If...Then Algorithmic Power and Politics

As we will see, the news industry constitutes a place in which the emergence of computation, machine learning, and data science have wide-reaching and persistent consequences. Although computerization has long been part of the news industry, the use of algorithms as an integral part of news work is a relatively recent phenomenon in this context. As I claimed in chapter 2, algorithms do not exist in a vacuum. Nor can their power and politics simply be framed in a top-down manner, for example, as something exercising power over another actor or object. Algorithms are always built and embedded into the lived world, at the level of institutional practice, individual behavior, and human experience. In chapters 2 and 3 I referred to the notion of ontological politics to denote how reality is never a given, but rather shaped and emerging through particular interactions. The purpose of this chapter is to provide more empirical evidence for the claim that algorithms are imbued with such ontological politics.

So far in this book, I have shown that algorithms mater in a variety of ways: in their capacity to govern participation on platforms, distribute information flow, embed values in design, reflect existing societal biases and help reinforce them by means of automation and feedback loops, and in their powers to make people feel and act in specific ways. As I have argued throughout, understanding the ontological politics of algorithms cannot simply be done by looking at the materiality of code, as if formulas have a deterministic force or causal relationship to what they are supposed to compute. Regarding the notion that we need to take materiality of technology seriously, I already take this for granted.2 While algorithms can be said to possess what Jane Bennet calls “thing-power” (2004: 348), a capacity to assert themselves that cannot be reduced to human subjectivity, here I want to focus on the specific circumstances of their material-discursive enactment—or matering— in the context of the news industry.

How exactly are algorithms enacted to shape the realities of contemporary journalism, and when do algorithms come to mater in the first place? To speak of mater, “matering” (Law, 2004b) and “material-discursive practices” (Barad, 2003; Orlikowski & Scot, 2015) is to be interested in the ways in which algorithms are materialized in diferent manners, made out and carved into existence at the intersection of technology, institutional practice, and discourse. Drawing on the process-relational framework that was introduced in chapter 3, what I am interested in here are the processes through which algorithms are demarcated as an entity and made separate from other things. What an algorithm does and the impact it may have is not tied to its materiality alone, but also to the ways in which it enters discursively into accounts, and makes things happen as part of situated practices. This eventfulness of algorithms is thus inextricably linked to the question of when algorithms mater, understood as the ways in which the algorithm emerges as objects of knowledge and concern, defining and being defined by those who are touched by it. Mater and matering are key terms in this regard, describing an ongoing process of becoming, whereby diferent actors and elements are brought together and aligned in ways that shape certain realities (Barad, 2003; Mol, 2002). To understand the realities that have allowed controversies surrounding “fake news,” Facebook’s trending topic, and its censorship of an iconic photograph to flourish, we need to first understand how algorithms come to mater at the intersection of journalism, social media platforms, and institutional practice. When, then, do algorithms mater and whom do they mater for?

120

I F . . . T H E N

Algorithms in the News

Espen Egil Hansen, the editor-in-chief of Aftenposten, did not call out what he called Mark Zuckerberg’s “Californian code” because of an opposition to algorithms as part of the news dissemination process, but because he disagrees with the ways in which Facebook uses them. In an interview about the newspaper’s role in tackling Facebook on censorship, Hansen described his growing frustration with the platform’s “unprecedented power as an editor,” contrasted with an unwillingness to admit its status as a media company: “I think that one of the core tasks of the independent press . . . is to expose people to diferent views. . . . What we see now is that the press is becoming weaker.” Social media platforms, the interview suggested, have a tendency to trap users in bubbles that prevent them from reading “the other side” of stories. Hansen furthermore claims:

By far, [Facebook] is the most important carrier of news today, and it is editing at least in two ways. Every post you see on your feed is there for a reason; it's not neutral at all, it's an algorithm that decides. (Sullivan, 2016)

