The Evolution of Facebook's News Feed

(1)

Master New Media and Digital Culture

The Evolution of Facebook’s

News Feed

On the Transformation of a Digital Object That Is Shaping Billions ________________________________________

New Media and Digital Culture MA Thesis 2015-2016 Ryan Jay de Ruiter | 10253254

Abstract

The Facebook News Feed is an important element of the Facebook platform. Billions of people communicate and consume information through the News Feed. As it is an influential element that uses the power of feeds to affect the way in which users consume information, I argue that it is necessary to understand how the News Feed is orchestrated, and how it has been orchestrated in the past. In this thesis I analyze this object from a historical perspective, in order to learn how the News Feed and its content has changed through the years. I analyze the News Feed’s user interface and underlying algorithm to create a genealogy of the News Feed from the moment it was instated until 2016. I draw from the field of software studies and I use Taina Bucher’s notion of the method of technography to map in which ways and for which causes the News Feed has changed, and what this means for the orchestration of content on the News Feed. The genealogy I have established enables the reader to understand how the News Feed was orchestrated at a specific moment in time. Among the important shifts that have been analyzed are the variety of content sorting filters, and the gradual shift from text-based messaged to visually oriented content. This paper is useful for anyone who is interested in the history, dynamics, and underlying software of Facebook News Feed.

Keywords

(2)

Date: June 24, 2016

Supervisor: mw. dr. A. Helmond

(3)

1 Introduction

“Facebook is always Facebook at a specific point in time, such as in 2004, 2008 and 2012. Which is why it is so important to know about the history

of Facebook.”

― Niels Brügger, “A Brief History of Facebook as a Media Text.” 2015.

Every month, over 1.5 billion people log in to their Facebook accounts (Statista). Users who log into Facebook will see their homepage, which shows the News Feed. A lot of information and news updates are digested by Facebook users through this News Feed. In 2014, Mark Zuckerberg explained how he wanted to turn the News Feed into the ‘perfect personalized newspaper’ (Kim n. pag.). In this essay I will discuss how the information that users consume through the News Feed has differently organized over the years, and what effects the algorithm alterations have had on this organization. The algorithm controls what Facebook users will see in their News Feeds and in which order. Through the consumption of this content the platform bears responsibility in the shaping of people’s worldview. Increasingly so, since for more and more people Facebook serves as a source of news (Barthel et al. n. pag.).

Nowadays, whenever you log in to Facebook, the algorithm browses and aggregates everything recently posted by each of your friends, all the pages you have liked, and all updates from the groups you are in. Depending on how many friends you have and the number of pages you have liked, the number of posts that the algorithm will scan for the average Facebook user runs up to several thousands (Oremus n. pag.). This number might sound big, but besides status updates, every time someone comments, someone is tagged, or when someone shares a posts or likes something it also counts as a post. All of these posts are then ranked through the algorithm, which leads to a specific order of most relevant posts, and in this process thousands of other posts are made invisible for the users. This is not necessarily a bad thing, on the contrary, it is an important process. If the filtering did not exist, Facebook users would be overwhelmed by posts, which would make it more difficult to find posts that have gained a lot attention, or posts of specific people close to you. The News Feed’s filtering algorithm tries to prevent important information from being drowned out. How the algorithm and the filtering exactly works is a company

(5)

secret, but many algorithmic changes have been disclosed by Facebook in the business and news sections of their website, and through other media outlets. Through researching the News Feed from a historical perspective I want to investigate how the News Feed has changed since the beginning of its existence, both visually and algorithmically. This allows me to analyze how Facebook has influenced the content arrangement and action potential of the News Feed, which I see as an important transformation when the News Feed is seen as a source of news and information for billions of people. The research question I want to answer through this is as follows: How has the evolution of the Facebook News Feed changed the way its content is orchestrated? Answering this question is meaningful in that the News Feed enables human communication and creativity, and the system directs and sometimes limits users’ options, overall constructing the users (Grosser n. pag.), and the “ways of seeing, knowing, and doing in the world [the software] ostensibly pertains to” (Fuller, Behind the Blip 6). Before performing the software analysis, it is essential to establish a framework of research that has been done in this area before, in order to position this research within existing fields and to take a critical look at earlier findings.

2 Software studies

I position this research within the field of software studies. I will explain why, starting with an exploration of what software studies encompasses and how it originated. An important role in this emerging field is played by Lev Manovich, who offers the first insight into software studies in his book The Language of

New Media (2001). Manovich calls for a move from media studies to something

that can be called software studies (65), which would offer the possibility to examine the materiality of new media. The concept of software studies is addressed by Manovich for the first time in the following passage:

New media calls for a new stage in media theory whose beginnings can be traced back to the revolutionary works of Harold Innis in the 1950s and Marshall McLuhan in the 1960s. To understand the logic of new media, we need to turn to computer science. It is there that we may expect to find the new terms, categories, and operations that characterize media that became programmable. From media studies, we move to something

(6)

that can be called “software studies” - from media theory to software theory. (65)

Through software studies Manovich wants to place new media within a larger historical perspective (36). He analyzed the relationships between old and new media, and marked five qualities of that are key trends in the age of new media proposed which show fundamental tendencies of the new media culture (44). These five ‘principles of new media’, according to Manovich, are what is shaping the development of new media over time; numerical representation, modularity, automation, variability, and transcoding. In his later work Manovich moved beyond these five principles, which I will address after analyzing one of his five principles that I find very interesting for this research. The automation principle that Manovich describes is applicable to my object of study, the News Feed, because algorithms in software create automated functions, through which according to Manovich “human intentionality can be removed from the creative process” (53). The News Feed itself was created to bring content to the user, instead of users having to surf to individual profiles (Facebook Help Centre1_{) –} an important moment in Facebook’s history, which will be later explored and analyzed in-depth. This development removed human intentionality from the social process. It was still possible to surf to specific individual profiles, but the standard way of browsing had changed. In another way however, the News Feed interface can be seen as both increasing and decreasing human intentionality, depending on the view. The interface does have various interactive and social options one can use, such as liking, commenting or for example hiding posts, which makes that through interaction the News Feed algorithm will receive data through which it can calculate what content to present the user. Now, human intentionality is dependent on users’ awareness of the algorithm. One might think that a user’s actions will be taken into account by the algorithm, which might encourage users to like posts they want to see more of in the future, and hide those they do not want to see anymore. By marking content, users would be training the algorithm, creating the basis for the algorithm’s decision-making (Rieder 11). In this case, the software can be seen as increasing human intentionality. However, if a user is acting on the platform without this understanding, the algorithm would still act on the data it receives and show posts accordingly, but this time without the user being aware, thus it would be 1 The content of the Facebook Help Centre has changed over the years. The source I refer to in this instance is the Facebook Help Centre on September 14, 2006:

