Computational Culture

a journal of software studies

Article information

    Irina Kaldrack and Theo Röhle

    Digital Cultures Research Lab, Leuphana University Lueneburg & Institute for Media Research, Braunschweig University of Art

    Publication Date
    9th November 2014

Print Options

Divide and Share: Taxonomies, Orders and Masses in Facebook’s Open Graph


The Open Graph protocol, introduced in 2010, has allowed Facebook to extend its reach far beyond the confines of the platform itself. It provides the basic technical infrastructure of connecting and sharing and encourages specific forms of analysis and usage. We argue that, if Facebook is to be conceptualized as a mass medium, the Open Graph is where media and masses mutually (re-)configure one another. In order to disentangle these relationships, we investigate backend and frontend practices from three different angles – descriptive, analytical and historical – and investigate how seemingly incompatible media promises converge.

‘Facebook helps you connect and share with the people in your life,’ according to the equally euphemistic and oddly reserved self-description of the most popular social media platform. If this slogan is taken seriously – which certainly seems opportune in the light of the ‘mass’ of a billion users – then the questions arise ‘if’ and ‘how’ Facebook is to be conceptualized as a mass medium in terms of connecting and sharing.

To pursue this question we focus attention on one of the central areas of Facebook’s technical infrastructure: the ‘Open Graph’ protocol and the applications, such as the ‘Like’ button, that build upon it. The Facebook ‘Like’ button has become an omnipresent component of the World Wide Web within only a few years. Not only entire websites, but also individual texts, videos and photos, can be ‘liked’: At first glance, the function of the ‘Like’ button seems primarily to consist of displaying the degree of popularity of online content by means of the number of likes any item of content receives. A closer look reveals, however, that the increasing spread of the underlying Open Graph protocol is establishing new classification schemata on the web.

We investigate how an interplay of connecting and sharing emerges in the application of the Open Graph, where media and masses mutually (re-)configure one another. We argue that Facebook Open Graph encourages specific forms of analysis and usage, thereby consolidating seemingly incompatible media promises. In order to disentangle these relationships, we make an analytical distinction between the technological level and the level of usage. This distinction serves as the basic structure of our paper and is elaborated through three subsequent stages – with a descriptive, an analytical and a historical outlook respectively.

By way of an introduction, we first describe the technical details of the markup protocols and procedures of analysis that are part of the Open Graph. What interests us here is the question of which forms of connecting and sharing the technical infrastructure permits. Then, we discuss what connecting and sharing means from a backend perspective, i.e.. as website owner, and a frontend perspective, i.e.. as ‘normal’ user. By contrasting these perspectives, we seek to identify the respective media characteristics that Open Graph develops on either side. We notice that the Open Graph handles the distribution of content and lets website owners trace the paths their content takes, thereby organizing the relation between content and users. For Facebook users, the circulation of findings, opinions and comments constitutes a space of co-presence, where findings and opinions can be shared among peers.

On the analytical level, we discuss epistemic characteristics of Open Graph – mathematic-algorithmic principles, staging and using – in terms of how ‘mass’ and ‘media’ are configured. We argue that, from a backend perspective, Open Graph becomes a means of observing emerging orders. It attains media quality via its specific scaling properties: merging the methodology of statistics and network analysis, it promises to offer lossless scalability, dispensing with the need to choose between the reduction inherent in each of these methods. From a frontend perspective, Open Graph gains its media quality by providing a simulation of co-presence, while structurally constituting micro-publics without clearly demarcated borders. While these micro-publics, quantitatively speaking, can accommodate ‘masses’ of users, we argue that they involve a mode of ordering that sets its apart from the mass proper.

In the third part of the paper, we engage in a historical reconstruction of specific epistemic traditions relevant for an understanding of Open Graph. Focusing on the work of Gabriel Tarde, we discuss a major epistemological shift in the late 19th century: whereas statistics in the tradition of Adolphe Quetelet’s Social Physics sought to identify laws of the social in the large-scale aggregation of the features of actors, Tarde laid the theoretical groundwork for an analysis based on relations between actors. We argue that the way in which Tarde envisaged the convergence of these perspectives, together with his notion of masses, publics, and the role of media, can be related to the phantasmagoric notions of scalability and emergence inscribed into Open Graph. However, rather than a smooth convergence, what can actually be observed in Open Graph today are tensions between these approaches. This insight leads us to our main argument that Open Graph, at its epistemological and infrastructural core, is characterized by an oscillation, a persistent switching of modes, that constantly produces lacunae begging to be filled with data. In the final part of the paper, we discuss how this ‘urge to fill in’ instilled in the technical infrastructure is negotiated on the user side.

From ‘Like’ button to Profile – The Technical Infrastructure of the Open Graph
Open Graph is a protocol developed by Facebook, on the basis of which content on the World Wide Web can be labeled with metadata and connected with Facebook. It draws on ideas developed in the context of Semantic Web initiatives, more specifically the Resource Description Framework (RDF) and the Friend-of-a-Friend project, but implements these in a proprietary way.1 The Open Graph Protocol was introduced by Facebook in April 2010, consolidating features that were earlier part of the Facebook Platform and Facebook Connect. Following Gerlitz and Helmond2, the Open Graph Protocol can be seen as part of a larger trend towards integrating external content into social media platforms. Until 2006, sharing links to other websites on Facebook had to be done manually by copying URLs into a form. The introduction of Facebook buttons, which could be placed on external websites, made sharing content less of an effort for users, but also expanded the reach of the Facebook platform much further into the World Wide Web. The ‘Like’ button, which today seems emblematic for this development, was first introduced as an internal feature as a means to comment positively on others’ status updates. It was introduced as an external feature in 2010 together with other Social Plugins that were part of the new concept of the Open Graph Protocol.

Today, the protocol is the technical basis for most of the possibilities for using Facebook. Its central role within the Facebook infrastructure first becomes apparent, however, when looking behind the latter’s graphical interfaces: the special feature of this protocol is that it makes it possible to formally describe and automatically determine content, both within Facebook and outside it. This can be demonstrated, for instance, with the contents of the movie database If information about a movie is retrieved, there is a ‘Like’ button on the relevant page. Clicking on this button causes the movie to be displayed automatically in a user’s own profile under the category ‘movies’, and the like then appears in the user’s Timeline and in the News Feed of the user’s ‘friends’.
Users and movies are stored as ‘objects’ in Open Graph, and the like indicates the time and the way in which these objects were linked with one another. On a technical level, Open Graph thus organizes connecting and sharing via addressing, categorizing and distribution: websites outside of Facebook and content elements both inside and outside of Facebook are rendered addressable as objects via an individual identification number. They are also categorized as specific object-types with a range of properties. Clicking the Like button or using another Social Plugin that is associated with an object leads to this action (potentially) being displayed in the newsfeed of contacts. This distribution is based on the users’ address management – their definition of ‘friends’ and their privacy settings.

Circulation and Relations – The Open Graph Backend
In order for content on websites outside Facebook to be integrated as objects within Open Graph, those responsible for the respective websites must include metadata on them, labeling their websites or parts of it as Open Graph objects. Marking up content with metadata according to the Open Graph Protocol has advantages for website owners: they can more precisely track who has visited the website and which elements have been viewed on an individual page. Each time an Open Graph object is accessed, this is recorded by Facebook along with any subsequent interaction with the contents or plugins of the object. This information is made available to the owners of the respective websites in two forms: on the one hand they can call up statistical compilations through the service Facebook Insights. These statistics consist of information about frequency, which can be filtered and sorted according to criteria of time, geography, and demographics3. On the other hand, Open Graph objects are given their own profile on Facebook, through which, for example, comments by individual users can also be compiled4.

However, the analysis tools that Facebook makes available are only a rudimentary variation of the possibilities for analyzing the recorded traffic. An application programming interface (API) also allows website owners to perform a direct and automated query of Open Graph, with which substantially more comprehensive statistics about the use of their contents can be compiled. In this way, website owners can determine, for example, the number of users who came across certain content through recommendations from a friend. Limitations that exist for this kind of analysis, e.g., concerning personal information, such as gender, age group, location, etc., can be easily circumvented by suggesting the use of certain apps to the users. When apps are installed, the app providers are given access permissions to personal data ranging from the name and friends list all the way to access to the message inbox5. Once these permissions have been granted to an app provider, the relevant information is made automatically accessible to them through the API.

This kind of traffic analysis is used primarily by major content providers who want to track more precisely where and how their content is circulating through social media. This also includes technical measures that automatize the distribution of content, thus advancing its circulation in the networks. One example for this is the app of the eReading device, Kobo. This app registers when Facebook users (if they are logged in) start reading a book, when they highlight a passage, take notes and when they finish reading a book. It subsequently informs the user’s circle of friends through the News Feed that ‘user X started reading [finished reading/highlighted a passage/wrote a note in] book Y’. Following the same schema, Tumblr, for example, informs a user’s circle of friends about which elements were posted or reblogged. At the same time, the chain of comments, likes and further interactions following from these announcements can be tracked by the staff of Kobo or Tumblr, respectively.