Hansen’s worry is that the world’s most powerful editor, the algorithm, and the people controlling it—Mark Zuckerberg and his associates—do not operate according to the same principles and professional ethics as the “independent press.” This worry concerns the power struggle over editorial responsibility, and the question of who should have a right to claim it. On the one side, we have the independent press, whose task it is to inform the public and expose people to diferent views. On the other side, we find the curated filter bubbles created by social media algorithms that relentlessly “trap users in a world of news that only afirms their beliefs” (Sullivan 2016). However, this reduces a complex issue to a false dichotomy of being either for or against the algorithm. Facebook does not connect users and provide them a platform due to an altruistic, benevolent motivation. Nor does Aftenposten and the news industry rally against Facebook because of mere anger over the removal of a photograph. And even though he partly construes such a black-and-white narrative himself, Hansen says in the interview that the problem is not the algorithm per se: “I of course use algorithms myself, and my ambitions are to use them on a much higher level than we do today.”

Programming the News

121

While algorithms in many ways have become an entrenched part of the logic and functioning of today’s news media, their status and purpose are still very much up for debate and contestation. In this chapter I look more closely at the content and circumstance of these debates by considering how key actors in the news media, such as Hansen, make sense of and negotiate algorithms in the institutional context of journalism. To understand the power and politics of algorithms in the contemporary media landscape, the argument is made that it is vital to see algorithms as neither given nor stable objects, but rather made and unmade in material-discursive practices. Sometimes algorithms are rendered important, while at other times deemed insignificant. Knowing more about processes of matering, I suggest, enables us to understand the multiple realities of algorithms, and how these relate and coexist.

Although the many recent controversies surrounding algorithms certainly call into question what algorithms do to the current informational landscape, what news is and ought to be, and what it means to have editorial responsibility, algorithms are already an integral part of journalism and journalistic practices. As is evident from much of the current literature on the intersection of technology and journalism (e.g., Anderson,  2013; Coddington,  2015; Diakopoulos,  2015; Dörr,  2016; Lewis, 2015), algorithms, robots, and data analytics are now contained in the standard repertoire of the way in which news outlets organize, produce and disseminate new content.3 However, to posit these changes and developments as new forces changing journalism is to miss the point. From the telegraph and the typewriter to databases, spreadsheets, and the personal computer, various forms of technology have for centuries been used to support and augment news work (Boczkowski, 2004; Pavlik, 2000; Weiss & Domingo, 2010). Indeed, computers have been part of journalistic practice at least since the mainframe computers of the 1960s, which also gave rise to the tradition of computer-assisted reporting (CAR). In his seminal book Precision Journalism, Philip Meyer ([1973]  2002) describes how computers and statistics would be the most efective tools by which journalists could strengthen journalistic objectivity and transform journalism more broadly.4 In order to fulfil the “real” goal of journalism, which in Meyer’s view was to reveal social injustices, journalists would have to work with statistical software packages, database programs, and spreadsheets to find stories in the data (Meyer, 2002: 86; Parasie & Dagiral, 2013).

122

I F . . . T H E N

Today, these ideas are no longer mere predictions of a distant future. The computerization and algorithmization of news and news work have increasingly become the norm. While many of the core tools identified by Meyers remain the same, the scale and extent to which computational tools and processes are incorporated into news work are what’s novel.

If the digitization of news once meant making analogue content digitally available, it now signifies a series of processes that go far beyond simple online newspaper distribution. The digitization and computerization of news can be seen at every stage of the production cycle: how journalists use computational tools to collect and produce news stories, the way in which systematized tagging and metadata practices are used to archive content, through the implementation of digital paywalls to ensure revenue and access to valuable user data, and in the algorithmic presentation and curation of news.

