Collective Likeness: Mimetic Aspects of Liking

When Facebook introduced six versions of the like button in 2016 (Facebook Reactions), it placated long-standing calls for a dislike button without directly complying with these requests. Ever since the like button had been introduced in 2009, users had come together in Facebook groups to demand a way of expressing “thumbs down” to complement the regular “thumbs up.” Many saw the absence of a dislike button as an exclusion of negative points of view. And even just a few years ago, there were still complaints that a lack of possibilities for expressing criticism, disapproval, and dissent on Facebook exerted pressure to conform.1 The vivid desire for a dislike button indicated that negative affects, too, could be a motivation for liking. Not only adorable kittens can become viral successes but also content that triggers passionate hate, outrage, and dismay. Because the like button carries connotations of a positive judgment, however, it did not seem to be the right operation for reacting, for example, to images of a natural disaster. Instead of introducing a dislike button, Facebook added a range of various emojis symbolizing love, laughter, amazement, sadness, and anger, in addition to thumbs up. Facebook thus multiplied the opportunities available to react to all different kinds of content. Just why the dislike button never had a chance becomes apparent if we understand the like button not as a value judgment or an expression of emotion but rather as an act of connecting, as a basic operation of social networking. Regardless of whether liking is given symbolically positive or negative connotations, it is always an act of affirmation, networking, or engagement. Accordingly, the opposite of the like button would not be a pejorative version of itself but an act of dis-connection, or dis-engagement. And indeed, a previously activated like can

be undone even without a dislike button. The difference articulated by the like is thus not: like/dislike but rather connected/disconnected. To put it pointedly: a user's opinion about something is completely irrelevant for the practice of liking as long as it affects them -be it positively or negatively.

Habit
But what does it mean to connect in social networks with content or other users? What is connection? The media studies scholar Wendy Chun has posed this question while rejecting models for explaining networks that she finds much too simplistic.2 How has the network become such a sweeping, universal model that it equally reduces phenomena as diverse as global financial flows, public transportation infrastructure, food chains in the animal kingdom, and social media platforms to an utterly ridiculously simplified diagram of nodes and edges (connections)? According to Chun, what makes network models so irresistible is that they make visible what would otherwise remain invisible: complicated, confusing social and physical movements. Network models translate dynamic interactions and events in time into a spatial structure of nodes and lines of connection and thus create an overview, an order, and a predictability of future events. But this always makes them, as Chun writes, "too early and too late."3 Networks are historiographic because they record previously occurring incidents of interaction as a line. And they are also predictive because they anticipate future interactions on the basis of previous, repeated interactions, which they project as lines of connection. This is why Chun also describes them as "imagined networks": "They describe future projections as though they really existed; they relay past events as if they were unfolding in the present."4 In this way, networks spatially represent events in time. And the basis for this spatial connection of nodes is the (potential) repetition of past events, which is what makes it possible to anticipate future events: the elucidation/imagining of connection. To be able to repeat is what links the machinic and the human.5 Understanding liking as an act of networking, as a link between a user profile and a piece of content, means counting on its repetition. Only if a like does not remain a singular event but is a repetitive practice carried out by the user-subject, can it be captured as a connection and hence as information. Facebook's newsfeed algorithms, which select and arrange content of a user's newsfeed based on their past likes, count on them liking things similar to what they have already liked. Wendy Chun thus suggests understanding connections that are projected based on repetition as habits: "Imagined connections are habits."6 With recourse to a number of well-known theorists of habit -ranging from Felix Ravaisson, Henri Bergson, and Gilles Deleuze to William James, Gabriel Tarde, and Pierre Bourdieu -Chun understands habits as conditions for predicting events, that is to say, for projecting networks. Habits are repetitive practices that can be acquired and learned through imitating others. For this reason, habits connect those who practice them, making these practitioners similar to one another.