There are certain technical conditions that must be fulfilled in order for an announcement like ‘user X started reading book Y’ to be displayed in the Facebook News Feed. The first precondition is that a definition exists in the Open Graph of what a ‘book’ is and what is to be understood as ‘started reading’. For these kinds of definitions Facebook provides a form for app developers to define similar combinations of actions and objects, for example, ‘running’ a ‘course’ or ‘rating’ an ‘episode’ of a ‘TV show’6. Criteria for what constitutes the relevant situation here may be any interaction that leaves a trace in the Facebook database, such as clicking on a link, looking at content that is marked up with the Open Graph code, or the location announced by a smart phone.

At this point, a convergence of states of interests can be observed: the spread of Open Graph over the entire web in the rudimentary form of the ‘Like’ button has already made it possible for Facebook to track patterns of linking and usage outside its own platform. Definitions of actions (e.g., cooking) and objects (e.g., recipes), when viewed together, additionally provide Facebook with a previously unattained insight into the ‘semantic’ dimensions of the web7. This is possible because the Open Graph offers sufficiently strong motivation for adding metadata to content by offering the owners of websites and apps substantially more extensive and detailed possibilities for analyzing traffic, as well as contributing to a faster and broader circulation of their content.

For website owners, Open Graph provides an infrastructure for the circulation of content and simultaneously lets them trace the paths that individual elements are taking, e.g. via sharing and commenting. Thus, Open Graph acquires media characteristics by allowing them to observe the relations developing between users and their content as well as between users via their content. For Facebook itself, the nature of Open Graph as an infrastructure of observation is even more apparent: based on the classification of objects on the one hand (which is provided by the website owners) and the interaction with objects as well as their subsequent distribution (which is initiated by the users) on the other, Facebook is able to gain an aggregated picture of how these relations develop over time.

Sharing and Co-Presence – Using and Staging the Open Graph
From the point of view of Facebook use, Open Graph is involved primarily when the ‘Like’ button or other Social Plugins are used and when apps are installed and used. The metadata remains hidden from the users; it is only evident in the automatic categorization of contents. Nevertheless, Open Graph also plays a role in use: the features and the way they are staged suggest integrating objects in one’s own profile – users collect ‘finds’, present them to one another and can exchange comments on them. As described above, apps can generate additional information on the kind of interaction with certain content. The automatic integration of objects in the News Feed and the Timeline directs the attention of the ‘friends group’ to a user’s activities, thus prompting comments on these activities, further sharing of them or posting one’s own recommendations in reaction.

In this way, the technical infrastructure of Open Graph enables specific (communication) practices resulting in a certain form of visibility8. On the one hand, Open Graph can be used to maintain one’s own profile and present preferences through the links to categorized content, on the other hand, the ‘Like’ button and other Social Plugins are used to refer to other web content from within Facebook. The links appear simultaneously in one’s own Timeline and in the News Feed of ‘friends’. Conversely, the links of ‘friends’ and the comments written by their ‘friends’ appear in the News Feed of Facebook users. Whether posts are included in the News Feed is determined by privacy settings, individual filters, and by the News Feed algorithm, formerly referred to as ‘EdgeRank’. The algorithm rates how relevant a post might be for a user calculating different measures, e.g., how often and in what form the receiving user has interacted with the sending one, how many comments a post already has, to which forms of posts the user reacts most often and how old posts are9.

Use of the News Feed engenders a visibility of showing and opining. What can be seen are links to contents that one’s ‘friends’ have seen, partial information about how they interact with these contents, and comments on the contents. This visibility is based on lists of friends and takes place as automatic distribution, partially constrained by the News Feeds algorithm, individual filters and privacy settings. Through the address management based on lists of friends and traffic between contents, ‘friends’ and ‘friends of friends’, Facebook arranges the profile, Timeline, and News Feeds of the participants. Once a post has made it into these areas of the interface, the whole chain of subsequent interactions, such as comments and likes, is accessible along with it. The resultant visibility aims for ‘friends’ to follow the links of others, to repeat and comment on their perceptions and interactions. Put in terms of connecting and sharing: to connect is to manage one’s own list of friends and to share is to show and comment. Thereby, the Open Graph stages a space of co-presence (especially in the form of the News Feed), in which the individual users make contact with their ‘friends’, see what they have seen or done, and allows them to express opinions about this.

Pertinent findings are also suggested by investigations of the use of new media by US American young people. The book, Hanging Out, Messing Around, Geeking Out, connects the results from various studies conducted from 2005 to 2008. For the field of communication media – which also covers Facebook use and the ‘Like’ button – the authors note that young people seek ‘to construct spaces for co-presence where they can engage in ongoing, lightweight social contact, that moves fluidly between online and offline contact’10. In an overview of research concerned with sociality in Social Network Sites, Ellison and boyd also state that:

‘Early research on the topic suggested that SNS users were more likely to articulate existing relationships on social network sites than meet new people […]. Recent research on Facebook suggests that connecting with close friends is more common than using the site to meet new people, but that using the site to find out more information about peripheral others, such as casual acquaintances or someone one has met socially, is also a strategy employed by users […].”11

Facebook Open Graph stages a visibility that prompts users to repeat the perceptions of others and exchange opinions about them thus becoming a medium of a space of co-presence on the user side.

Knowledge Forms of the Masses
Facebook’s Open Graph is the basis of a constellation, in which technical functionalities of addressing and distribution are interlocked with categorization, staging and usage. At a media level, it enables the observation of relations between users and contents for website owners and appears as a space of co-presence for users. Regarding our main question, whether Facebook can be conceptualized as a mass medium in terms of connecting and sharing and if so, how, it seems necessary to determine what notions of mass are produced by Facebook on the different levels we are analyzing. In order to do this, we first reflect generally on the co-constitution of medium and mass in the case of the Open Graph. Then, we proceed to define more precisely the media qualities of Facebook for backend users like website owners compared to ‘normal’ frontend users.
Fundamentally, Open Graph is only imaginable in the mass: for the effective circulation of content there needs to be a sufficiently large number of users; the semantic markup of content depends on the mass mobilization of website owners; the crystallization of orders at the social level requires the mass interaction of users with online elements and with one another. Our central argument is that Open Graph interlocks classifying and communicating in such a way that new arrangements of order emerge. Thus, by organizing the relations between users and content, and visually staging these relationships in a specific way in the interface, Open Graph simultaneously establishes a specific understanding of ‘mass’ and gains media characteristics.

If one focuses on the question ‘which kinds of order are engendered through the technical infrastructure?’, two points can be initially noted: the Open Graph is accompanied, first of all, by a massive expansion of addressability. Web content is no longer exclusively accessible through a URL, which applies to an entire page or section of a page, but also through an ID that is administered by Facebook. This lets website owners track which Facebook users are interested in certain elements of a page. From a more abstract point of view, the expansion of addressability provides a significantly increased quantity of traceable data, but at the same time many differences are being leveled. The integration of heterogeneous elements – user profiles, text communications, photos, film descriptions, recipes, etc. – into a common space of addressability requires a starkly reduced model. In the present case, this reduced model is analogous to network analysis – nodes and edges correspond to the objects and connections of the Open Graph. However, this reduced model is not the final result, since classifications are subsequently carried out in the data material, thus creating new categorizations. Following the idea behind the RDF specification formulated by the World Wide Web Consortium, Open Graph merges the model of network analysis with semantic categorization. Categories are defined either by Facebook itself, for example, when a movie from the Internet Movie Database appears in a Facebook profile under the category ‘movie’. Or, as in the case of the Kobo eBook being ‘read’, they are provided by app providers on the basis of the Open Graph protocol. Thus, Facebook Open Graph provides an infrastructure that is based on the expansion of the quantities of data that are addressed and categorized. At the same time, the Open Graph apparently aims to generate traffic, in the staging and in the usage offers, which is then available as data and metadata material for further analysis.
Addressing, leveling, categorizing, and mass traffic are the crucial characteristics of the Open Graph at the technical level. From an infrastructural perspective, these characteristics firstly produce masses of data through the expansion of addressing, the leveling of differences and the production of mass traffic. At the same time – and this happens on the level of protocol and standards – Open Graph serves as a basis for a specific ordering of these masses: stored addresses and links are being read as objects and actions (i.e. relations) between them, categorization serves as a means of allocation meaning to these addresses and links. Finally, the recorded traffic allows the usage of links to be analysed in terms of frequency.

Together with the Open Graph protocol and the mass storage of data by Facebook and app providers, this provides the foundation for a mathematical analysis. It is the application of this analysis that renders Open Graph a medium of observation in the service of the website owners.