(7)

further reducing human intentionality. Taina Bucher has researched these different situations, and she has shown that “a plethora of ordinary affects” are generated through these different experiences (Algorithmic Imaginary 13). In a way Matthew Fuller’s argument complements this, as he notes in his book that naturalized technologisation, indicating smooth and imperceptible software structuration, lessens users’ ability to understand changes and engage with them (Software Studies 4). Manovich has argued that we live in a world of permanent change where software is always in flux (Software Takes Command 2) and where Facebook updates its code daily (1). This would according to Fuller not benefit the comprehension of the software. Constant changes constitute less engagement, and ultimately less awareness of the News Feed algorithm overall, which might be seen as an unfavorable observation, as this cultural software is used by billions of people (Company Info).

Since Manovich’s classification, the field of software studies has expanded and this key author in software studies has proposed new ideas and adaptations of his own definitions. Manovich has written his second book on software titled

Software Takes Command, in 2013. Already in 2011 though, in the introduction

of what was then his upcoming book, Manovich altered his previous formulation of software studies:

Reading this statement today, I feel some adjustments are in order. It positions computer science as a kind of absolute truth, a given which can explain to us how culture works in software society. But computer science is itself part of culture. Therefore, I think that Software Studies has to investigate both the role of software in forming contemporary culture, and cultural, social, and economic forces that are shaping development of software itself. (6)

This new description encompasses, according to Taina Bucher, “the need to shed light on an object long overlooked within media and cultural studies” (Programmed Sociality 26). I can accordingly identify my object of study in this notion of what software studies should be about. Other notions of software studies include Dodge, Kitchin, and Zook’s understanding that software studies seeks to understand code beyond the technical, offering cultural and theoretical insights as to how code, through algorithms and data, encapsulates the world (1285). The technical algorithmic augmentation of the News Feed algorithm, which I will identify in this research, is examined precisely for the reason Dodge et al. articulate. Facebook’s sorting algorithm holds power that operates

(8)

through the News Feed and plays out in the lives of individual users. In investigating this I also answer the ‘pressing need’ pointed out by David Beer (2009) to “explore with some detail this vision of power through the algorithm” (999), because the software constrains and enables, formulates hierarchies and shapes the things people encounter (1000). This research serves to help fill the knowledge gap identified by Langlois et al. (2009), namely the need to explore specific elements and actors that are mobilized in specific ways, that shape sociality online (416). Still, now that different definitions, notions and aims of software studies have been shown, there is not yet described how software studies has looked at algorithms. It is now necessary to explore how software studies research previously has approached algorithms.

2.1 Software studies approaches to studying algorithms

While the term interface might be used more broadly than the term algorithm, this term is arguably more abstract, for algorithms are invisible layers. The Oxford dictionary defines the term algorithm as “a process or set of rules to be followed in calculations or other problem-solving operations, especially by a computer.” Tarleton Gillespie, however, sees the term ‘algorithm’ as something more complex than the previous definition. He argues that the term has many meanings, depending on its use, and that in the case of Facebook’s algorithm it can stand for a “complex sociotechnical assemblage” (6):

Facebook's algorithm really means “Facebook,” and Facebook really means the people, things, priorities, infrastructures, aims, and discourses that animate them. But it may also be a political economic conflation: this is Facebook acting through its algorithm, intervening in an algorithmic way, building a business precisely on its ability to construct complex models of social/expressive activity, train on an immense corpus of data, tune countless parameters, and reach formalized goals extremely efficiently. Facebook as a company often behaves algorithmically. (Gillespie 6)

Gillespie argues that saying ‘Facebook’s algorithm’ might thus be a way to assign accountability to the company (6), which indicates that the News Feed algorithm might also carry a complex political meaning. Besides this analysis, multiple other researchers have had an interest in the News Feed algorithm for a variety of reasons; Birkbak and Carlsen recently wrote about how the actively constructing algorithm is justifying its own behavior, by analyzing how various actors allocated, reviewed and justified the algorithm (n. pag.). Zeynep Tufekci wrote about the implications and the harms that the algorithm can bring in

(9)

various case studies within Facebook and Twitter. The ‘power of the algorithm’ is seen as something that deservers scrutiny by more researchers. Taina Bucher attested that it lies within the algorithms capacities to “define certain regimes of visibility by assigning more weight to edges that generate a higher degree of user participation” (Want to Be on Top? 150). She also wrote about the effects the algorithm can have on users by interviewing people who talked about it on Twitter (The Algorithmic Imaginary), and Martin Berg looked at how interpersonal relationships are affected by the algorithmic structures. He did this through analyzing self-reflexive diaries that participants had written. Berg notes that this is a feasible way to study social effects, whereas a lot of other ethnographic analysis on Facebook is very hard to execute outside of one’s own Facebook network due to the platform’s privacy settings (n. pag.). The focus of these research papers make a lot of sense, since a lot of people are not aware of the News Feed curation (Eslami et al. 1). This leads to the consequences that Facebook users attribute wrongly “the composition of their feeds to the habits or intent of their friends and family” (9). Eslami et al. also carry out their research through interviewing users. All these user studies provide answers through samples of users, which are hard to generalize. Although, the same might be said about software studies research: Tufekci argues that analyzing how content is arranged through social media algorithms is an interesting field. There are no simple, correct answers, because the output is generated through subjective decision makers (206). Similarly, Taina Bucher writes that as with any new established field there is no ‘right way’ and there are no established means as how to engage in software studies (Programmed Sociality 27). My interest is captured by the reciprocity between the user interface and the invisible algorithmic structure. This interest accords with Benjamin Grosser’s argument that in order to understand the effects of software, both code and interface need to be examined to see what the software makes possible. He argues that software studies should exist of a multifaceted analysis of both the interface that users see, as well as the underlying algorithms and data structures that together are the elements that make up software (n. pag.). This is also stressed by R. Stuart Geiger, who notes that analyzing algorithms separate from their environment is like ‘talking about the law without talking about the courts or the police” (348). Complementing this statement, Rob Kitchin has written that when approaching algorithms from a software studies perspective, they should be framed “within a wider socio-technical assemblage” to study the powers and

(10)

politics, and “the way [algorithms] reshape particular domains” (7). Following from these statements, one might say that studying Facebook’s News Feed might mean the scope has to be as broad as the entire platform, or maybe even further. However, such a scope is not desirable for this research because I do not want to take away the specific focus on the News Feed’s evolution. Furthermore, even though the News Feed is not a piece of isolated software, it does consist of both a visual interface and an algorithmic structure, which is why this research can follow Grosser’s dichotomous structure for analysis.

