'the algorithm as always already an ethico-political entity by virtue of it being immanently formed through the relational attributes of selves and others' (p. 7). What are we to make of digital disconnection in light of social and technologically tethered beings? If the impetus to disconnect stems from a desire to reconnect with oneself and others, yet is not pragmatically durable, what other options are there?

Being singular plural
The writings of the French philosopher Jean-Luc Nancy (1991Nancy ( , 2000 offer some useful insights into what an associative life engendered by algorithmic systems might imply. Against more commonly held conceptions of community as aggregates of individuals, Nancy offers a theory that understands community, or plurality, as originary. It is not the individual that comes first but community, or better, being-with. As Nancy (2000) writes in Being Singular Plural, 'existence is with: otherwise nothing exists' (p. 4). However, it is not so much that one comes before the other, or of an origin that signifies 'that from which the world comes', but rather an 'infinity of origins' (Nancy, 2000: 15). By positing 'being singular plural' without any determined syntax, Nancy exemplifies how these are equivalent terms. 'Being is singularly plural and plurally singular' (Nancy, 2000: 28).
Being-with as an ontological commitment is particularly interesting in our discussion on digital disconnection for (at least) two reasons. First, Nancy's notion of being singular plural and being-with offers a theoretical basis for an understanding of machine learning. Although Nancy did not have machine learning in mind when proposing the notion of being singular plural, his ideas could certainly be read as descriptive of the inescapable relationality at stake in contemporary algorithmic arrangements. Second, if being-with for Nancy implies 'the inherent and indeed inescapable relations between singular beings through their transient but inevitable exposures to one another' (Willson, 2012: 286), there is nothing to disconnect from.
While community is often understood as a 'common being', a kind of fusion or communion of individuals, Nancy (1991) sees it as a 'being-in-common' (p. 29). As Willson (2012) writes, 'it is the in where community resides; in the relations or exposures that both touch but are also distinct and resistant: incompletely shared and thus partly and mutually constitutive; singular but also plural' (p. 286). If we accept these ontological commitments, then the ethico-political response cannot be to seek to fuse individuals together but to strive for a recognition that we are already responsibly and inescapably tethered together. Accordingly, common bonds are not a matter of belonging to the same political institutions or nations, but are developed through an ongoing process of communication and sharing. Incidentally, this is also what is happening in the case of machine learning. Because, for Nancy, singular being is already a touching and sharing, the nature of the political involves 'a community consciously undergoing the experience of its sharing' (Schwarzmantel, 2007: 463). Yet we rarely think of our algorithmic associative lives in this way. We should.
As people link, share, like, write, stay silent, lurk, listen, unsubscribe or take a break, they touch and are touched upon. This touching and sharing is not just what happens existentially, but is also commercially and politically exploited, as we have seen in Amoore's account of how deep learning is used by various actors to infer intent and to refine and edit the code for future uses in unknown future places. In a datafied and algorithmic world, what we do with and without our digital devices contributes to making the world appear in certain ways rather than others. What we experience when checking our Facebook feeds, reading an online newspaper or applying for a mortgage is community in practice. This is the community (un)consciously undergoing the experience of its sharing. On the somewhat more obvious level, it is the world brought to our attention by other people's sharing practices and processed by algorithmic infrastructures. But more fundamentally, what is experienced is the existential sharing of being-in-common, understood as 'circulating in the with and as the with of this singularly plural coexistence' (Nancy, 2000: 3).

What radical relationality means for disconnection studies
If being-with is the starting point, where does it leave the notion of digital disconnection, and what are the repercussions for disconnection studies? The fact that we cannot escape others makes prioritising the individual as the locus of voluntary disconnection seem naïve at best. It also complicates arguments about the degree of agency that actors have in disconnecting. With algorithmic arrangements and machine learning, notions of voluntary and involuntary actions get rather blurred. If digital disconnection is understood as an act of resistance and of mindful media use, then what is needed is to come to terms with the kind of 'radical relations' (Willson, 2006: 214) encompassed within such arrangements − that is, to articulate a stance on how seemingly mindful practices of detoxing, leaving or slowing-down partake in sharing and exposing particular others, whether consciously or unconsciously, explicitly or implicitly.
Because, as Nancy (2000) suggests, 'every position is also dis-position', there is an opportunity for disconnection studies to consider 'the appearing that takes the place of and takes place in the position' (p. 12). This would imply developing an ethico-political stance with regard to the co-creation and complicit shaping of the very algorithmic arrangements that one may seek to resist or refuse. The question is how to conceptualise digital disconnection when the very acts of avoidance and disengagement may place the actor not on the outside of algorithms but on the very inside. As Bucher (2018) argues, there is no clear outside to machine learning algorithms in the sense of a clear distinction between machinic and human agency. This is because the 'authorship of the algorithm is multiple, continually edited, modified, and rewritten through the algorithm's engagement with the world' (Amoore, 2020: 22). The being-with that we are concerned with, then, is not exclusively between human beings, but a mode of being-together that includes the more-than-human as well.
It is by starting from more-than-human co-existence that disconnection studies can deepen their conceptualisation of refusal and resistance. Even when platforms do their best to give the opposite impression by offering personalised news feeds and tailored recommendations, what Nancy's work alongside the logic of machine learning shows is that nothing is ever that personal. 'It is never about only one person, just me or just you', Lury and Day (2019) write, 'but always involves generalization ' (p. 18). It is about the attributes of one data population that comes to act on the propensities of another (Amoore, 2020: 87). What a user sees is never just a reflection of their past choices and clicks but the image of multiple moments and residues of data involving others. We might be able to control the use of a particular platform or app for a certain amount of time, but not the traces left by our non-use, relapse or resistance.
The opportunity for disconnection scholars, then, is to disconnect from the seductive idea of individual and voluntary non-use, and to start instead from the question of what it means to be many. If politics involves 'a community consciously undergoing the experience of its sharing' (Schwarzmantel, 2007: 463), then media resistance would entail staying with the trouble of bringing the performative shaping of our algorithmic environments into the conscious foreground of experience. Rather than positing individual practices to be futile, Nancy's notion of being singular plural locates the ethico-political response firmly in the with. How to rethink disconnection when all data touch and are touched upon? What to respond given that one is always already in an ethical relationship through the extant exposure to others? How to be comfortably and responsibly positioned when there is nothing to disconnect from?

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