Observing Emerging Orders
In the following, we outline the basic concepts of these procedures and the knowledge forms inherent to them. The specific orderings of masses intersecting in the Open Graph contribute to establishing it as a medium for website owners. We argue that, from a backend perspective, Open Graph promises to allow the observation of self-organizing masses, i.e. the emergence of order, and to be able to do this with lossless scalability.

Building on the data provided by Open Graph, methods from statistics and data mining can be combined with methods from network analysis based on graph theory12. The basis for the statistical analysis is the data collected for an actor. This data corresponds, for example, to properties such as male/female, age or other profile information from Facebook users. In this way, a list of features can be attributed to each Facebook object. A set of Facebook objects can be ordered in terms of one of these collected features using specific metrics and various calculation procedures. Here the metric determines the way in which the collected data is to be made comparable to a feature, thus providing a measure for proximity or similarity13. The deviation from the mean value can be used as the order of the Facebook objects considered in terms of a certain characteristic.

A further question is whether there are interrelations between different features, which are also a form in which order is installed within a mass of observed objects. Two of the most important methods for determining these kinds of relations are: correlation analysis and (linear) regression. The method of correlation observes the extent to which two features of a data set (in other words two different features of an object) deviate from the mean value of these features, simultaneously. Regression is used to attempt to determine the connection in the form of a function and to make it calculable, in other words, to determine the dependencies. Thus, statistical knowledge is based on registered features and a posited measure of similarity. On this basis, elements can be ordered according to their proximity or deviation, and the considered features in terms of their mutual dependencies.

Mathematical graph theory investigates graphs in terms of their structures14. It starts from pair relationships between nodes and certain conditions, such as the total number of edges or the smallest degree occurring in a graph. Following from this, the next questions graph theory asks are whether certain sub-graphs occur, how long the walks are in the graph, or how many nodes (or edges) must be removed, so that the graph disconnects (i.e., sub-graphs emerge that are not connected by walks).

Thus, Graph Theory asks which structures can be deduced from the existence of pairs of relations between elements of a set. This is less a matter of making statements about the concrete nodes than generally establishing the existence of certain (sub-) structures in networks. Network analysis then shifts the focus somewhat. Starting from the model of the graph, it deduces structures in the empirical networks it has collected, asking about subgraphs, connectivity, etc., in networks. What is crucial, however, are the graph-theoretical measures of the ‘importance’ of nodes, especially due to different centrality measures. These examine how ‘central’ a node is, for example, the extent to which a node has especially many adjacent nodes relative to the set of edges in the graph, or whether a node is connected with all the other nodes of the network through especially short paths. In addition, adjacency relationships can be made calculable on adjacency matrices using matrix operations 15 – in a sense these calculations involve logical connections of relations between (not necessarily adjacent) nodes.

The fundamental logic of network analysis is also relevant to Open Graph: the expansion of the space of addressing creates the possibility of grasping data traffic as the production of links, or edges, between addresses, or nodes. More specifically, network analysis interprets nodes as actors and edges as the communication events between them. From this perspective, the knowledge network analysis develops is based on relationships between actors, rendering structures which are evolving out of these relationships observable. In this respect, network analysis is an explicit counter-program to a kind of sociology, which seeks to describe actors based on attributes and statistically measure them: ‘The actor is primarily interesting as an abstract concept, with which the different relationships can be determined and observed, which are interesting for network research’16. Although network analysis based on graph theory focuses on single nodes and their pair relations, it focuses specifically on their function or value in relation to the network as a whole. From the network-analytical perspective, this value does not arise from personal characteristics, or exclusively from its direct relations to others. Instead, the value is based on adjacency relations in interplay with the relations of other actors. In this way, Open Graph promises to make structures which emerge from the interactions between a mass of web users and their content observable. This promise is based on the combination of the more abstract mathematical analysis and the more actor-centered sociological analysis. As Open Graph provides data that can be employed for both statistical and network analysis, there are also various ways to blend these approaches in combined analyses.

The statistical analysis of profile data allows both for the determination of types of profiles and for the subsequent classification of every available profile. If profile data is connected with traffic, the value or the function of the node in the network can be initially determined17. If this analysis is conducted with sufficiently many nodes and sub-networks, by means of statistical analysis, we can ask how the network-analytical value of a node correlates with its profile or profile type. Open Graph also provides the possibility of bundling certain constellations of nodes and edges into a node. The profile of a user, for example, can be considered as a network of nodes and edges and can be examined in this respect in terms of relationships between the elements. However, the complete profile of the user can also be grasped as a node and examined network-analytically for relationships to other profiles.

Statistics and network analysis thus provide very different kinds of perspectives: statistical methods assert similarities, normalities or deviations, focusing on the features of actors. Network-analytical measures supply statements about the value of individual nodes (or node associations) for the structure or topology of the network under consideration. Whereas the taxonomy based on statistics employs a type as representing a group, thus severing the connection to the micro-level, in network analysis the topological structure underlying the group remains at least theoretically visible. Through the combination of these two classifications, the difference between the micro and macro-levels is dissolved into a question of scaling. The scaling method decides whether a phenomenon is perceived as a micro or macro-phenomenon, whether an entire group, for example, is considered as a single node in a network, or whether the focus is on group members as nodes. The statistical aggregations can be used to define characteristics of a group on each of these levels18.

The statistical analysis thus fills the gaps in the network-analytical arrangement and vice versa. Since both methods allow for specific means of scaling and hierarchization, their combination in Open Graph renders ‘connecting’ and ‘sharing’ observable in a new way. It suggests the potential to attain lossless scalability – to be able to zoom in and out of features, nodes, and networks without having to commit to one particular perspective. In this respect, Facebook Open Graph becomes a medium for observing the self-organization of masses via orders of groups, types and topologies. The promise of observing and explaining the bottom-up emergence of structure is fueled by the supposed potential for lossless scalability.

Following our reflections on the relationship between medium and masses at the backend of Open Graph, with the observation of emergent orders as the core media characteristic, we now turn our attention to the relationship between medium and masses on the user side. In the description of manners of use, we have noted that the Open Graph is a medium offering a space of co-presence, which simulates shared presence and promises participation. These groups emerge because their members engage with similar content as a collective and exchange opinions about it, and through these kinds of interactions and reciprocal effects a ‘we-feeling’ crystallizes19.

At the level of usage, Facebook Open Graph thus does not initially address ‘masses’ per se, but rather the constitution of groups. It promises shared presence and participation in the life of ‘friends’ as perceptions and opinions circulate among them. In reference to the relationship of individuals, groups and masses, Facebook seems to create a space of co-presence for a collective: a group is formed in which everyone is looking at similar contents and exchanging and balancing opinions about them. At a closer look, however, this is more the imagination of a collective. Every single Facebook user can imagine their ‘friends’ as a group. Nevertheless, the ‘friends’ of one person are not necessarily friends with one another. In some cases they become visible to one another in comments on the contents from the mutual ‘friend’, but otherwise perceive no content from one another. If Facebook Open Graph appears on the user side as the medium of a space of co-presence, then this must be qualified from a structural perspective: it is more a matter of overlapping spaces of co-presence. Additionally the borders of these overlapping spaces sometimes get perforated or porous: Via the chains of interaction and comments on findings, ‘friends of friends’ become visible in one specific space of co-presence, even though they do not actually belong to the relevant friends list.

In the case of Open Graph, it thus might be more apt to grasp the relationship between medium and mass as the constitution of publics. Based on our findings so far, we can generally subscribe to danah boyd’s definition of ‘networked publics’ as ‘publics that have been transformed by networked media, its properties, and its potential’20. However, boyd also states that ‘social network sites are publics both because of the ways in which they connect people en masse and because of the space they provide for interactions and information.’ Following from our discussion, we see a need to provide a sharper distinction between publics and masses.

We have argued that the spaces of co-presence offered by Facebook are rather the result of the users’ imagination than them being actually implemented in a technical way. Instead, Open Graph organizes masses by dividing them into groups that one could describe as micro-publics. The key feature of these micro-publics is that their boundaries are vague; since there is always the possibility of comments on comments spawning off into other users’ networks, the actual audience for a status update or comment remains invisible. However vague these boundaries are, micro-publics are nevertheless characterized by a certain order, or at least half-order, in a way that distinguishes them from the mass outside of them. While there is always the possibility that an update or comment reaches the mass, thus integrating a certain part of this mass into the order of the micro-public, the mass essentially remains at the periphery of the users’ perspective – much like the parts of the visual field blurred by a scotoma. In this perspective, seen as a medium, Facebook Open Graph is not so much about masses as about subjects confronted with masses, dividing themselves from the mass by constituting micro-publics and imaging these as groups in spaces of co-presence.