In addition to Grosser’s argument for both code and interface analysis in order to investigate the possibilities software offers, Matthew Fuller extends the question of what software makes possible by referring to, as he puts it,

conditions of possibility (Software Studies 2). By this he meant that software,

which may seem like a neutral actor, actually creates options and possibilities, but simultaneously constrains its users by limiting options and directing the ways users should act. Social relations are made invisible through this ostensible neutrality of software (3), which is why a closer look is necessary. On this matter, Manovich also discusses how software can be seen as invisible, as hidden in multiple ways. The algorithm is of course directly hidden as in not accessible for anyone outside of the company, but also the interface and its possibilities become naturalized through usage of the platform. The News Feed is completely taken for granted, becomes invisible, and hides in plain sight (n. pag.). It is important to take this into account in this research, because through the conditions of possibility the News Feed is overall constructing the ways users interact with the software, and it is doing so invisibly. Furthermore, even if the code would be available, it could be “too big to analyze and discuss its ‘politics’ in any meaningful way” (Manovich in Szilak n. pag.). Taina Bucher has devised a way, not to analyze the code itself, but to understand “the performance of algorithms through the ways in which they are being articulated, experienced and contested in the public domain” (The Algorithmic Imaginary 11). She notes that algorithms are more than “abstract computational processes,” but the presence of a Facebook algorithm has the power to “shape social life to various degrees” (11), through which the algorithm can be studied. By stating this, Bucher complements Fuller, who says that a closer look is necessary because social relations are made invisible (Software Studies 3).

(11)

While for Grosser software studies entails investigating code and interface, Manovich argues in an interview with Michael Connor that this is just the surface of software studies and that this field for him means asking broad questions about our world and everything that runs on software. He declares that interface and code investigation could only be of value if it helps answering these questions and creates an understanding of how software shapes us as individuals and as society (n. pag.). In an article titled Software is the Message (2014) Manovich gives an example of such a question which intersects with this research: “How do particular algorithms used by Facebook to decide what updates from our friends to show in our News Feed shape how we understand the world?” (80). This research can be seen as complementing Manovich’s overarching question through investigating the changes from the developers’ side. It is interesting to analyze the evolution of the Feed from this perspective because it enables me to make claims about the intentions of the company, what seems important to Facebook, and how Facebook tries to accommodate its users with refined and expanded software tools. Through the News Feed, Facebook is growing the means for people to communicate and express themselves. In other words, software, as the means for communication, has turned the medium into the new McLuhanian ‘message’ (81). Furthermore, Manovich writes that the properties of digital media, or content, are defined by the application software through which the interaction happens (Manovich 2013, 152). The News Feed, that acts as the container of media, is software developed by a large group of people and the software is constantly being refined, expanded, and altered. These continuous mutations might have multiple motives, such as the company’s ideology, the business model, or the social and cultural shaping forces. José van Dijck similarly talks about ‘the platform’s architecture’ – interface, design, and code – as the result of the owner’s effort to steer users in a certain direction (4). The software alterations should be seen as deeply social, says Manovich (149). The changes are not the consequence of individual minds, but they are the consequence of motives from software developers that market their product to millions of users in Facebook’s case, which creates the need to analyze how this software is orchestrated and how this process developed over time. In the next chapter I will explain how I intend pursue a historical approach, but first I need to address obstacles around researching the News Feed.

(12)

3 Methodological Framework

The first problem that needs to solved methodologically, is how the News Feed can be demarcated. The most logical solution for defining the News Feed is also the approach I have chosen to proceed with: looking at how Facebook itself demarcates the News Feed. However, the complication which then arises, because of using a historical approach, is that the definition given by Facebook in the Help Centre also changes throughout time. This is a factor that complicates the demarcation, but also benefits the research, as the Facebook Help Centre can also be seen as an object for analysis itself, because it provides questions and answers related to the News Feed which change throughout time. A second obstacle is that a choice has to be made between the analysis of the News Feed’s web interface, mobile interface, and the app-based interface, or any combination of those. Can I still analyze the News Feed if I neglect one or the other platform? For a variety of reasons, I have chosen to research the web interface exclusively. For one, the mobile platform did arrive a year later than the desktop version (Hewitt 2007), and the first app-based version came out another year later in 2008 (Hewitt 2008), but more importantly, the desktop interface receives wider attention from media outlets as well as from the company itself. There is a lot of documentation, visual content and discussion about Facebook’s web interface. Also, since the desktop provides a larger screen, more features are available in the desktop version than in the mobile interface. This is useful for analysis, but it also creates another problem in the context of setting boundaries for the research object, in that it is difficult to keep the different elements of Facebook’s platform apart, because the interface links a lot of parts together. For example, in my own personal experience I have seen short surveys from Facebook in my News Feed, in which I am asked which posts I deem more relevant (see figure 1). As written when I explained Grosser’s notion of software, a multifaceted analysis of interface and algorithms is necessary, because it together makes up the software. A Facebook survey can thus not be categorized in either one or the other section, which is why I will not create separate timelines for the user interface and the algorithm. Instead will analyze the News Feed as a whole, from its launch in 2006 until today. Because of this chronological analysis, instead of an analysis based on different elements, there will be more clarity which aids the understanding of the timeframe of

(13)

changes on interface and algorithmic level and seeing them in context of each other.

Figure 1: Survey question on Facebook, April 2014.

