If...Then Algorithmic Power and Politics

Jyllandsposten, says it has been “absolutely fantastic to discover the strength in having something 100% algorithmically controlled,” he was quick to point out the danger of compromising the public ideal of journalism by creating filter bubbles. Similarly, Møller says that filter bubbles are exactly what they want to avoid. The question for many of the news organizations is how to solve this conundrum. On one hand, there is an increasing need to cater to the specific interests of individual readers. On the other hand, as Møller contends, “If everything were based on algorithms, we would be worried that you were missing some of the other things we would like to tell you about” (Interview 5, March 2014).

There is a general wariness among editors and managers of granting too much agency to algorithms. What really maters is how algorithms can be used to augment and support journalists in accomplishing their work. Jørgen Frøland, project manager for personalization at Polaris Media and developer of the algorithm at iTromsø thinks that much of the fear and concern surrounding algorithms stems from the fact that many people still do not understand what they really are. In most instances, Frøland points out, “algorithms are just plain stupid” (Interview 19, August 2016). It is not an either/or but a mater of using technology for the things that technology does well and people for what they know best. For example, Frøland adds, algorithms cannot make decisions based on common sense. For Ingeborg Volan (who works in the Polaris Media–owned regional newspaper Adresseavisen) the relative dumbness of algorithms means that their role in journalism should not be considered too problematic. She draws on a familiar argument voiced in debates about robot journalism: the algorithm is not as a threat to the job security of journalists but something that has the potential to make the job more exciting by “relieving journalists of some of their most tedious tasks” (Interview 4, August 2016). Because “good journalism is about asking critical questions, being able to judge the trustworthiness of sources, and to go out into the world and talk to people,” Volan does not think that algorithms pose a very serious threat to the journalistic profession.

The words people choose to talk about algorithms mater. Ola Henriksson, web editor and project manager at Sweden’s national newspaper Svenska Dagbladet, thinks that the word “algorithm” is too confusing and abstract. When I ask what the word “algorithm” means to him, he says “mathematics” and “calculations” (Interview 18, June 2016). However, this is not what journalism is really about, he says, suggesting that the term itself may do more harm than good. The Schibsted-owned Svenska Dagbladet became the first newspaper to implement the algorithmic infrastructure and valuation practices developed by Omni. Rather than talking about an algorithm, the Omni-inspired approach became more widely known as an “editor-controlled algorithm.” I asked Henriksson whether he thinks calling it editor-controlled makes a diference. “Yes, for some, it actually seemed to make a diference,” he replied. The notion of “editor-control” helped ease people’s minds, and ensure them that they would still play an important role, despite the emergence of an algorithmic media landscape.

140

I F . . . T H E N

It is not dificult to see the appeal of the notion of an “editor-controlled algorithm” as described by several of the news professionals. It conveniently limits the power of code, while allowing for human agency at the same time. The result is something less threatening, yet modern and innovative. Since its implementation at Svenska Dagbladet, the algorithm and new publishing platform has widely been heralded as a success. Partially guided by editors and partially by reader preferences the algorithm delivers a personalized homepage for each website visitor based on timeliness, news values, and interest. Since launching the algorithm the newspaper claims to have seen a boost in trafic by up 15% in page views and 9% in unique visitor. This is the story of an algorithmic hero, not a villain. While the Swedish newspaper was practically pronounced dead, the algorithm made the future look bright again, or so at least the story went (Garcia,  2017). What makes this algorithm applaudable are its owners. That is, what sets the journalistic and “editor-controlled algorithm” apart from Facebook’s “Californian code” lies in the former’s moral salience. Fredric Karén, editor-in-chief at Svenska Dagbladet, puts it in this way in an interview with Storybench:

I believe that you need to control and own your own technology. Google and Facebook would be thrilled for us to join their ad platforms. But to be a publisher is also about integrity: To own the way you publish, when you publish and where you publish is really important moving forward. (Rodrigues, 2017)

Let’s pause for a moment and consider what the word “integrity” is doing in the above account and how it works to make the algorithm mater in the right way. In light of the many recent controversies surrounding the Facebook algorithm, especially regarding the discussions about the extent to which Facebook should assume editorial responsibility, the news media find themselves in the welcome position of highlighting and strengthening their institutional legitimacy. For Espen Sundve at Schibsted, the prospect of becoming too absorbed with Facebook’s news feed algorithms is one of his biggest concerns about the future of news. Although Sundve has Facebook’s publishing platform “Instant Articles” in mind when he talks about the risk of being engulfed by the Facebook algorithm, these concerns have ramifications far beyond its financial accountability. For news professionals, the problematic algorithm does not merely manifest itself inside news organizations but, perhaps more importantly, in terms of the external algorithms governed by Facebook and Google. Sundve says of the Facebook news feed that, “from a democratic perspective,” it is very problematic to “give away your editorial position to someone who clearly does not take any editorial responsibility” (Interview 7, May 2016).