Disposition and Mass
We have analyzed what kind of relations between connecting and sharing are being produced by Facebook Open Graph, both on the different levels of mathematical-technical procedures and on the different levels of usage. We have also discussed how notions of masses and media are interrelated in these settings: For website owners, Open Graph represents a medium for observing and analyzing a self-organizing or self-ordering mass. It holds the promise of enabling lossless scalability between individuals, types and groups. For Facebook users, Open Graph seems to be a medium of co-presence, although it might be seen more precisely as a medium for micro-publics. For us, this analysis triggered an irritation: How does a mere technical constellation of addressing, categorization and distribution acquire this status as a two-fold medium? In our view, it does not suffice to map this two-folded nature onto the frontend/backend-distinction we employed. Rather, in order to answer this question we need to discuss in more detail the genealogies of media promises that can be found at the intersections of figures of thought, formalization and their operationality. Engaging in such a historical reconstruction, we will argue that Open Graph represents a folding together of specific epistemic traditions.

The following spotlights on the history of statistics and network analysis show how the mathematical practices we have been discussing tie in with traditions of sociological thinking and their corresponding forms of control. Additionally, these formations interconnect with (electro-) physical conceptions of mass, as well as with specific notions of mass media. Our genealogical path starts with social physics using statistics as an instrument for observation and government of populations – thus moving an epistemic apparatus from astrophysics, where it had been an instrument to calculate observation errors, into the social realm. As a second tradition we discuss the advent of a different mass-order starting from relations between actors. The important point is that this understanding of mass starts to pervade media discourses on the one hand and becomes a central tenet of network analysis on the other. We discuss how these different epistemological traditions have become entangled in Open Graph and argue that the resulting oscillation constantly generates lacunae as well as a corresponding ‘urge to fill in’. The media quality of the Open Graph is thus its ambiguity, allowing it to merge different media promises.

As a starting point for looking at the historical developments in statistics, we chose the works by Adolphe Quetelet (1796-1874). In his main work Essay on Social Physics 21, Quetelet takes up approaches from error theory, which was prominent at that time in astronomy. This theory was based on the assumption that measurement errors in the observation of the orbits of planets followed a certain distribution and that the actual orbit of a planet could be calculated on the basis of this distribution. Quetelet transferred this idea to social processes and systematized them into a law of deviation, which stated that human characteristics, such as height, for example, followed a fixed distribution curve. Quetelet considered the mean value of this distribution – the average man – as both an abstract law, which became evident behind the mass of empirical observations, and as an ideal, which could be applied as a normative measure to society as well as to individuals.

Quetelet’s recurrence on the law of large numbers allows him to ignore the variability of individual actions as well as the motivations guiding these actions, because they converge in the mean value. This perspective is also accompanied by a specific notion of governing: a well-governed society is distinguished by its orientation towards the mean value, because this is what represents ‘all that is great, good and beautiful’22:

‘Quetelet elevated his ‘average man’ […] specifically to both an aesthetic and a political ideal type by linking the collective symbolism of balance, stability, the optimum and beauty to him. […] Quetelet’s emphasis on invariant normalities had to tend toward a purely proto-normalist strategy: empirical and mathematical statistics were supposed to discover ‘natural’ normalities, to which policies could then be oriented.’23

In this sense, the individual is ignored in order to delineate types in terms of the ‘Law of Large Numbers’.24 Following Harré25, such group definitions can be called ‘taxonomic collectives’: collectives whose coherence is achieved exclusively through an external observer and in which no internal structural formation processes have taken place26. This approach is still relevant as one of the ordering mechanisms implemented in the Open Graph. The categorization of objects, as well as of data-traffic – the metadata – is a ‘top-down’ operation of taxonomy. These features, together with the metrics used, put the traffic in order and allow the suggestion of friends’ groups based on certain attributes, as well as measuring the importance of instances of communication, thus creating “taxonomic collectives”. But it is important to keep in mind that this is only one part of the infrastructure.

We have argued that Open Graph holds the promise of rendering the emergent ordering of masses observable. This represents a departure from the top-down approach involving ‘taxonomic collectives’ and a focus on structures evolving out of relations engendered by data traffic. In a historical reconstruction, we can identify a similar epistemic shift most apparently in the work of Gabriel Tarde, who we have already mentioned. Besides his profession as a statistician and his involvement in late nineteenth century discussions on mass psychology, Tarde also provided a strong impetus for the later development of network analytical thinking. Therefore, he is a key figure for understanding the promises entailed in the amalgamation of these methodological approaches.

Tarde’s early works encompass mass psychology as well as criminology. In the late Nineteenth Century, a decisive question for mass psychology was how a crowd, an unordered assembly of individuals, could suddenly merge into a unit, which functions as a single acting subject, integrating individuals into itself, thus dissolving their individuality (and rationality) to a certain extent. Trying to explain these kinds of phenomena, Tarde proposed an understanding of sociology according to which society is emerging from reciprocal effects among individuals27’. See Georg Simmel, Soziologie. Untersuchungen über die Formen der Vergesellschaftung (München: Duncker & Humblot, 1922), 4. Own translation.]. The central reciprocal effect is imitation, which is primarily the repetition of ideas and opinions28 through suggestion and transmission. What was decisive for the early Tarde regarding masses is that normal social interactions of imitation and suggestion increase in intension and speed in the mass. Individuals, consequently, enter into a kind of common mental state, which turns the mass into a unit: ”A mob is a strange phenomenon. It is a gathering of heterogeneous elements, unknown to one another; but as soon as a spark of passion, having flashed out from one of these elements, electrifies this confused mass, there takes place a sort of sudden organization, a spontaneous generation. This incoherence becomes cohesion, this noise becomes a voice, and these thousands of men crowded together soon form but a single animal, a wild beast without a name, which marches to its goal with an irresistible finality.“29

With his conception of the transmission of beliefs and desires through imitation and suggestion, Tarde – as well as other mass psychologists like Le Bon and Sighele – is connecting to contemporaneous discourses on transmission in relation to electricity and magnetism, hypnosis and contagion30. At the same time, the irrationality that appears in masses and seems to be constituent for them, is encapsulated and made operationalizable. Gamper points out that early cinema theory also ties into this knowledge. Perception, hypnosis and transmission become the mode of operation in these theories and explain how an audience becomes a community. In this way, the early theories on the emergence of masses gradually turned into a knowledge about the guidance of masses31.

Regarding our analysis of Facebook as a medium for a space of co-presence and overlapping publics, it is decisive that Tarde developed a notion of publics that is related to the circulation of beliefs and opinions through media. Tarde understood the public as the distributed mass of physically separated individuals. Nevertheless, these form a collective at a mental level. Suggestion – the mutual influencing of individuals among one another – results in a common opinion that unites the collective. In the concept of publics, the shared presence of individuals is replaced by circulation via media. Here, the newspaper as a ‘mass press’ is a crucial medium of remote influence, as it distributes opinions and allows them to circulate32.

Early discourse on the interaction of media, masses and public is thus marked by three figures of thought. What is fundamental is the notion of transmission, which establishes a connection between individuals and orders them into a commonality. From this follows a notion of perception, where the transmission event is set in motion by common perception (which can be apparatively controlled as needed). At the same time, Tarde’s concept of the public includes a concept of communication, or at least of distribution and circulation, in which the formation of a common opinion is inherent.

What is important for considering the Facebook Open Graph as a space of co-presence is how mass psychology ideas of media continue to be passed on. As Gamper summarizes: ‘The world of the “publics” was thus the world of first mechanical, then thermodynamic and, finally, electromagnetic transmission media, which were the actual carriers of the suggestive exchange in these social units.’33 Christina Bartz investigated how television discourses from the 1950s/60s tie into mass psychology theories, so that television first appears discursively as a mass medium. Consequently, the mass medium theory of television correlates the distributed mass with the perception-hypnosis discourse of mass theory and early cinema discourses. Television is thus proposed as a medium of perception, rather than of communication and can become a simulator of presence34. We will argue that this media conception migrates into the Facebook Open Graph as a phantasmagorical undercurrent, enabling the camouflage of micro-publics as spaces of co-presence.

Tarde’s thinking about masses, sociology and publics starts with the relation between individuals conceptualized as imitation of thought, beliefs and desires. Interestingly enough, he does not develop this perspective in opposition to social statistics. Rather, as Bruno Latour and Vincent Lépinay (2009) argue in their recent attempt to revive his Psychologie Économique (1902)35, Tarde seeks to reform statistics by extending its realm into the domain of psychology. As Tarde himself remarked early on: “In looking over the work of statisticians, it is most important to remember that the things which are under calculation are essentially subjective qualities, desires and beliefs, and that very often the acts which they enumerate, although equal in number, give expression to very different weights among these things.”36

Looming behind Tarde’s reformative zeal is an epistemological shift: social statistics according to Quetelet did not consider the individual important in order to gain knowledge about a population in terms of the mean value. Being confronted with mass phenomena which can hardly be explained by mean values and normal distributions, such as mass panics and mass politics, it seems that the need for a different perspective arose: With his focus on repetition through imitation, Tarde plays a decisive role in establishing a bottom-up perspective, with the aim of turning statistics into a tool for observing the emergence of social facts out of psychological forces and their mode of circulation.