I will first describe how Facebook was organized in its early years, from the start in 2004 until the moment Facebook opened its platform for a broader public, because this is when many new digital objects were added to the website, including the News Feed (Philips n. pag.). It is important to involve this period to understand how the platform worked, what features it contained, and how users interacted with the software. Then I will analyze in-depth the moment the News Feed was implemented. This means I will use a historical approach to interpret the changes made by Facebook over time, to investigate the intrinsic reasoning for this changes. I draw on the approach from software studies scholar Taina Bucher, used in her dissertation Programmed Sociality, which she names

technography. The term is used for “a way of describing and observing

technology, on its own material-discursive terms, in order to examine the interplay between a diverse set of actors (both human and nonhuman)” (69). The term is partly derived from the method called ethnography, but there is a clear and distinct difference between the two methods. Where ethnography has its focus on “documenting people’s beliefs and practices from the people’s own perspectives” (Riemer 205), technography has a different perspective. Technographic research does not seek to understand culture through the meanings that people derive from it, but the approach asks what the software itself is suggestive of, as Bucher puts it (Programmed Sociality 69). The method is not confined to software and interface analysis, but it also includes the analysis of technical documents, web articles discussing the News Feed, conferences and developer talks, interviews, and the developer pages within the website itself. Through these external sources, this method also takes into account the perspectives of human actors who are well-informed about the

(14)

software, and those making the software. Bucher does not explicitly draw a line between technography and ethnography: she sees technography rather “as a mode of inquiry into the meanings embedded in the mechanics of technology that make use of ethnographic methods” (69). I agree with this, since I found that in Facebook’s case, software related changes, have a lot to do with user feedback. While the software itself has the focus, certain ethnographic analyses cannot be completely ruled out. The method of technography does not necessarily constitute a historical approach, but Bucher draws from Manovich’s wordings that software studies has to investigate the role of software in forming present-day culture (26), which he relates to a historical viewpoint. As Matthew Kirschenbaum puts it, meticulous documentary research about the past of software is also a pre-condition for a ‘theory of the present’ (150). This theory of the present is a term that Manovich uses in his first book on software studies, by which he aims at the need for a genealogy for computer media, for a way to theorize about the present, instead of speculating about the future (The

Language of New Media 33). The term genealogy as used here stems from

Michel Foucault, for whom it entailed “a method of writing critical history” (Garland 372). Through using historical data and materials – in this case recordings of the News Feed – a genealogy can help reassess current values. David Garland notes that a genealogy “intents to problematize the present by revealing the power relations upon which it depends and the contingent processes that have brought it into being” (372). Garland’s interpretation of a genealogy strengthens the reasoning for the use of this method in this research, because the ambition of this writing is to critically inform and provide understanding about how the News Feed has come to be what it is today. I think that a genealogy, through the means of historical documentary research from a technographic perspective, can serve to answer the question that is the subject of this thesis.

I will use a technographic approach to first and foremost study the digital objects and the way the News Feed worked, as well as technical specifications on developer pages of Facebook. The difficulty in analyzing the platform itself through time, lies in the fact that no earlier versions of Facebook can be accessed. The analysis of Facebook in 2006, for example, can only be based on what is documented by others. The Internet Archive is a website that focuses on documenting existing versions of websites, which was founded in 1996 (About).

(15)

This archive has documented a lot of webpages, dating back to 1996, or in Facebook’s case, August 6, 20052_{. Web pages are captures in a way that looks} like screenshots, but often the links on the captured web page still work, and lead you to other pages. However, using this way to examine Facebook’s user interface is problematic, since one needs to be signed in on the platform with one’s Facebook account, in order to reach further than the landing page. When you then sign in, you will see the current version of Facebook, instead of the captured one. This means it is impossible for the Wayback Machine to capture the News Feed interface at a certain moment in time. Although the Internet Archive is of no use for investigating what the interface looked like at a certain time, the archive can be used by comparing different versions of Facebook’s open subdomains, which are documented by the Internet Archive. These subdomains include the Facebook Help Centre and Facebook Newsroom, which are useful resources that inform their users about the News Feed and its changes. The material I can retrieve through the Internet Archive from these pages will be the main data for analysis of how the News Feed previously was structured and organized. Furthermore, web articles discussing Facebook changes often refer to blogs written by Facebook, and the links to these blogs often do not direct to the corresponding page anymore. For this problem the Internet Archive can also be consulted. Another part, for which I do not need the Internet Archive’s aid, is Facebook’s own documentation in its Newsroom pages of the website. The Newsroom holds the category ‘News Feed FYI’, which consists of an archive dating back all the way to News Feed’s inception in 2006. A lot of changes have been documented and justified here, both on design level and algorithmic. Facebook’s so-called f8-developer conferences offer illustrations and demonstrations of these changes as well. Also, part of the images and data that will be subject to analysis will be from third party sources that have documented the platform, as long as there is a timestamp available as to confirm the timeframe of the data. After this, as a part of the technographical method established by Taina Bucher, I also analyze media reports, such as news articles and blog posts from sources like TechCrunch, Mashable and The

2 This is the first capture of what was then called Thefacebook. The domain name has been captured earlier as well, dating back to December 12, 1998, but at that time the domain was registered to a company called AboutFace. Also, a lot of captures from the timeframe between AboutFace and Facebook show error pages, which is the first usable capture dates from August 6, 2005.

(16)

Guardian, to get a grasp of how Facebook manifested its new product to various

media outlets. Google’s advanced search options allows for queries to be looked up within specific timeframes, which aids the exploration of specific dates and years. Also, in an effort to provide simplified and accessible information about the News Feed, Facebook launched a user-friendly informative and interactive page called newsfeed.fb.com. This is another specific page from which I can derive how Facebook illustrates its News Feed to the users. Through forming a genealogy, I can look at the News Feed from a historical perspective and use this as an investigative method to analyze the web interface and algorithmic developments of the News Feed, which helps creating a better understanding of how these important objects have evolved and for what purposes this has happened. This research might be of significance for the bigger picture, in the context of our digital society, but I do not want to focus on making large claims around (micro)political practices. Such questions belong more to the field of cultural studies, whereas software studies positions its research around the computational basis of our digital world. Research on a social platform so ingrained in our contemporary society is bound to give insight into, and make claims about our digital society. Before I can start my analysis however, I need to create an understanding and demarcation of important concepts of this study. This is why first I will provide a short history on the term ‘feed’, after which I will demarcate the physical element of my research object, News Feed’s interface.