Through embodiment, culturally acquired habits become "second nature." They thus transcend dichotomies such as nature/culture, inside/outside, conscious/automatic, human/nonhuman, and individual/collective. They furthermore oscillate between stability and change. Through repetition, they give permanence to what is ephemeral, and at the same time, habits are also open to change. For one thing, this is because they are only acquired through external impulses, which means that different habits can also be learned. And for another, this is because every repetition produces difference. Habits form the subject who practices them in relation to an environment; they generate a "way of being"7 or a "habitus."8 Habitualized liking is the social media subject's way of being. One cannot use Facebook or other social media apps in a passive manner. Without likes, friends, or connections, a profile remains amorphous, empty, and isolated. Only through repeated liking (or other acts of networking) do user-subjects take shape. And only through repeated liking can subjects' newsfeeds become their habitat, which supports their existence by encouraging them to express even more likes. "Like breeds like."9

Homophily
Others use the terms "echo chamber" or "filter bubble" to denote what I am here calling "habitat." Discussions about echo chambers or filter bubbles are predominantly shaped by concerns about the loss of democratic values, i.e., access to heterogeneous content and diverse perspectives and spaces for difference and controversy. These discourses accuse the algorithms of the dominating platforms of only showing us what we already know and like, that is, of merely reflecting back what we ourselves shout out, just like an echo chamber. They argue that the culture of likes has herded us into gated communities with people like ourselves by filtering out everything that disrupts this unity. Eli Pariser, who coined the term "filter bubble," writes: "the filter bubble confines us to our own information neighborhood, unable to see or explore the rest of the enormous world of possibilities that exist online."10 Wendy Chun follows up on this criticism by scrutinizing the networking mechanisms that allegedly produce echo chambers.11 According to Chun, a basic assumption for explaining the formation of clusters in networks has been incredibly influential in network research for years: homophily, the love of the same. In a 2001 sociological study that has proven authoritative for network science,12 homophily is defined as the idea that people prefer connections with people like themselves. Birds of a feather flock together. This can be seen, the study argues, in all kinds of personal networks: "Similarity breeds connection."13 The supposed similarity structuring connections is found here in the usual sociodemographic categories, although ethnic homophily caused the strongest differences in the formation of personal networks, closely followed by age, religion, education, profession, and sex. Chun argues that the axiom of homophily represents a grossly simplified notion of the influence between individuals, as well as a problematic reversal of cause and effect that attributes and thereby also justifies real injustices -such as racist or classist segregation of residential areas in modern cities -to a naturalized love of the same. According to this point of view, Muslim migrants, low-wage workers, or white middle class families simply like being together in their respective groups, and therefore in different parts of the city. Yet actual political causes for housing segregation remain hidden.
Homophily has subsequently become a foundational concept in network science -a simple answer to the question of how stable social networks are constituted and maintained. One reason is that the principle enables predictions: when network analysis identifies similarities, it is assumed that they will reproduce themselves, and thereby create connections. As a result, in all areas in which network analyses are applied (for example, as machine learning algorithms), historically developed forms of social discrimination are reproduced without the need for racist, classist, or sexist categorizations to explicitly come into play. Networks project past connections into the future and thereby reproduce what they claim to only describe. This makes them self-fulfilling prophecies. Chun names a self-learning prediction algorithm used by the Chicago police for crime prevention that predicts possible wrongdoing on the basis of social proximity and sociodemographic similarity, thus exposing minorities, and in particular African-Americans, to a higher risk of being arrested or murdered, without effectively lowering the crime rate.14 This bias problem in network science is enormous, and Chun is right to point it out. Yet despite her criticism of the one-sided axiom of homophily, she herself appears to remain wedded to thinking in dichotomies of homogeneity/heterogeneity, similarity/difference, or control/freedom in a way that does not adequately capture the transgressive, mimetic dimensions of network operations.
If we follow the arguments of Chun and other critics, the supposed love of the same leads precisely to the unity, homogeneity, and harmony of segregated filter bubbles. Pariser calls this "friendly world syndrome."15 The experience of everyday social media use, however, is of anything but harmony. Facebook, Twitter, and their ilk make opposing viewpoints and opinions, as well as modes of perception that are completely different from one's own, more available than ever. In comments to posts, one can take a seat at the kitchen table of political opponents. 16 If filter bubbles are a new phenomenon in times of social media, then this is only because they become perceptible in the first place in colliding and overlapping with one another. These kinds of encounters between different filter bubbles often erupt in brutal verbal hostilities and hate speech. Does this hate, then, truly lead to mutual rejection and a withdrawal into the homogeneous bubble, as Wendy Chun claims for the phenomenon of white flight?17 In social media, at least, other mechanisms are apparently at work. Hate and outrage are also emotions that create connections. Even hate breeds likes. Just think about the torrential power of shitstorms. According to the logic of the network, these waves of negativity are also breeding grounds of positive connections, just like any cute cat meme. The most bitter enemies find themselves in close proximity. Is Chun's claim, then, really true that networks do not model conflicts?18 Does the supposed love of the same actually eradicate all difference?
We get a different idea if we look beyond the duality of sameness and difference by instead focusing on the mimetic dimension of repetitive liking. In doing so, it is important to distinguish between sameness and similarity. Theoreticians of homophily, as well as Wendy Chun, imprecisely use the term "similarity" when they actually mean "homogeneity." Yet "the like is not the same" ("le semblable n'est pas le pareil"), as Jean-Luc Nancy writes in his reconceptualization of the concept of community.19 Things that are similar are also different. In a reversal of the naturalized causality of the axiom of homophily, the hypothesis would thus be: connection breeds similarities, and therefore differences.