What is decisive for this kind of statistical observation is that the interweaving of micro-processes is presumed to form structures – it is no longer characteristics of actors that are queried, but rather the processes in effect between them. Tarde thereby laid the theoretical groundwork for the development of sociometry by Jacob Moreno and the subsequent rise of network analysis37 Here, too, it is not attributes such as age or gender that are of interest, but instead it is a matter of identifying certain patterns in the relations between the actors. Sociometry and network analysis, however, radicalize this perspective. Moreno positioned sociometry explicitly as a counter-proposal to the application of macro-sociological categories38. Also methodologically, Moreno’s development of the sociogram as a means of visualizing groups structures introduced forms of reduction, formalization and calculability that ran counter to those employed in statistics. Tarde’s approach, on the other hand, held the promise of resolving the tensions between the perspectives, or at least to provide a common theoretical framework where the two diverging perspectives could meet.

Returning to Open Graph, we can now notice that the two intersecting, yet different, ways of ordering masses we identified earlier – statistics and network analysis – can be traced back to two distinct epistemic constellations, which, from a historical perspective, entered into a competition with one another: Whereas statistics in Quetelets sense was a search for determinist macro-forces, Tarde and network theory were rather interested in how interacting micro-forces resulted in the emergence of structure.

Apart from Tarde’s work, there was one other contemporary discursive space, where the two epistemic constellations met: like Quetelet before them, Tarde and later, Moreno, took recourse to physical terminology, using it to refer to analogies between physics and society. What is instructive is the different character of these analogies: whereas for Quetelet the analogy with physics was intended to prove the universality of laws in the social field, Tarde and Moreno were concerned with describing interactions that exist beyond those connections between actors that can be directly observed.

The terminology from physics thus provided a kind of common language or translatability between the two different epistemic constellations, even if they were used for different purposes: the search for a global natural law on the one hand and the search for emergent structures on the other. Yet the physical terminology also has another, less epistemic and more normative aspect: similar to the way Quetelet wanted ‘social physics’ to be understood as a foundation for governing, Moreno’s sociometry also contains the aspect of control. Despite all the differences in the objectives, physical metaphors, thus, seem to be coupled with the idea of making it possible to explain, or at least calculate forces, thus being able to influence and shape them.

When network theory and statistics now have moved into (media-) technical infrastructure as the foundation for automatized data analysis, phantasms of natural laws, emergence and controllability are also attached to it. At the same time, this is accompanied by the contradictions inherent in the fields of knowledge, as a kind of undercurrent.

Infrastructural Oscillation
We examined how the idea of alleviating tensions between different epistemological traditions emerged around the turn of century in the work of Gabriel Tarde. On this basis, we can now entangle our historical sketch and relate it to our discussion of Facebook as a mass medium. Regarding Open Graph as part of Facebook’s technical infrastructure, the crucial point is that data is adjusted – addressed, labeled, tracked and stored – according to the needs of statistics and network analysis. At the same time, this technological setting carries with it the two complementary epistemological traditions as undercurrents. This includes both the different evocations of natural laws and physics, the phantasm of rendering emerging structures observable and controllable and the notion of simulated presence in the tradition of mass media discourses.

Following this line of reasoning, the entanglement of epistemic traditions in Open Graph combines the physical laws of determinist macro-force on the one hand and a multitude of interacting micro-forces on the other correlating with the top-down approach of statistics and the bottom-up perspective of network analysis. The contradictions between the approaches serve as a motor for a constant switching of modes. Whenever the moment of reduction becomes too obvious, the perspective tips in the other direction in a kind of compensatory gesture, suggesting that it is possible fill the respective ‘blind spot’ by means of a methodological adjustment. The dynamics of these tipping movements result in a permanent ‘urge to fill in’: getting more addresses, more categories and more data traffic finally promises lossless scalability.39. In this sense the table can serve as an instrument of insight, because it identifies acute knowledge gaps and desiderata directly through missing entries.’ See Markus Krajewski, “In Formation. Aufstieg und Fall der Tabelle als Paradigma der Datenverarbeitung,” in Nach Feierabend. Daten, ed. David Gugerli et al. (Zürich/Berlin: Diaphanes, 2007), 45. Own translation.] The overlapping of analytical procedures enabled by and built into the technical infrastructure increases the number of lacunae and desiderata on the side of the website owners exponentially. This results in a permanent situation of insufficiency, which concretely calls for filling up with data and metadata on the part of the users.

We have argued that the medial promise for the website owners is loss-free scalable observability and controlability, whereas the medial promise for the users is the creation of spaces of co-presence – although it is more an overlapping of micro-publics. Looking at network analytical approaches from a media perspective, we identified concepts of transmission, perception and communication as central figures of thoughts, which also migrate into the mass media discourse concerning television.

To put it in terms of infrastructure: Facebook Open Graph re-orders this mixture of transmission, perception and communication. Here, the historical genealogies inscribed into the infrastructure resurface as effects on the level of usage: communication and transmission have rigidified into a technical apparatus and produce a simulation of presence. Speaking infrastructurally (and thus not necessarily in terms of the interface), this creates a void begging to be filled in – somewhat like an empty stage for the users to enter and act upon.

In/Coherent Subjectivities
Reconsidering the dominant use form of the Open Graph – sharing one’s own perceptions via the ‘Like’ button – at first glance, this seems like an actualization of this prompt in practice. As a last step, however, we want to argue that this practice does not necessarily follow a dictate of the technical infrastructure. Following Adelmann and Winkler40 and Adelmann41, the relation between use and infrastructure can be grasped in a more differentiated way. Introducing the concept of the ‘double subject’, these authors argue that constellations of digital media are not prone to an unambiguous development of subject positions. Whereas a postmodern critique of the subject might portray the digital realm as a welcome support in the erosion of the autonomous subject, forms of the coherent subject capable of agency do predominate, for example, in computer games.

Adelmann and Winkler explain these ‘paradoxical’ phenomena by noting a reciprocal relationship between the real and the imaginary level. The subject is able to compensate her/his real fragmentation and distribution (e.g., in the form of many single units of information stored in databases) with an imaginary coherence, so that they continue to experience themselves as capable of agency. What is decisive is, therefore, specifically the interplay of both levels: ‘Metaphorically speaking, the utopian re-production of the bourgeois subject requires the real stage of the communication and distribution of the subject in databases.’42

Corresponding tendencies can also be identified in the case of group membership on Facebook: the real fragmentation of the subject, which, in the Open Graph, only exists in the form of disparate database entries, is compensated with the imagination of a group that shares common perceptions and opinions. When Adelmann/Winkler speak of a pleasure-gain, which results from alternating between dispensing with agency and capability of agency, this precisely delineates the characteristic feature of Facebook as a medium of collectivization: in the mode of fragmentation, the transmission event is given ‘free rein’, so to speak, the subject finds itself in the midst of a cascade of impressions, where openness, unlimitedness, perhaps, even freedom, are suggested. These impressions are ordered in the context of the group: first of all, one’s own News Feed can serve as a kind of filter, showing what must be seen, because the group (and the News Feed algorithm) finds it ‘relevant’. Secondly, web contents are already evaluated as to whether they can generate attention in the group. Thirdly, adding to the News Feed, showing one’s own finds can be regarded as a kind of short chain of action at the level of self-presentation and attention: commentaries from ‘friends’ on one’s ‘likes’ confirm the selection, devote attention and enable the crystallization of a common opinion. The circulating of one’s own ‘likes’, and the reactions to them, mirrors the individual as a perceiving, opining and acting subject. Moments of causal effectiveness appear, the impression of co-presence is created, and the decentered, fragmented subject is given ‘something to stand on’ again, in a sense. In this way, the space of co-presence of a commonly perceiving group supports the notion of a non-fragmented subject capable of agency, one which has control over its own perception and actions.

Seen it this way, the ‘urge to fill in’ is not exclusively forced onto users via an infrastructure which embodies the power of the backend users, but it (also) emerges out of the practices and the media promises of the usage of Facebook. In this sense, the concept of the double subject goes beyond approaches that consider social media as part of a history of loss, in which the postmodern potential of digital media was squandered43. In addition, the interlocking of the real and the imaginary level emphasized here can be grasped as an addition to approaches that emphasize a ‘desire for dividuality’44 from a philosophical-anthropological view, or internalized demands for self-query, in the sense of technologies of the self45.