4 Background of core concepts

4.1 ‘Feed’

The term ‘feed’ takes an important role in this research, which is why should be researched how the feed as a term, and as part of software culture came into being. This should be considered before performing the main research, because as opposed to Facebook’s News Feed, the concept ‘feed’ has a relatively long history. Building a historical pathway of the etymology and the use of the term will aid the understanding of the most banal component of the News Feed: its name. The Oxford dictionary refers to the word feed in the field of computing as a “facility for notifying the user of a blog or other frequently updated website that new content has been added.” The term was originally only used in the

(17)

context of nourishment, as in food and feeding, and this operation can also be seen in the new use of the term. A user is being fed new information, and it is precisely this being fed, instead of actively having to look for content that is the reason the RSS protocol was developed. This RSS protocol, for which the acronym stands for both Rich Site Summary and Really Simple Syndication (Wusteman 404), was the first type of feed that came into being in the digital world, which was first released in 1999 by Netscape and later taken over by Dave Winer’s company UserLand Software (Winer n. pag.; Weltevrede et al. 133). This protocol was then used as a tool by publishers to spread their content by removing the user’s need to manually browse for content. This content could be news items, blog post, sports content, or any other information element (404). Through subscribing to an RSS-feed, one would automatically receive the new or updated content in a so-called aggregator or feed reader. What a user then would receive would be called a news feed, or a blog feed, and so on. RSS might be reason the term feed has reached the large public, because websites have promoted their feeds, for example through the orange RSS logo, to get people to subscribe to them. Google queries for the term ‘news feed’ show that gradually, from late 2003 and onwards, the term is being used. The first time ‘news feed’ is mentioned on a website that is relatively known, it is used as a synonym for RSS feed3_{, in 2005 (ESPN.com). Then in 2006, the same year that} Facebook implemented the News Feed, Wikipedia articles came into existence for the term ‘web feed’, to refer to ‘news feed’. Software company Sharethrough, which works with in-feed ads, recognized a number of historical moments that have been important for the evolution of the feed (The Rise of the

Feed). For example, in 2001 Apple laid the foundation for a new way of

navigation on mobile devices, by introducing the thumb scroll on the first iPod. This new way of navigating through a device’s content, instead of using buttons, can be seen as the start of a new type of action, namely scrolling, which nowadays is an important action when interacting with feeds. Furthermore, in 3 The first time the term ‘RSS feed’ was used on the web is hard to figure out, as Google returns for example news websites that offer RSS feeds, with time stamps that date before the actual invention of RSS. My investigation points to the first uses of ‘RSS feed’ on developer blogs and web pages explaining how RSS operates and how it can be used in early 2000. It is interesting that the use of the term ‘feed’ is not explained, argued or reasoned on these pages. It must also be noted that it is not always possible to find out when these pages were updated, and when the exact moment was ‘RSS feed’ was first typed on these kind of pages. Sources:

http://search.cpan.org/~sburke/XML-RSS-SimpleGen-11.11/lib/XML/RSS/SimpleGen.pm http://web.resource.org/rss/1.0/spec

(18)

2004 the website Digg was launched, which displayed user-submitted content on the homepage of the website. The website started out as an experiment, where content put forward by users could be voted up (‘digging’) or down, which created a constantly updated list of popular web content. Digg itself described its website in 2005 as “a technology news website that combines social bookmarking, blogging, RSS, and non-hierarchical editorial control” (FAQ, Internet Archive). Digg can be seen as a curation site that uses feeds, although these feeds are put forward and ranked by the site’s users. The website also gave the ability to save search phrases as RSS feeds, to be read in a feed reader, thus it combined existing RSS uses with a new online display of feeds where Digg serves as a feed-based curation site. The concept that Digg used turned out to be working, because Sharethrough notes that in 2013 the top 21 news sites and the five most used social media platforms all work with web feeds to deliver content. Some platforms’ feeds deliver content purely based on user-selected creators, just like a feed reader does. An example of such a platform is Instagram. However, Facebook’s News Feed also delivers content like Sponsored Stories or ‘Stories you might like’, which have not been selected by the feed’s users (more on this in Chapter 4). Another difference between the feeds of the two platforms is that Instagram only shows content created on the platform, whilst Facebook’s News Feed contains content from both internal and external sources. Facebook and Instagram are part of the reason that Sharethrough’s statistics show that the Internet’s most visited websites indicate a serious shift from static web pages to constantly updated content. However, while this is the case right now, the reception was drastically different when Facebook launched the News Feed in 2006, which is the starting point of the News Feed analysis. Before I can analyze the software, I need to explain what Facebook’s user interface is, and how it has been studied before.

4.2 User interface

Researching the News Feed interface asks for a more specific frame of what this element is. What is the News Feed interface, and for that matter, how do I define ‘interface’? A logical term to talk about this element seems to be the graphical

user interface (GUI), which is still often used regarding the ‘front-end’ of

software, where interaction through text, objects and images takes place (Stalder 2012). The News Feed, then, is part of this user interface, which allows for user input and content consumption. To analyze this part of the interface, I

(19)

will specifically focus on what Langlois and Elmer call digital objects, a concept for those objects that allow for communicative acts on interface level (5). A digital object can for instance be a ‘like’ button, as well as a comment or any other kind of text. The authors do not see the visible interface as a boundary, digital objects can also be formatting specifications, or algorithms (11). The objects are multifaceted and can portray a lot of different things. However, when I talk about digital objects within the interface, I specifically mean the visible objects, such as buttons, hyperlinks, Facebook posts, comments, or Reactions (which I will address later). Langlois and Elmer have identified three layers within digital objects (11); at the first level these objects are media objects, which can be analyzed for content or form. Deeper analysis, about how an object influences or shapes communication, happens at the second level, that of the

network object. A digital object links several networks together. For example,

when a Facebook user interacts with another user through a friend request, the friend networks are linked together. In another instance, clicking on a Facebook posts links the specific user to another website. Not all these connections are happening in the visible interface, the visible objects also relate to the algorithmic software through what Langlois et al. see as ‘double articulation’: when communicative acts that take place on one level also create articulations on another level (Langlois and Elmer 5). In the case of the News Feed, communicative acts from users create more user data for the algorithm to work with, in order to calculate which new content to present. Langlois et al. see double articulation as something that has to be critically looked at, because according to them it is a corporate social media strategy that on the surface seemingly offers free tools, while in the back-end acts are translated into valuable data (6). This critique is often heard, most commonly through the phrase “If you are not paying for it, you are the product.” Certainly, Facebook’s business model relies on its users, which is why the News Feed’s evolution is also strongly tied to this business model. The analysis will show if News Feed adaptations also have double articulations and how these adaptations are put forward and presented through time. The third layer of digital objects that Langlois and Elmer recognize is that of the phatic object. A digital object is phatic in that an object enacts presence. Users are made visible and positioned according to their actions. Phatic objects establish relations among users through the user’s actions. These different layers of the digital objects within the News Feed will prove useful for my analysis, for example I can historically

(20)

look at different versions of media objects, and see how in different times digital objects act as network and phatic objects.