Let’s at this point recall the common understanding of an algorithm as a set of instructions for solving a problem. What are the problems that algorithms are supposed to help solve in this context? One obvious problem for journalism is the “challenge to the economic viability of newspapers triggered by the digital revolution in publishing and news distribution” (Alexander, 2015: 10). Sociologist Jefrey Alexander observes how many “leading journalistic institutions in the West have experienced great economic upheaval, cuting staf, and undergoing deep, often radical reorganization—in eforts to meet the digital challenge” (2015: 10). These cuts and reorganizations have also been felt acutely in the Nordic context, where many newspapers, including Aftenposten, are currently experiencing a dramatic reduction in staf—primarily those working with the print edition. As Alexander notes, the cultural mantra “information wants to be free” that came along with the Internet has become an economic headache for journalism today.5 Declining advertising and the rise of blogs and social media platforms as fierce market competitors have contributed significantly to the economic crisis in journalism in recent years. Rasmus Kleis Nielsen notes how over “half of all digital advertising revenues globally go to Google, Facebook” (2016: 86), and the stigma that was once atached to paywalls continues to be an obstacle for their economic and cultural feasibility (Alexander et al., 2016).

The rise of computational journalism, here broadly understood as news entangled with algorithmic practices, comes at a time of crisis not only for the newspaper industry and journalism in general, but also in terms of broader social and economic unrest. Not unlike the rhetoric surrounding the rise of CAR in the 1960s, the discourses of computational and algorithmic journalism are characterized by a continuum of utopian and dystopian visions. Algorithms, data science, and everything that comes with them are distinguished by a great degree of techno-optimism and promise, but also marked by a notable skepticism and worries about machines taking over jobs. On the optimistic side, computational journalism is talked about as a way of upgrading and equipping journalism for the demands of the 21st century (Hamilton & Turner, 2009; Cohen et al., 2011). As Lewis and Usher note, “many argue that computational journalism will both lead to beter investigative journalism and create new forms of engagement with audiences” (2013: 606). The sentiment here is that journalism needs “to adapt to technological and economic changes in order to survive” (Peters & Broersma, 2013: 2). More than just as a celebration of technical possibilities, many see these digital transformations as a changing condition of news to which one simply needs to adapt in order to stay in the game.

Programming the News

123

But as with all technical change and innovation, discussions surrounding the computerization of newsrooms and the digitization of news have typically been accompanied by worries about the changing status quo. The emergence of social media, for example, fostered discussions around the potential replacement of journalists and other news workers by bloggers (Lowrey,  2006; Matheson,  2004; Rosen, 2005) or non-professionals engaging in citizen journalism (Goode, 2009; Rosenberry & St. John, 2010). These debates and worries have been replaced by similar discussions and worries about the rise of the so-called robot journalism (Clerwall,  2014). Companies such as Narrative Science or Automated Insights, which train computers to write news stories using automated algorithmic processes, have spurred much discussion about the sustainability of journalist as an enduring profession and the vulnerability of the human work force in the age of the purportedly “smart machine” (Carlson,  2015; van Dalen,  2012). Without commiting to news robots being either a threat or a salvation, it is clear from the now burgeoning literature on what Anderson calls the “shaggy, emerging beast” of computational journalism (2013: 1017; see also Rodgers, 2015: 13) that algorithms are here to stay. The question, then, is not whether algorithms play an important part in journalism and news work, but in what way this role is playing out in practice, how it is accounted for and made relevant, and when it, indeed, comes to mater to specific actors in a given seting.

Making Algorithms Matter

A SCANDINAVIAN CASE STUDY

What is at stake in this chapter is an exploration of algorithms as enacted in practice—in other words, how news organizations are dealing with algorithms, both in terms of the algorithmic logics of digital intermediaries such as Facebook, and the journalistic algorithms developed internally within the organizations in question. Drawing mostly on qualitative expert interviews with news editors and digital product managers from ten diferent Scandinavian newspapers and media outlets, the aim is to explore how algorithms come to mater in the specific seting of news work.6 The data sources on which I rely are 20 interviews I conducted personally among key actors in news organizations in Norway, Denmark, and Sweden from 2014 to 2016.7 The interviewees were digital project managers, editors-in-chief, technology directors, social media editors, developers, and journalists at leading news organizations. These organizations include the Scandinavia-based publishing house Schibsted, the Norwegian media group Polaris Media, Norway’s national newspapers Aftenposten and Verdens Gang (VG) and the Norwegian local newspapers Adresseavisen and iTromsø, Denmark’s national newspapers Jyllandsposten and Politiken, Sweden’s national newspaper Svenska Dagbladet, the public broadcasting service Sveriges Radio and the digital-only news service Omni.8 The news organizations reported on in this chapter all use algorithms in some way or another to select, create, edit, and publish the news. While there are many similarities among the diferent news organizations and their atempts to make sense of algorithms in journalism, the aim here is not to tell one master story of algorithmic journalism in Scandinavia. What is of interest are the politics entailed in trying to establish the meaning of algorithms. As such, it is a story of the ontological constitution of algorithms, of what happens when its status is up for grabs.