New Masses?
Our exploration of the Open Graph raises the question of the referent of a concept of ‘masses’, which has already been preconfigured by media, but is reflected in media at the same time. The matrix we employed to structure our investigation – with the distinction between backend and frontend on the one hand and the descriptive, analytical and historical perspective on the other – has provided a range of clues in this respect. It has made it clear that the Open Graph is not principally a new phenomenon; it appears instead to be a provisional climax of sociological querying practices, which are equipped with an extremely expanded space of observation, made permanent and cast in a media infrastructure.

Nevertheless, a new understanding of ‘masses’ seems to have crystallized in these constellations. The interplay between the technical infrastructure and use practices, to which new forms of a phantasmagorical surplus inhere on both sides, have especially contributed to this new understanding. The case of infrastructure involves the idea of being able to make emergence actually comprehensible on the basis of technical procedures, thus being able to overcome the reductive moment of conventional methodological approaches. The case of use, on the other hand, involves the imagination of a space of co-presence through ‘sharing’ contents and the subsequent opinions. As we have argued, this imagined space of co-presence can be distinguished both from the socio-technical production of micro-publics and the mass per se.

In both cases, phantasmagorical surpluses are coupled with alternating between two different modes, through which a process of continual compensatory ‘filling in’ is set in motion. This was expressed at the analysis level in an increasingly dense sequence of method adjustments, leading into a layering of originally contradictory forms of knowledge and orders. Each of these adjustments engenders new lacunae that are to be filled in, thus concretely contributing to enriching Open Graph with the most diverse kinds of data. The meagerness, or reduction, inherent to these layerings, despite the quantities of data at work, is obscured at the level of user interfaces by an elaborate staging. However, this does not necessarily mean that all users simply ‘fall for’ this staging. Instead, micro-publics provide a space of agency, which is characterized by the alternation between dissociation and coherence. However, this does not necessarily mean that all users simply ‘fall for’ this staging. Instead, micro-publics provide a space of agency, which is characterized by the alternation between dissociation and coherence. Here, ‘filling in’ provides compensation for the fragmentation of the subject by imagining connectedness within a group. As the discussion of the ‘double subject’ has shown, this form of compensation can be understood as contributing to ones own agency as a coherent subject.

Our investigation suggests that the production of lacunae and the phantasmagoric surpluses that fuel these dynamics are to be considered as some of the decisive media characteristics of social media. Paradoxically, however, this view is systematically blocked by current discourses on Big Data exclusively preoccupied with the expansion of massive data storage and mining. In dealing with mass, as both the descriptive, analytical and historical angle of our exploration have suggested, reduction and inclusion as well as connecting and sharing are interrelated in much more intricate ways.


We would like to thank the editors and two anonymous reviewers for valuable comments. We are also grateful for the suggestions and ideas we developed in discussions at the 2nd International DFG Symposium of Media Studies “Social Media – New Crowds?”, Lueneburg, 2nd – 4th of February 2012, where we presented an earlier version of the paper. The text was translated by Aileen Derieg.


Adelmann, Ralf. “’There is no correct way to use the system.’ Datenbanklogiken und mediale Formen.” in Sortieren, Sammeln, Suchen, Spielen. Die Datenbank als mediale Praxis, edited by Stefan Böhme, Rolf R. F. Nohr and Serjoscha Wiemer, 253-267. Berlin/Münster: LIT Verlag, 2012.
Adelmann, Ralf, and Hartmut Winkler. “Kurze Ketten. Handeln und Subjektkonstitution in Computerspielen.” Ästhetik & Kommunikation 148 (2010): 99-107.
Albrecht, Steffen. “Knoten im Netzwerk.” in Handbuch Netzwerkforschung, edited by Christian Stegbauer and Roger Häußling, 125-134. Wiesbaden: VS Verlag für Sozialwissenschaften, 2010.
Ang, Ien. Desperately seeking the audience. London, New York: Routledge, 1991.
Bartz, Christina. “Subliminal Masses. The Knowledge of Social Control.” Soziale Systeme 2 (2003): 285-297.
Bilton, Nick. “Price of Facebook Privacy? Start Clicking.“ New York Times, May 12, 2010. Accessed December 3, 2013,
Bondy, J.A., and U.S.R. Murty. Graph theory. London, New York: Springer, 2008.
boyd, danah “Social Network Sites as Networked Publics. Affordances, Dynamics, and Implications.” in A Networked Self. Identity, Community and Culture on Social Network Sites, edited by Zizi Papacharissi, 39-58. New York: Routledge, 2011.
boyd, danah, and Nicole N. B. Ellison “Sociality through Social Network Sites.” in The Oxford Handbook of Internet Studies, edited by William H. Dutton, 151-172. Oxford: Oxford University Press, 2013.
Bucher, Taina. “Want to be on the top? Algorithmic power and the threat of invisibility on Facebook.” New Media & Society 7 (2012): 1164-1180.
Coté, Mark, and Jennifer Pybus. “Learning to Immaterial Labour 2.0. MySpace and Social Networks.” ephemera 1 (2007): 88-106.
Creswell, Max. “Formal Semantics.” in The Blackwell guide to the philosophy of language, edited by Michael, Devitt and Richard Hanley, 131-146. Malden, Mass.: Blackwell Pub., 2006.
Desrosières, Alain. The Politics of Large Numbers. A History of Statistical Reasoning. Cambridge, Mass.: Harvard University Press, 1998.
Dollhausen, Karin, and Josef Wehner. “Virtuelle Gruppen – Integration durch Netzkommunikation? Gesellschafts- und medientheoretische Oberlegungen.” in Virtuelle Gruppen. Charakteristika und Problemdimensionen, edited by Udo Thiedecke, 68-87. Wiesbaden: Westdeutscher Verlag, 2000.
Facebook. “Extended Permissions – Facebook Developers.” Accessed December 2, 2013.
Facebook. “Insights for Websites. Product Guide.” Last modified March 2011,
Facebook. “How News Feed Works.” Accessed December 2, 2013,
Facebook. “Open Graph Action Types – Facebook Developers.” Accessed December 2, 2013,
Facebook. “Page – Facebook Developers.” Accessed December 2, 2013,
Freeman, Linton C. The development of social network analysis. A Study in the Sociology of Science. Vancouver, BC; North Charleston, S.C.: Empirical Press; BookSurge, 2004.
Gamper, Michael. Masse lesen, Masse schreiben. Eine Diskurs- und Imaginationsgeschichte der Menschenmenge 1765-1930. München: Fink, 2007.
Gamper, Michael. “Charisma, Hypnose, Nachahmung. Massenpsychologie und Medientheorie.” in Trancemedien und neue Medien um 1900. Ein anderer Blick auf die Moderne, edited by Marcus Hahn and Erhard Schüttpelz, 351-373. Bielefeld: Transcript, 2009.
Gerlitz, Carolin and Anne Helmond. “Hit, Link, Like and Share. Organizing the social and the fabric of the web in a Like economy.” Last modified January 26, 2011, accessed December 2, 2013,
Gross, Jonathan L., and Jay Yellen. Graph theory and its applications. Boca Raton, Fla.: CRC Press, 1999.
Hacking, Ian. “How Should We Do the History of Statistics?” in The Foucault Effect. Studies in Governmentality, edited by Graham Burchell, Colin Gordon and Peter Miller, 181-195. Chicago: University of Chicago Press, 1991.
Halpin Harry, Mischa Tuffield, Daniel Appelquist, Dan Brickley, Melvin Carvahlo, Renato Iannella, Alexandre Passant, Christine Perey, and Henry Story. „A Standards-based, Open and Privacy-aware Social Web.“ Last modified December 6, 2010, accessed August 29, 2014,
Harré, Rom. “Philosophical aspects of the micro-macro problem.” in Advances in Social Theory and Methodology. Toward an Integration of Micro- and Macro-Sociologies, edited by Karin D. Knorr-Cetina and Aaron V. Cicourel, 139-160. Boston, London, Henley: Routledge & Kegan Paul, 1981.
Hui, Yuk, and Harry Halpin. “Collective Individuation: The Future of the Social Web.” in Unlike Us Reader. Social Media Monopolies and Their Alternatives, edited by Geert Lovink and Miriam Rasch, 103-116. Amsterdam: Institute of Network Cultures, 2013.
Ito, Mizuko, Sonja Baumer, Matteo Bitani, danah boyd, Rachel Cody, Becky Herr, Heather A. Horst, Patricia G. Lange, Dilan Mahendran, Katynka Martinez, C.J. Pacoe, Dan Perke, Laura Robinson, Christo Sims, and Lisa Tripp. Hanging out, messing around, and geeking out. Kids living and learning with new media. Cambridge, Mass.: MIT Press, 2010.
Katz, Elihu. “Rediscovering Gabriel Tarde.” Political Communication 3 (2006): 263-270.
Krajewski, Markus. “In Formation. Aufstieg und Fall der Tabelle als Paradigma der Datenverarbeitung.” in Nach Feierabend. Daten, edited by David Gugerli, Michael Hagner, Michael Hampe, Barbara Orland, Philipp Sarasin and Jakob Tanner, Zürich/Berlin: Diaphanes, 2007.
Latour, Bruno, and Vincent A. Lepinay. The Science of Passionate Interests. An Introduction to Gabriel Tarde’s economic anthropology. Chicago: Prickly Paradigm Press, 2009.
Link, Jürgen. “From the ‘Power of the Norm’ to ‘Flexible Normalism’: Considerations after Foucault.” Cultural Critique 57 (2004): 14-32.
Link, Jürgen. Versuch über den Normalismus. Wie Normalität produziert wird. Göttingen: Vandenhoeck & Ruprecht, 2006.
Lovink, Geert. “A World Beyond Facebook: Introduction to the Unlike Us Reader.” in Unlike Us Reader. Social Media Monopolies and Their Alternatives, edited by Geert Lovink and Miriam Rasch, 9-15. Amsterdam: Institute of Network Cultures, 2013.
Mayer, Katja. “On the Sociometry of Search Engines. A Historical Review of Methods.” in Deep Search. The Politics of Search beyond Google, edited by Felix Stalder and Konrad Becker, 54-72. Innsbruck: Studienverlag, 2009.
Peters, John D. “Satan and Savior: Mass Communication in Progressive Thought.” Critical Studies in Mass Communication 3 (1989): 247-263.
Quetelét, Adolphe. Sur l’homme et le développement de ses facultés, ou essai de physique sociale. Paris: Bachelier, 1835.
Raunig, Gerald. “Desiring Dividuality.” Paper presented at the presentation of Open #19: Beyond Privacy: New Perspectives on the Public and Private Domains, Berlin, June 12, 2010.
Sennett, Richard. The Corrosion of Character. The Personal Consequences of Work in the New Capitalism. New York: Norton, 1998.
Simmel, Georg. Soziologie. Untersuchungen über die Formen der Vergesellschaftung. München: Duncker & Humblot, 1922.
Tarde, Gabriel. “Empirical Bases of Sociological Theory,” in Gabriel Tarde On Communication and Social Influence. Selected Papers, edited by Terry N. Clark, 277-294. Chicago; London: The University of Chicago Press, 1969.
Tarde, Gabriel. Penal philosophy. Boston: Little, Brown & Co., 1912.
Tarde, Gabriel. Psychologie économique. Paris: F. Alcan, 1902.
Thovex, Christophe, and Francky Trichet. “Semantic social networks analysis. Towards a sociophysical knowledge analysis.” Social Network Analysis and Mining 1 (2013): 35-49.