Several authors have focused on analyzing Facebook’s user interface. Benjamin Grosser has taken the interface of Facebook, and within it the News Feed, as an object of study. He developed the Facebook Demetricator in order to explore the social performance of metrics in the interface. In his research he analyzes how various affordances within the timeline perform, and how they are designed in a specific way in order to draw upon “our deeply ingrained ‘desire for more” (n. pag). For example, he points out that when a timestamp on a post gets older, they get less specific and increasingly hide their placement, as to conceal the temporal rearrangement that the algorithm performs (n. pag.). Other authors have looked at how Facebook’s interface and features evolved over time. Niels Brügger has developed a brief history of Facebook, from the beginning until 2013. This work is meant to tell Facebook’s history through a chronological representation, with the focus on how new features were developed from older media forms in a Web-historical perspective (n. pag.). Although Brügger does not use methods of technography, user studies, or algorithmic analysis, it is useful to draw from this paper how he has divided Facebook’s evolution into three phases, by comparing the changes within the platform. These phases are classified in years; ‘2004-2006’, which can be seen as the ‘educational’ one, where originally Facebook was only available for Harvard students and later other universities; then the phase ‘2006-2008’ where the site was not limited to educational institutions anymore and a lot of new features were added; and the phase ‘2008-2013’, where according to Brügger a lot less new functionalities were added in contrast to the ever growing user base. Brügger wrote his paper about the platform as a whole, so it is interesting to see how I can compare his work in relation to the evolution of the News Feed. In his dissertation, Rob Heyman also analyzes Facebook’s evolution and innovations, though with a focus on what this means for the relation between users and advertisers. Heyman focuses on how innovations are deployed in view of Facebook’s business model that is based on free use. In other words, he analyzes how Facebook’s features aggregate user generated content. Consistent with Brügger’s analysis, he first notes a phase of adding features in order to become indispensable (4), after which “increased advertising, targeting, and tracking” is introduced (5). Even though Facebook’s business model is not the focus of my

(21)

research, it should be clear that a lot of Facebook’s doings probably derives from their business strategy, which is why this underlying motive will also play a part in my analysis.

(22)

Brügger and Heyman look at Facebook as a whole, but research specifically aimed at the News Feed comes with the problem of demarcation. Where should boundaries of the News Feed be set, and where is a digital object is not part of

the News Feed anymore? The ‘feed’ or list of the News Feed itself consists of all kinds of posts; status updates, shares, photos, tags and likes are the most common, but we can also distinguish check-ins, app-generated posts, and some other types of posts (How Sharing Works). Also, the News Feed shows advertisements and other recommended content like ‘pages you might like’. Besides all these different types of posts, there are other important elements in the News Feed that users can interact with. The newly updated like-button with its ‘Reactions’, commenting and share-buttons, and also the ‘post options’ in the upper right corner of a post, giving users the options to hide or save posts, for example. All these digital objects create a problematic environment for researching the News Feed, because where does researching the News Feed stop? Should and could we research the News Feed without researching functions reaching further than the News Feed? And what about the ticker in the upper right corner? Does this count as a part of the News Feed, since it has

(23)

commonalities? Figure 2 shows some of the different elements that make up the Facebook homepage.

Figure 2: Facebook’s interface as of June 10, 2016 (frames edited).

The green frame highlights the News Feed, as it is demarcated on the Facebook Help page in June of 2016: “News Feed is a stream of updates from friends and Pages you follow and is your default home page view” (Navigating Your Home

Page). The purple frames highlight specific features that are part of the News

Feed, such as hiding, liking, sharing, and commenting on posts, as well as Sponsored Stories, all of which will be later addressed. Now that I have introduced software studies, algorithm approaches, my research methods, and an understanding of important concepts, I can begin my analysis at the moment Facebook was launched.

5 The News Feed over time

On a Wednesday, February 4th_{in 2004, Mark Zuckerberg launched what was} then called ‘Thefacebook’, a social networking site created for Harvard students where one could sign up through his Harvard email address. The site was designed to look like the pages that are often found in booklets that schools use, with names, photos, and addresses of students in them (Tabak n. pag.). In October of 2004, the team of Thefacebook enlisted the help of ad sales firm Y2M to give investors an idea of what the site was. This slideshow included an image of the interface when a user searches social network (see figure 3). This interface shows similarities with the offline ‘facebook’ used by schools, and gives insight into the original intention of the website.

(24)

Figure 3: Thefacebook, October 2004 (Marshall)

Quickly the network extended to other universities, then high schools, and in October of 2005 universities in the United Kingdom were reached (Philips n. pag.). Still though, this means that the site that by now acquired the domain Facebook.com was exclusive. This changed in September 2006, when the network started allowing users with any email address. Within the first two years of (the) Facebook, there is not yet a News Feed algorithm or News Feed interface to study. Facebook had approximately 9 million users at the moment that they were on the brink of implementing the News Feed (Kornblum n. pag.). These 9 million people were used to a static homepage, instead of a constantly updating one that is now available for the 1.6 billion users in the first quarter of 2016 (Statista). In this stage, the website could be seen as a directory, which is also how the company was described in an early presentation: “Thefacebook.com is an expanding online directory that connects students, alumni, and faculty/staff through social networks at colleges and universities” (Edwards 2015, n. pag.). In this pre-News Feed version of Facebook, a user had to find a specific friend in the friend list to browse to his or her individual profile, as was the case in other social media platforms like Myspace and Friendster. Having to perform these actions appeared to give users some sense of privacy because a user had to actively look up their profiles in order to see any updates (Leyden n. pag.). It can be argued that in the same way people had

(25)

to browse to blogs or news websites to see updates, until RSS technology removed this necessity. In the same way, Facebook’s static homepage changed into a continuously updating one because of the News Feed implementation, which is arguably the start of a whole new phase of Facebook.