124

I F . . . T H E N

While the interviews certainly ascribe meaning to algorithms, meaning should not be seen as a “property of individual words or groups of words but an ongoing performance of the world in its diferential intelligibility” (Barad, 2003: 821). There are at least two sides to this that we need to acknowledge before moving on. First, materiality and discourse do not stand in opposition to each other; they are mutually constitutive (Barad, 2007; Orlikowski & Scot, 2015). The notion of a constitutive relation between the material and discursive implies that relations enact entities as part of situated practices. Like discourse and meaning, then, materiality or mater should not be understood in fixed terms. As Barad puts it, “mater does not refer to a fixed substance; rather, mater is substance in its intra-active becoming—not a thing, but a doing” (2003: 822). For Barad, “materiality is discursive [. . .] just as discursive practices are always already material” (2003: 822). Mater and materiality, in other words, are not substances but accomplishments made pertinent in a given situation (Cooren et al., 2012). This also means that “any distinction of humans and technologies is analytical only, and done with the recognition that these entities necessarily entail each other in practice” (Orlikowski & Scot, 2008: 456). In other words, studying how algorithms come to mater implies an understanding of how algorithms are made relevant and pertinent in a given situation— for example, by examining how and when diferent aspects of the algorithm are variously enacted in people’s accounts of them.

This brings me to the second point that we need to take into consideration when examining the meanings that news professionals ascribe to algorithms. Echoing ethnomethodological sentiments, we need to reflect on speech itself. News professionals necessarily speak as editors, managers, executives, and so on, which means that there are specific preoccupations and concerns that animate the person who occupies these positions. Consider the “Terror of War” incident again. As the editor-inchief of one of Norway’s leading newspapers, Hansen does not merely speak up against the power of social media platforms, but also as a spokesperson for the independent press more broadly. The perceived wrongdoing of Facebook’s algorithms is here used as a justification and legitimation of the news media. As Cooren et al. (2012) note, speaking as also implies a speaking for, not just in terms of being a spokesperson for a specific company or brand but also in terms of speaking to the particular interests and issues that such a position is supposed to entail. The preoccupations and concerns that news professionals speak to, then, are not mere reflections of a material referent but must, rather, be understood as marking the agency of what appears to them. In this sense, I follow Lucy Suchman’s call for “the need to atend to the boundary work through which entities are defined” (Orlikowski & Scot, 2008: 464):

Programming the News

125

Beginning with the premise that discrete units of analysis are not given but made, we need to ask how any object of analysis—human or nonhuman or combination of the two—is called out as separate from the more extended networks of which it is a part. (Suchman, 2007: 283)

Finally, there are two things to note with regard to this being a Scandinavian case study, as opposed to an investigation with a more global focus. In the way in which leading staf from select Nordic news organizations describe the coming together of journalism and algorithms, I believe that these interviewees’ strategic roles and, in many cases, their position as experts beyond the Scandinavian context, provide a valuable and convincing account of how, where and when algorithms have come to mater.