Irina Kaldrack is a postdoctoral researcher at the Digital Cultures Research Lab, Leuphana University Lueneburg. Before that, she was Postdoc at at ‘eikones – NCCR Iconic Criticism‘ in Basel and at the Research Training Group ,Automatisms‘ at University of Paderborn. Her research interests are: mediality of technical media, methodologies of media studies, media history, history of knowledge of human movement and cultural history of mathematics. Irina holds a diploma in mathematics and received her PhD in Cultural Studies from Humboldt-University Berlin. Pertinent publications: ed. with Bu­blitz, Röhle, Ze­man ‘Au­to­ma­tis­men – Selbst-Tech­no­lo­gi­en’, 2013; ‘Ge­hen in der Da­ten­bank – Der BML­wal­ker’. in Böhme, Nohr, Wie­mer (ed.) ‘Die Da­ten­bank als me­dia­le Pra­xis’, 2012; ‘Ima­gi­nier­te Wirk­sam­keit. Zwi­schen Per­for­mance und Be­we­gungs­er­ken­nung’, 2011.

Theo Röhle is a postdoctoral researcher at the Institute for Media Research, Braunschweig University of Art. Before that, he held a postdoc position at the Research Training Group ‘Automatisms’, University of Paderborn. His research focuses on digital knowledge and epistemology (Search, Digital Humanities, Social Media), surveillance and control, software history, and business games. He received his PhD in media culture from the University of Hamburg and his MA in Communication Studies from Stockholm University. Pertinent publications: ed. with Oliver Leistert ‘Generation Facebook. Über das Leben im Social Net’, 2011; ‘Der Google-Komplex. Über Macht im Zeitalter des Internets’, 2010; ‘Desperately seeking the consumer: Personalized search engines and the commercial exploitation of user data’ in First Monday, 2007.