5.1 Launch of the News Feed

Before Facebook opened up for everyone on September the 26th_{of 2006} (Abram), the News Feed launched on the 5th_{of September. I want to specifically} look at how Facebook marketed the product to its users, and at how the company responded to its many critics. The first time Facebook talked about its new feature, was right after the launch through the blog post “Facebook Gets a Facelift” on the 5th_{of September, 2006 (Sanghvi). It is interesting that Facebook} chose to explain the feature right after its implementation, instead of giving notice of the change beforehand. This means that while everyone was used to the directory based interface (figure 4), suddenly the homepage showed all kinds of updates from anyone in your friend list (figure 5).

(26)

Figure 4: Facebook interface 2005 (Buck).

Figure 5: Facebook’s News Feed in 2006, highlighted (Manjoo). This post stated the following:

(27)

You've probably noticed that Facebook looks different today. We've added two cool features: News Feed, which appears on your homepage, and Mini-Feed, which appears in each person's profile.

News Feed highlights what's happening in your social circles on

Facebook. It updates a personalized list of news stories throughout the day, so you'll know when Mark adds Britney Spears to his Favorites or when your crush is single again. Now, whenever you log in, you'll get the latest headlines generated by the activity of your friends and social groups. […]

News Feed and Mini-Feed are a different way of looking at the news about your friends, but they do not give out any information that wasn't already visible. Your privacy settings remain the same – the people who couldn't see your info before still can't see it now.

These features are not only different from anything we've had on Facebook before, but they're quite unlike anything you can find on the web. We hope these changes help you stay more up to date on your friends' lives. (Sanghvi)

The News Feed was introduced as a feature bringing the latest headlines, as opposed to a directory one had to manually search through. However, it can be seen as an overstatement that the News Feed is “quite unlike anything you can find on the web.” Apart from the social aspect, the News Feed basically can be seen as a feed reader, showing posts from people who the user has befriended, and from pages the user has liked, just like a feed reader is showing news from the outlets one is subscribed to. Support for the claim can however also be found, through Chris Cox, who was back then engineer at Facebook (McGirt). He outed in 2010 in David Kirkpatrick’s book the The Facebook Effect, in which was written about Facebook’s early years, what the founding idea for the News Feed was: “The Internet could help you answer a million questions, but not the most important one, the one you wake up with every day: How are the people doing that I care about?” (181). Besides this blog post, Facebook’s support page was also updated to answer questions people might have about the News Feed. Under the header ‘Features’ the sections ‘News Feed’ and ‘Profiles and Mini-Feed’ were put on top of the list (figure 6).

(28)

Figure 6: http://www.facebook.com/help.php in September 2006, highlighted (Internet Archive).

When a user clicked on News Feed on this support page, Facebook first gives a description of the new object:

News Feed highlights what's going on in and around your Facebook network by listing the latest stories about your friends on your Facebook homepage. You will only be notified of actions that you would have been able to see by clicking around the site. We also display external news articles that might be of interest to you. It's like we started delivering the mail to you instead of forcing you to pick it up on your own. (Facebook Help Centre, September 14th₂₀₀₆₎4

The first sentence of this description tells us that the News Feed lists ‘the latest stories’, which seems like the News Feed is not using any algorithmic sorting, but instead shows all content in reverse-chronological order. However, this is not

4 Source:

(29)

the case. The Facebook Help page lists the question “How does Facebook choose News Feed stories?”, for which the answer was the following:

If I could explain the algorithm, I wouldn't be writing the Help page. Basically, we use information about how you interact with your friends on the site to automatically generate stories that could be interesting or relevant to you. (Facebook Help Centre, September 14th₂₀₀₆₎5

This description shows that Facebook did not communicate openly about how the algorithm operated. What is also interesting from the description of what News Feed is, is that it accurately points out the similarity between the News Feed and the RSS protocol, which allows a feed reader to aggregate items within one piece of software. The News Feed is described with words of a similar meaning to this, “delivering mail instead of picking it up on your own.” An important difference with the existing use of RSS is that the News Feed not only shows updates from friends, but also external news articles were being displayed. Through this implementation one might already see a hint of the vision Zuckerberg stated seven years later in 2013: “We want to give everyone in the world the best personalized newspaper in the world” (Kim n. pag.).

This first version of the News Feed provided updates from friends, and gave users the ability to bookmark (save) and share these posts. This last feature indicates a specific ability that is not found in existing uses of RSS feeds: further distribution across other News Feeds. Furthermore, blue hyperlinks within the News Feed linked to specific Facebook profiles and other Facebook pages, and users could also click on photos to see the photo on a user’s profile (kkns, YouTube). This new homepage with real-time updates was a move questioned by many critics according to Ruchi Sanghvi (Focus on Feed). Tech critics asked the question why Facebook would launch a product that would cut down on page views? Instead of browsing to individual profile pages, loading more advertisements, and keeping people busy with the content, every update would show on the homepage. It was not seen before in social media and more importantly; it would negatively impact revenues because less web pages would be visited by the users (Focus on Feed). Facebook CEO Mark Zuckerberg explained in a keynote conversation with TechCrunch founder Michael Arrington that this theory was not true. In fact, page views and traffic went up 50% in the

5 Source:

https://web.archive.org/web/20121210233611/http://www.facebook.com/help/166738576 721085/

(30)

weeks after News Feed’s introduction. The new feature created a more integrate experience that brings people back the website more often (Biggs n. pag.).

5.1.1 Reception

Even though traffic went up, from the users’ point of view the News Feed implementation caused a large uproar, resulting in various online protests (Leyden; Schmidt; Arrington 2006a). Many users experienced the update as ‘creepy’, and ‘too stalker-esque’ (Leyden). Both within Facebook by creating ‘groups’ (Arrington 2006a), and outside of Facebook through protest websites hundreds of thousands showed their dismay (Leyden). The protest was formed around the experienced intrusive nature of the News Feed, and the thought that it is out of the user’s control. Every update of a user would show up on another person’s News Feed, without the ability to alter this. Questions were raised, ‘Would you trade your privacy for the attention of others?’ One of the main comments users had, was that they did not receive the ability to make this decision themselves. The other new feature that was implemented at the same time, Mini-Feed, did not create the same backlash even though the same kind of updates were shown in this feed. This is for one, because the feature was only visible on one’s wall, and probably more important, the Mini-Feed was easy to curate (Leyden) which might mean that the users felt like they did have control over this feature. Within a day after the launch of the News Feed and the Mini-Feed, Mark Zuckerberg responded to the criticism in a post on Facebook (2006a): “Calm down. Breathe. We hear you.” The message of this post was that Facebook would listen to everyone’s concerns, but that nothing had changed regarding users’ privacy. “Nothing you do is being broadcast; rather, it is being