Also, in emphasizing the notion of enactment, I reject the notion of context as an explanatory factor. This may seem somewhat counterintuitive at first, given that I have been talking about “the particular context” of news media as my starting point. However, as Steve Woolgar and Javier Lezaun point out, there is a crucial diference between what ethnomethodologists call “context of action” and “context in action”—“between context as an explanatory resource available exclusively to the analyst, and the context as an emergent property of interaction available to  its participants” (2013: 324). Although algorithms need to be understood in context (in this case journalism) the meaning of algorithms cannot be reduced to context either. Just as news media professionals’ talk about algorithms cannot be understood separately from the context of journalism, journalism itself needs to be understood as the contingent upshot of material-discursive practices.9 The interviews provide rich and important data about journalism’s current condition, especially as it relates to the Scandinavian countries and to algorithms. Yet, this chapter is not about journalism or specific news organizations per se. To invoke the way I put it at the end of chapter 2: “Yes, journalism. Yes, Scandinavia. Yes, technology. Yes, economic challenges. Yes, social media platforms. Yes, algorithms. Yes, to all of the above, and no one in particular.” The story is more about the relationship between the elements and the new realities that emanate from that, rather than any one of these in isolation.

126

I F . . . T H E N

THE ALGORITHMIC LANDSCAPE OF NORDIC NEWSROOM

On the basis of interviews with news media professionals in senior positions, in this chapter I provide an in-depth case study on the ways in which algorithms mater and how organizational actors are handling, making sense of, and diferently orienting themselves toward the algorithmic logics of the current media landscape. In the previous two chapters I considered how algorithms are productive of certain subjectivities and orientations. Here I extend these conversations, by looking in more detail at the ways in which algorithms afect core understandings of institutional practice and professional values. Within the news media organizations studied, algorithms are put to work in two main ways. On the one hand, algorithms are enlisted as part of the production of news—primarily, in terms of automating certain processes or otherwise making the workflow more eficient by helping journalists find, produce, and disseminate the news. On the other hand, algorithms are applied to find new and beter ways to connect audiences and readers to revenue—primarily, by means of personalization and recommendations. In the following sections, I go beyond considering algorithms to be problem-solving devices (as the standard definitions would have us), and analyze them as problem-creating devices, and point to the social and ethical dilemmas that emerge when news media use algorithms. Beyond simply solving or creating problems, algorithms also “problematize.” There are at least two aspects to this: On the one hand, algorithms are rendered problematic, that is, judgements are passed, moral domains carved out, and certain concerns allowed to circulate (Arribas-Ayllon & Walkerdine, 2008). On the other hand, algorithms can also be seen as problematization devices, working to question the accepted boundaries and definitions of journalism itself.

The Algorithmic Potential

When people think of algorithms in journalism, it very often has to do with how they influence the final presentation of the news. Although it is important, this is only one of many ways in which algorithms are entrenched in journalism. For example, to list but a few of the ways in which algorithms are currently being used as part of the news media, algorithms are applied to find and research stories, verify sources, write and produce a story, and track the performance of articles after they have been published. They can also serve in A/B-testing headlines and the automatic production of news items by robots (Diakopoulos & Koliska, 2017; Lewis & Westlund, 2015; Linden, 2017). While it’s still early days with respect to algorithmic journalism, there seems to be a growing sense among news media professionals that algorithms have great potential and many areas of application. The interviews suggest that algorithms are deemed particularly useful in the production of news, both in the ways in in which they transform editorial workflows and the dissemination of news, by way of making sure the right content reaches the right persons. When algorithms are articulated as having potential, what is generally emphasized are the various possibilities and promises that they seem to entail. For many of the interviewees, algorithms promise to reduce the cost of production, make it more eficient, automate tedious and repetitive tasks, provide valuable insights into users and readers, help predict future preferences, help find links and correlations in the data, and present the news in new, personalized ways.