  1. Harry Halpin et al., “A Standards-based, Open and Privacy-aware Social Web,” (2010), last modified December 6, 2010, accessed August 29, 2014, On the conflictual relationship between the Open Graph protocol and other approaches in the field of the semantic web, see also Yuk Hui and Harry Halpin, “Collective Individuation: The Future of the Social Web,” in Unlike Us Reader. Social Media Monopolies and Their Alternatives, ed. Geert Lovink and Miriam Rasch (Amsterdam: Institute of Network Cultures, 2013), 103-116.
  2. Carolin Gerlitz and Anne Helmond, “Hit, Link, Like and Share. Organizing the social and the fabric of the web in a Like economy”, last modified January 26, 2011, accessed December 2, 2013,
  3. “Insights for Websites. Product Guide,” Facebook, last modified March 2011,
  4. “Page – Facebook Developers”, Facebook, accessed December 2, 2013,
  5. “Extended Permissions – Facebook Developers”, Facebook, accessed December 2, 2013,
  6. “Open Graph Action Types – Facebook Developers”, Facebook, accessed December 2, 2013,
  7. Obviously, ‘semantic’ can only be understood here in the sense of formal semantics, cf. Max Creswell, “Formal Semantics,” in The Blackwell guide to the philosophy of language, ed. Michael Devitt and Richard Hanley (Malden, Mass.: Blackwell Pub., 2006), 131-146.
  8. A crucial aspect of the visibilities produced is the privacy settings, in which users determine what exactly is visible for their ‘friends’. For an overview of the various possibilities for settings, it is worth taking a look at a chart from the New York Times: Nick Bilton, “Price of Facebook Privacy? Start Clicking,“ New York Times, May 12, 2010, accessed December 3, 2013,
  9. “How News Feed Works”, Facebook, accessed December 2, 2013, A detailed critical investigation of the EdgeRank algorithm in relation to the question of visibility is provided in Taina Bucher, “Want to be on the top? Algorithmic power and the threat of invisibility on Facebook,” New Media & Society 7 (2012): 1164-1180.
  10. Mizuko Ito et al. Hanging out, messing around, and geeking out. Kids living and learning with new media (Cambridge, Mass.: MIT Press, 2010), 38.
  11. danah boyd and Nicole B. Ellison “Sociality through Social Network Sites,” in The Oxford Handbook of Internet Studies, ed. William W. H. Dutton (Oxford: Oxford University Press, 2013), 163.
  12. Since the procedures actually applied by Facebook and the app providers are not published, this is an area of speculation, but it is well founded by taking current discourses in engineering sciences in this field into consideration.
  13. A metric or distance function assigns a real number ≥ 0 to each pair of elements from the set considered. A metric must fulfill the following conditions: the distance between two elements is exactly 0, if the elements are identical; the distance between x and y corresponds to the distance between y and x; and triangle inequality applies.
  14. See, for example, the textbooks J. A. Bondy and U.S.R. Murty, Graph Theory (London: Springer, 2008); Jonathan L. Gross and Jay Yellen Graph Theory and its Applications (Boca Raton, Fla.: CRC Press, 1999).
  15. In figurative terms, a matrix is a tabular arrangement of symbols in rows and columns. Matrices can be added and subtracted in places and multiplied using a certain procedure. In an adjacency matrix the columns and rows designate the respective nodes of a graph. If two nodes ‘i’ and ‘j’ are adjacent, a ‘1’ is entered in the ‘i’-row and the ‘j’-column, otherwise an 0.
  16. Steffen Albrecht, “Knoten im Netzwerk,” in Handbuch Netzwerkforschung, ed. Christian Stegbauer and Roger Häußling (Wiesbaden: VS Verlag für Sozialwissenschaften, 2010), 129 (own translation).
  17. It makes sense that only sub-networks are considered here, not Facebook in its entirety.
  18. The definitions of objects and actions that are part of Open Graph could provide a further level of analysis, as recent approaches in ‘Semantic Social Network Analysis’ suggest. One deficiency of conventional social network analysis is seen here in the reduction to nodes and edges, which can result, for example, in the influence of certain nodes being underestimated, because the measures used only take into consideration the number of connections. To solve this dilemma, the authors propose a statistical analysis of statements that can be attributed to the nodes in a network. If agreements can be identified in these statements, which can also be attributed to a certain category, then this can be taken as an indication of belonging to a group, which cannot be read from the analysis of the network constellations. See Christophe Thovex and Francky Trichet “Semantic social networks analysis. Towards a sociophysical knowledge analysis,” Social Network Analysis and Mining 1 (2013): 35-49.
  19. This figure of thought is also discussed in the social sciences. An overview of social theory and media theory ideas about virtual groups is provided by Karin Dollhausen and Josef Wehner “Virtuelle Gruppen – Integration durch Netzkommunikation? Gesellschafts- und medientheoretische Oberlegungen,” in Virtuelle Gruppen. Charakteristika und Problemdimensionen, ed. Udo Thiedecke (Wiesbaden: Westdeutscher Verlag, 2000), 68-87. They consider the relation of social ties, communication and technology from their social sciences perspective against the horizon of the changing constitution of society. Consequently, modern forms of (as)sociation are accompanied by an increase in individualization and the dissolution of traditional and firmly established social forms. This results simultaneously in the advantage and the problem that compulsions or orientation are dissolved. In this perspective, the technical infrastructure of social networks and their possibilities for interaction correspond to the needs of ‘flexible human beings’ (cf. Richard Sennett, The Corrosion of Character. The Personal Consequences of Work in the New Capitalism (New York: Norton, 1998)) to form casual, fluid, but sometimes stable social forms, on the basis of which they can situate themselves through identity. In other words, Open Graph could be seen as a medium of fluid social groups that exchange experiences and opinions and enable orientation for the individuals within them, while maintaining flexibility and mobility at the same time. However, these groups are no longer to be understood as ‘social groups’ in the narrower sense, which are distinguished by diffuse, yet close, and especially stable, reciprocal relationships.
  20. danah boyd, “Social Network Sites as Networked Publics. Affordances, Dynamics, and Implications,” in A Networked Self. Identity, Community and Culture on Social Network Sites, ed. Zizi Papacharissi (New York: Routledge, 2011), 42.
  21. Adolphe Quetelet, Sur l’homme et le développement de ses facultés, ou essai de physique sociale (Paris: Bachelier, 1835).
  22. Quetelet, Sur l’homme, 408, translated from a quote in Jürgen Link, Versuch über den Normalismus. Wie Normalität produziert wird (Göttingen: Vandenhoeck & Ruprecht, 2006), 197.
  23. Link, Versuch über den Normalismus, 195. Own translation, emphasis in original. For an English introduction to Link’s concept of normalism, see Jürgen Link, “From the ‘Power of the Norm’ to ‘Flexible Normalism’: Considerations after Foucault,” Cultural Critique 57 (2004): 14-32. The question of why political steering would even be needed, if nature strives for mean values of itself, was discussed in the contemporary debates under the heading ‘statistical determinism’. On this, Ian Hacking explains that the aims of the reformatory movements related to a more indirect level of steering: ‘to reorganize “boundary conditions” under which a population was governed by statistical laws’, cf. Ian Hacking, “How Should We Do the History of Statistics?” in The Foucault Effect. Studies in Governmentality, ed. Graham Burchell, Colin Gordon and Peter Miller (Chicago: University of Chicago Press, 1991), 188.
  24. This is true even for the more differentiated approach of Francis Galton. Instead of positing the mean value as an ideal, he divides the normal distribution in parts and identifies these parts with sub-groups of the observed population. For example, in his eugenically motivated investigation, Galton takes recourse to a classification from Booth for determining the ‘civic worth’, which is dissipated on the x-axis of the normal distribution; cf. Alain Desrosières, The Politics of Large Numbers. A History of Statistical Reasoning (Cambridge, Mass.: Harvard University Press, 1998), 114f.
  25. Rom Harré, “Philosophical aspects of the micro-macro problem,” in Advances in social theory and methodology. Toward an integration of micro- and macro-sociologies, ed. Karin D. Knorr-Cetina and Aaron V. Cicourel (Boston, London, Henley: Routledge & Kegan Paul, 1981), 139-160.
  26. In reference to user surveys in the field of television, see also Ien Ang, Desperately seeking the audience (London; New York: Routledge, 1991).
  27. Another genealogical path involving a notion of society made up of reciprocal effects among actors would lead via Georg Simmel (1858-1918). According to him, the crystallization of stable social constellations could first be explained from an analysis of these reciprocal effects: ‘For unity in the empirical sense is nothing other than an interplay of elements: an organic body is a unit, because its organs are in a closer reciprocal exchange than with any external being […
  28. The second central concept is innovation or invention, which marks the moment of the new in society.
  29. Gabriel Tarde, Penal philosophy (Boston: Little, Brown & Co., 1912), 323.
  30. Michael Gamper, Masse lesen, Masse schreiben. Eine Diskurs- und Imaginationsgeschichte der Menschenmenge 1765-1930 (München: Fink, 2007), 412; Michael Gamper, “Charisma, Hypnose, Nachahmung. Massenpsychologie und Medientheorie,” in Trancemedien und neue Medien um 1900. Ein anderer Blick auf die Moderne, ed. Marcus Hahn and Erhard Schüttpelz (Bielefeld: Transcript, 2009), 354f.
  31. Gamper, “Charisma, Hypnose, Nachahmung,” 366.
  32. On the history of the impact of Tarde’s ideas on the newspaper as part of the ‘social organism’ on the US American mass communication discourse, see also John D. Peters, “Satan and Savior: Mass Communication in Progressive Thought,” Critical Studies in Mass Communication 3 (1989): 247-263.
  33. Gamper, “Charisma, Hypnose, Nachahmung,” 368.
  34. Christina Bartz, “Subliminal Masses. The Knowledge of Social Control,” Soziale Systeme 2 (2003): 285-297.
  35. Bruno Latour and Vincent A. Lepinay, The Science of Passionate Interests. An Introduction to Gabriel Tarde’s economic anthropology (Chicago: Prickly Paradigm Press, 2009); Gabriel Tarde, Psychologie économique (Paris: F. Alcan, 1902).
  36. Gabriel Tarde “Empirical Bases of Sociological Theory,” in Gabriel Tarde on Communication and Social Influence; selected papers, ed. Terry N. Clark (Chicago: University of Chicago Press, 1969), 214.
  37. Linton C. Freeman, The development of social network analysis. A Study in the Sociology of Science (Vancouver, BC; North Charleston, S.C.: Empirical Press; BookSurge, 2004); Elihu Katz, “Rediscovering Gabriel Tarde,” Political Communication 3 (2006): 263-270.
  38. Katja Mayer, “On the Sociometry of Search Engines. A Historical Review of Methods,” in Deep Search. The Politics of Search beyond Google, eds. Felix Stalder and Konrad Becker (Innsbruck: Studienverlag , 2009), 58.
  39. Markus Krajewski notices a similar phenomenon of an ‘urge to fill in’ with regard to the historical development of the table as an instrument of inquiry: ‘The grid of a table is consequently oriented to the completeness of what is to be covered. Cleverly selected categorial subdivisions into rows and columns allow, not only the classification of each existing element, but also the anticipation of lacunae, which result from a strict combinatorics of the categories […
  40. Ralf Adelmann and Hartmut Winkler “Kurze Ketten. Handeln und Subjektkonstitution in Computerspielen,” Ästhetik & Kommunikation 148 (2010): 99-107.
  41. Ralf Adelmann, “’There is no correct way to use the system.’ Datenbanklogiken und mediale Formen,” in Sortieren, Sammeln, Suchen, Spielen. Die Datenbank als mediale Praxis, ed. Stefan Böhme, Rolf F. Nohr and Serjoscha Wiemer (Berlin/Münster: LIT Verlag, 2012), 253-267.
  42. Adelmann, “’There is no correct way to use the system.’,“ 255.
  43. This impression appears e.g. with Lovink, who laments that social media are ‘not postmodern machines but straightforward modernist products of the 1990s wave of digital globalization turned mass culture’, cf. Geert Lovink, “A World Beyond Facebook: Introduction to the Unlike Us Reader,” in Unlike Us Reader. Social Media Monopolies and Their Alternatives, ed. Geert Lovink and Miriam Rasch (Amsterdam: Institute of Network Cultures, 2013), 12.
  44. Gerald Raunig, “Desiring Dividuality“ (paper presented at the presentation of Open #19: Beyond Privacy: New Perspectives on the Public and Private Domains, Berlin, June 12, 2010).
  45. cf. Mark Coté and Jennifer Pybus, “Learning to Immaterial Labour 2.0. MySpace and Social Networks,” ephemera 1 (2007): 88-106.
Series Navigation<< No SQL: The Shifting Materialities of Database TechnologyWhat Do Metrics Want? How Quantification Prescribes Social Interaction on Facebook >>

  • Find articles by author