shared with people who care about what you do – your friends” (2006a

[emphasis added]). User Interface Engineering expert Jared Spool wrote about the backlash Facebook received, and he stated the following: “you don’t just have to design the change, you have to design the process of introducing the change and ensuring it’s embraced by the users” (n. pag.). According to him, the update was the result of a groupthink and not the result of a user assessment (n. pag.). An assessment of Facebook’s users is exactly that wat mattered, according to Spool. Social media researcher danah boyd wrote in a blog on her website that she understands why Facebook implemented the News Feed, but she perceives the feature to be “a privacy trainwreck” (n. pag.). It is unhealthy and socially disruptive because data of Facebook actions is now easily

(31)

accessible, which according to boyd makes all the difference (n. pag.). Also, when data is accessible and presented, a user might be more inclined to pay attention to it, which makes that users know a lot more about their Facebook friends. This changes relationships with people, which boyd calls invasive (n. pag.).

All these critics made that two days later, Zuckerberg spoke again. He wrote a slightly longer blog post starting with the sentence “We really messed this one up” (2006b). In this post Zuckerberg wrote that the company failed at explaining the new features and at giving the users the ability to control the features. The actions that Facebook had taken in the 48 hours after the launch in order to meet the users’ wishes regarded extending people’s privacy controls. These controls were not directly implemented in the News Feed interface, but in the settings menu. These new settings gave users control as to which type of stories would go into friend’s News Feeds and also listed which type of actions would never show up in other people’s feeds6_{. The way the News Feed was received by} the public can already tell us that public opinions of the feature, however much the News Feed changed over the next ten years, have drastically shifted. So what has changed since News Feed’s inception?

5.1.2 Feedback response

Twenty days after the News Feed launch, Facebook opened up its platform for the whole world (Abram n.pag.). This means that new users could register with any email address, instead of college email addresses exclusively. Unlike with the News Feed launch, this change was communicated beforehand through various media outlets (Arrington 2006b; Kornblum; Hansell). Opening up Facebook for everyone did not directly go accompanied with any changes to the News Feed, but the newly acquired users helped Facebook receive feedback about what people want from their News Feed (Wang) Because of the conclusion that “everyone wants something a little bit different out of his or her News

6 These extended privacy controls were available in the settings of a user account. The Internet Archive unfortunately cannot capture these pages, since one has to be signed in to access this page. This keeps us in the dark about the specific enlisted actions that for example ‘would never show up in other people’s feeds.’

(32)

Feed”, Facebook brought two new changes; a News Feed Preferences page, and friend lists. The News Feed Preference page gave users the ability to tweak the content of their News Feed (figure 7).

Figure 7: News Feed Preference page as of November 16th_{2006 (Wang).}

Because of this update, users were now able to modify the quantity they wanted to see of specific types of content. The subject of content is not taken into account, but users could indicate whether they wanted to see more photos, less events, or more wall posts for example. It is interesting to note that this new update was not picked up by the online media. A Google query around the period the new settings were introduced shows that almost no outlets covered the new update, which means it was only communicated through Facebooks blog post “News Feed just got better” (Wang). Similarly, the feature ‘friend lists’ received a low amount of media coverage as well. This feature allowed users to select up to twenty friends of whom they want to see more content, and the same for friends of whom users want to see less activities. The lack of public attention for these new features is interesting, since one can assume Facebook wants to promote and encourage the use of newly developed features. This assumption is also aided by the light tone the blog post author tried to give the article, like in the following sentence: “Or maybe one of your friends is dominating your News Feed by always writing boring notes on what she ate yesterday; if you add her to the ‘Less About These Friends’ list, we'll try not to subject you to any more of her culinary ramblings” (Wang). Using these features gives users a certain power over their News Feed which was requested by a lot of users according to Wang. Still, while this feature gives the ability to alter News Feed content, it is yet unknown to anyone outside of Facebook how the News Feed selects its content.

The first time Facebook brings up the News Feed’s sorting process is in February of 2007 through a new blog post entitled “News Feed is a Robot!” (Boz). This new post explains the quantity of data the News Feed has to deal with, and assesses that the software is doing so effectively. However, there is no mention yet of an algorithm, which makes it difficult to investigate how the News Feed operated in its earliest phases. Deriving from Google, as of February 2007 the terms ‘News Feed’ and ‘algorithm’ have not been used together anywhere on the web yet. The first time the News Feed in combination with the

(33)

term ‘algorithm’ is used online, according to the Google search results, is in July of 2007, in an interview with a Facebook employee who talks about the mechanism behind the News Feed as a “ranking algorithm that uses signals based on user behavior throughout the site” (Insider Perspectives n. pag). However, no details about its workings were revealed yet. Also, this is not a quote officially stated through Facebook, but a talk with Ex-Google employee Justin Rosenstein, who discussed his recent move to Facebook. One update that Facebook published during this time where algorithms play a role, is that in August 0f 2007 the News Feed time stamps were removed (What’s New, Internet Archive). These timestamps were confusing and not useful, Facebook reported, which has to do with the non-chronological display of posts through the algorithm7_{. Finding out about the early workings of the News Feed algorithm} requires looking into Facebook’s next statements and writings involving the News Feed.

In November of 2007, a new blog post by Facebook with the title “News Feed Feedback” shows Facebook’s next alteration: feedback controls (Whitnah). Facebook wrote that the number of stories and story types has been growing, which makes it harder to select the stories that the users would find the most important. By implementing feedback controls the selecting process would be improved. Because of the new feature, a ‘thumb up’-icon and an ‘x’-icon now appear next to every post, through which users are able to show their liking, or rejection of a specific post (Figure 8).

Figure 8: News Feed thumbs up-icon and x-icon as of November 2007 (Whitnah). This is not the same as the Like button that was launched in 2009, since this feature is only meant to train the News Feed algorithm about a user’s liking or disliking of specific posts, and it does not inform users who posted the content that their posts have been liked. While the News Feed Preferences sliders that launched one year earlier gave users the ability to change the amount of specific 7 Source:

The Evolution of Facebook's News Feed