Programming the News

127

Interestingly, there seems to have been progress over time. When first contacting people in the news media to talk about algorithms and their role in journalism, people seemed less articulate about their potential than they seem today. While the early days of new technologies are always accompanied by a certain level of skepticism, particularly in fields engrained with strong professional values and cultures such as journalism, the interviews conducted in 2016 suggest a degree of normalization with regard to algorithms in news. This gradual normalization of algorithms is also reflected in the rate at which journalism conferences, industry seminars, panel discussions, and interest organizations put algorithms on the critical agenda. At the same time, news media professionals seem less apprehensive about algorithms than they were only a few years ago. As Hans Martin Cramer, former product developer at Aftenposten and now product manager at Schibsted Media Platform noted when I interviewed him for the second time in August 2016, the algorithms themselves are not a suficient explanation for this change in atitude. What changed since we first spoke in early 2014 was not so much the actual implementation of algorithms in the news (although things had certainly been happening) but, rather, the gradual normalization of discourses surrounding algorithms.10 “In addition,” he pointed out, “we are faced with a much more challenging economic situation, compared to just two years ago” (Interview 1, August 2016). As Cramer sees it, the media industry needs to move beyond the question of whether algorithms or robots will diminish or save the news. What is at stake is whether the new, emerging forms of journalism can be made sustainable from an economic point of view. Despite a great deal of talk about algorithms, the last few years, Cramer suggested, have shown that there are no easy solutions for developing algorithmic news production and dissemination: “it is either too complicated or too challenging.”

Yet, Scandinavian news outlets have long been at the forefront of digital innovation. Many of the managers and key stafers I talked to work for some of Scandinavia’s largest news organizations and publishing houses, including Schibsted, Polaris Media, and JP/Politikens Hus. As described by the Reuters Institute for the Study of Journalism in their annual digital news report 2016, these news outlets have a reputation for digital innovation in content and business models. Or, as The Columbia Review of Journalism recently wrote: “Leave it to the Scandinavians to be open to change and innovation, even in the ever-traditional world of newspaper journalism” (Garcia,  2017). For Espen Sundve, vice president of product management at Schibsted, news organizations essentially find themselves at an important crossroads:

128

I F . . . T H E N

Either they make themselves even more dependent on platforms like Facebook “that dictate the editorial and business rules without claiming any editorial and financial accountability for independent journalism,” or, they need to reinvent themselves by dictating the terms of what and whom algorithms are being used to serve (Sundve, 2017).

For Sundve, the algorithmic potential lies in making journalism more personal, which is not to be mistaken for what we usually think of in terms of personalization. In my interview with him, he worries that, by using a concept such as personalization, which primarily belongs to the world of technology, one may too easily lose sight of what is important from a journalistic point of view. Though “journalism can be a lot of things [. . .] algorithms are primarily there to complement the human understanding of $\operatorname { i t } _ { \ l _ { 1 } } ^ { \dag \mathnormal { \gamma } }$ Sundve says (Interview 7, May 2016). A central theme in Sundve’s discussion of journalism and the role of algorithms is the fundamental difference between news organizations and technology companies such as Facebook and Google. He likes to compare journalism and algorithms by talking about both of them as optimization tasks. It helps, he says, when talking to programmers in the newsrooms to speak about journalism in these terms. Essentially, journalism is about “closing the gap between what people know and should know but also want to know,” he adds. In principle, this is an optimization task that can be solved both technically and editorially, but “What distinguishes us from a technology company is that we include that which we think you should know” (Interview 7, May 2016, emphasis added). As with Hansen and his critique of Zuckerberg’s “Californian code,” Sundve’s description of the diference between journalism and technology companies is essentially about legitimizing and sustaining the role of news media as powerful channels of information and communication. Where social media platforms fail, journalism succeeds.

Faced with the algorithm’s arrival, news media professionals have a choice: either develop a proactive stance, or reactively adapt to the new technological landscape. As is evident in most of the interviews, simply dismissing algorithms is no longer an option. As we will see later in this chapter, one of the most pervasive storylines when it comes to algorithms in news, concerns their potential threat to the informed democratic public. Yet, there are limits to how sustainable these arguments about so-called filter bubbles and echo chambers are (a topic to which we will return shortly), as the news organizations themselves are increasingly using algorithms to present the news. The challenge, therefore, is to balance critique with actual practice, to make algorithms good as it were, and to engage in practices of making them mater in the right way. In talking to diferent news media professionals, it became clear how the algorithm in part derives its value from what is already valued. That is, algorithms are judged, made sense of and explained with reference to existing journalistic values and professional ethics: Algorithms and objectivity, algorithms and newsworthiness, algorithms and informed citizens, and the list goes on.

Programming the News

129