Collectivity
The mimetic dimension of habitualized liking lies in the repetitive structure of habit, which produces similarities and differences with each repetition. But it also exists in the term's polysemy: "to like something" and "to be like something."  For Warhol, liking is a repetitive, machinic practice that makes people more alike. His interlocutor worries that this might hide a communist idea threatening individuality. But Warhol argues that the dissolution of individuality in similarity has already happened right within North America's liberal, capitalistic consumer culture -and it's something he likes. For Warhol, liking is a queer relationality that undermines the opposition of sameness/difference, homo/hetero, or friend/ foe, thus making possible collectives that harbor both similarities and differences. "Warhol is encouraging us to forget the sense that we must relate to others by way of either identification ('being') or desire ('having'), which itself relies on the opposition of same and different."21 Warhol makes a habit of liking things. This can be seen in his serial works of visual art and film (such as the silkscreen prints of Marilyn Monroe or his Screen Tests), his imitation gags (instead of accepting invitations to give talks, he sent friends dressed as himself), or his intense practice of collecting all kinds of things: from everyday objects and photos to expensive pieces of furniture and works of art. In these practices, Warhol shows a particular interest for flawed, imperfect objects. These flaws allow the objects within the collection to remain singular; they are not identical, but similar.
For Warhol, liking is a practice of perceiving similarities, producing similarities, and becoming similar. Liking constitutes a fundamental mode of being and perceiving that Flatley describes, in recourse to Walter Benjamin, as "a collector's relationship to the world."22 This mode of being is characterized by an affective openness and a search for correspondences. "[E]verything concerns"23: everything gains the potential to affect and become part of a collection, to relate to other things and people, to become like other things and people. Something becomes part of a collection because it is in some way similar to the things that have already been collected. But it also becomes similar to them by becoming part of the collection. The collection is slightly changed by every new item. Collecting is thus not only conservative but transformative. Together with the collection, it also forms and changes the collector.
Flatley summarizes this affective openness of the collector with a series of questions that could also be posed by social media users: "How am I like this? How is this like other things? How can I relate to this thing as somehow imitable?
In what way are we alike? How do we (mis)fit together?"24 Social media users can also be characterized as possessing an affective openness and a collector's relationship to the world. Everything concerns, everything has the potential to affect and become part of a collection. Liking is collecting and the collection of likes constitutes the likeness of the user. The user composes themself as a collection that transforms with every like. At the same time, the user is also an object of other collections. By being liked or befriended they become part of the likeness of others.
The practice of liking forms a complex network of relations of similarity that are constantly being renewed and changed. Hence, we could speak of a "collective likeness": first, in the sense of an individually attributable appearance as a collection; and second, as a collective similarity, i.e., a collective whose elements are like one another.
Liking is a practice that erodes and transcends the distinctions between same/different, self/other, individual/collective, thereby shifting the focus from nodes (individuals) to connections. As the legal scholar Antoinette Rouvroy has shown, the form of government of social media platforms -which she calls "algorithmic governmentality" in recourse to Michel Foucault -operates precisely through the avoidance of individuality. The "object … of algorithmic governance" is no longer individuals but "precisely relations: the data shared are relations and only subsist as relations; the knowledge generated consists of relations of relations; and the normative actions that derive from it are actions on relations (or environments) referred to relations of relations." 25 Facebook is not interested in individuals. The identity and personality of user-subjects, what they like, and what they believe, play no role in the implementation of Facebook's business model. Although the social media giant processes enormous quantities of personal data, which users knowingly share and leave behind as digital traces, the company's power is not based on an in-depth knowledge of its users -of their characteristics and personalities. Facebook sells targeted advertising that is much more differentiated thanks to network analysis than target groups determined by means of conventional demographic information. But this is more of a microsegmentation of the market than an individualization. Users are not addressed individually but rather assigned to certain, nuanced target groups (profiles) based on their habits. For example, Facebook offers advertisers the service of identifying "lookalike audiences." Lookalike audiences are Facebook users who are like, or act like, the advertiser's customer base and who -it is assumed -will probably also become customers. Individual users are targeted only inasmuch as they have connections to other users. By means of network analyses, correlations are determined that make it possible to construct the profile of a collective even based on interests that seem to be completely singular and particular. People like me.

Regulation
In the logic of the network, liking and hating are connected by the same mimetic operation. Both affirmation and rejection are expressed through an act of repetition, i.e., imitation, thus becoming part of one's own profile. Any user can take a look at the diverse patchwork that is their online likeness, since Facebook allows users a limited degree of insight into their own profile, which is automatically created based on their online behavior and serves as the basis for the selection of targeted ads.
In the settings of every account, it is possible to browse through and alter the categories and interests assigned to one's profile -categories such as "frequent traveler," "iPhone user," or "engaged shopper" and interests such as "Bauhaus," "Adidas," or "music videos." The latter form a disparate collection of certain fields of interest that are quite specific and accurate and others that appear to be completely arbitrary.26 In some instances, Facebook's assessments seem to directly contradict one's own inclinations and convictions. For example, the profiles of liberal atheists might contain categories such as "Mormonism" or "Republican Party."27 This is evidence for the fact that, as was explained above, networks are indifferent to the symbolic meanings of connections. Both love and hate are infectious, provoking likes and fostering engagement, resulting in similarities not only with friends and neighbors but also with enemies.
In the preface to the second edition of his Laws of Imitation, Gabriel Tarde recognizes the necessity of supplementing his social theory -which he bases on mimetic practices -with a negative form of imitation. He notes that societies are constituted by two forms of imitation. People imitate models by doing the same thing or the opposite.
In counter-imitating one another, that is to say, in doing or saying the exact opposite of what [people] observe being done or said, they are becoming more and more assimilated, just as much assimilated as if they did or said precisely what was being done or said around them.28 Following Tarde, we can use the term "counter-imitating" to designate repeated acts of connection that are motivated by hate, as well as the targeted provocations of trolls. Counter-imitation makes conservatives and liberals resemble each other. It connects misogynists and feminists. It joins together climate change deniers and climate-protection advocates in a collective. "But both kinds … have the same content of ideas and purposes. They are assimilated, although they are adversaries, or, rather, because they are adversaries."29 Yet this portrayal is not meant to suggest a harmonious picture. Hate directed toward people in social networks regularly results in existential threats for groups and individuals making it impossible for them to participate and sometimes even leads to physical violence, mass shootings, mental health issues and suicides. Platforms find themselves compelled to intervene by instituting regulations -deleting content and blocking users -in order to remain hospitable for as many users as possible.
Mimesis possesses two poles, as Friedrich Balke demonstrated, as do mimetic operations in social networks: one that regulates and one that is excessive. As a principle of regulation, mimetic practices support an already existing order. Pierre Bourdieu's concept of habitus can be located at this regulatory end of the mimetic spectrum.30 According to Chun, habitualized liking as an embodied mode of being is both the basis for predicting user behavior and the condition for modeling networks. Driven to an extreme, however, regulative mimesis turns into mimetic excess that destroys systems of order. This polarity is also found in the mimetic structure of habit. When they are excessive, good habits become addictions and obsessions.31 In 2014, Mat Honan (an editor for Wired magazine) conducted an experiment in which he liked everything he encountered on Facebook. Very quickly, not only his newsfeed changed but also that of his Facebook friends. Postings from his friends disappeared. Instead, his screen was flooded Platforms such as Facebook must protect the order of the network, which is based on consistent, predictable user behavior, from excessive hate and excessive liking, because excessive mimesis destroys so-called good order by transforming it into a hostile environment and undoing predictability. To accomplish this, platforms must carry out policing measures that institute regulations and enforce their own community standards. Users are admonished to accept good social networking habits and let go of bad habits in order to achieve a "healthy" and satisfactory user experience and remain predictable for the platform. 33 For example, since 2015 Facebook has been carrying out a large-scale multimedia campaign encouraging users to claim sovereignty over their own Facebook profile: "Mach Facebook zu deinem Facebook!" (Make Facebook Your Facebook!), is the call.34 This includes a series of didactic videos informing users about how they can control and shape their own account. The first versions of the campaign in 2015 and 2016 aimed, in particular, to encourage reserved German users to become more active in their behavior, and to communicate the idea that repeated liking, following, and sharing -as well as unliking (hiding a post, unsubscribing to a page, blocking users) -could make Facebook into a personal habitat: "Mit mehr von dem, was dir gefällt. Und weniger von allem anderen" (With more of what you like, and less of the rest). 35 Since 2017, the campaign has aimed at communicating options for purging the results of bad habits. In the advertising videos, Facebook users complain, for instance, that they are being shown too many cat videos and not enough art and culture, and Facebook shows them how liking and unliking can help. The goal is to animate users to create an environment for themselves through connective activities that foster the production of more connections, making users more valuable for Facebook.
The mimetic structure of liking forms the basis for the permanence and predictability of network order. At the same time, it possesses the power to threaten the existing order. Facebook reacts to this danger with regulatory measures. But critics of homogeneous filter bubbles, too, act out of concern for an order of sameness (control, censorship) and difference (freedom, democracy) that the mimesis of liking questions. In doing so, they ignore the collectivizing function of liking. Contrary to what Pariser claims,36 one is not alone in a filter bubble but part of a collective likeness with others. Particular niche interests establish a relationship of similarity between users who otherwise share nothing with one another. Acts of connection motivated by affinity or rejection bring us closer together -to our friends as well as our enemies.