Becoming too dependent upon algorithms that do not reflect the right values is risky. As Sundve says, technology companies may atempt to give people what they want, but journalism is primarily about providing people with information they should know and taking responsibility for that. Giving people information about what they should know, Sundve adds, also implies giving people the opposite view of an issue. “If you are a vaccination opponent, you should also be exposed to facts on the other side of the spectrum and not just find 50 other people supporting your views on your Facebook news feed,” Sundve says (Interview 7, May 2016). One way to avoid filter bubbles, many informants suggest, is to think more carefully about the design of algorithms. It is not that algorithms always create bubbles and echo chambers. The question is what a specific algorithm is actually optimized for. As Gillinger sees it, there is no necessary opposition between algorithms and diversity—“it all depends on what it is that you want your algorithm to reflect.” It is entirely possible to optimize for clicks if that is what you want; “but, in my world, that is a badly written algorithm,” Gillinger says (Interview 17, May 2016). While critics worry about Facebook’s alleged power to create filter bubbles and misinformed publics, there is also the sense in which the democratic crisis can be resolved by writing beter and journalistically sound algorithms.

Programming the News

141

Algorithms as Problematization

Algorithms do not merely solve or create new problems, as the interviews make perfectly clear. They also need to be understood as “problematization” devices that work to question the accepted boundaries and definitions of journalism itself. It is here that we see most clearly the eventfulness of algorithms at play. As I argued in chapter 3, the notion of event allows for an analytical position that is not so much interested in establishing the “empirical accuracy” (Michael, 2004) of what the algorithm is or what it means, but which instead looks at how algorithms have the capacity to produce new orderings or disorderings.14 As Mackenzie points out, “event thinking” has helped scholars “think about the contingency of new formations or assemblages without ‘front-loading’ a particular ontological commitment” (2005: 388). In talking to news media professionals and observing the ways in which algorithms have emerged as an object of concern within journalism, it seems to me that the power and politics of algorithms are manifested in the way they problematize accepted boundaries. Paul Rabinow says of an event that it makes “things work in a diferent manner and produces and instantiates new capacities. A form/ event makes many other things more or less suddenly conceivable” (1999: 180). Just as “events problematize classifications, practices, things” (Rabinow, 2009: 67), an algorithm, I suggest, “puts into question or problematises an existing set of boundaries or limits within” journalism (Mackenzie, 2005: 388). In going forward, the question is not just what happens to particular domains when they introduce an algorithm, but how algorithms as events make it “possible to feel, perceive, act or know diferently” (2005: 388).

142

I F . . . T H E N

In the previous sections, I identified at least two distinct ways in which the understanding of algorithms is realized in Scandinavian newsrooms: as forms of potentiality and as something that is articulated as more or less problematic. In this final part of the chapter, I suggest that the notions of potentiality and the problematic do not necessarily stand in opposition to one another, but rather should be thought of as a “relational play of truth and falsehood” (Rabinow, 2009: 19). That is, when algorithms become contested, when they fluctuate between discourses of promise and closure, achieve recognition outside of accepted boundaries, and become imbued with certain capacities for action, they emerge as forms of “problematization.” A problematization, Foucault writes:

Does not mean the representation of a pre-existent object nor the creation through discourse of an object that did not exist. It is the ensemble of discursive and nondiscursive practices that make something enter into the play of true and false and constitute it as an object of thought (whether in the form of moral reflection, scientific knowledge, political analysis, etc.). (1988: 257)

The term “problematization” suggests a particular way of analyzing an event or situation—not as a given but, rather, as a question. As Rabinow suggests, the analyst’s task is to understand how, in a given situation, there are “multiple constraints at work but multiple responses as well” (2009: 19). For something to be problematized, a number of things must first be made uncertain or unfamiliar or “have provoked a number of dificulties around it” (Foucault, 1994: 117). As we have seen, the ways in which actors respond to algorithms in the news industry need to be understood against several constraining and enabling factors—most evidently, the dire economic circumstances in which most news organizations find themselves. Moreover, the responses of news professionals need to be understood against the broader landscape of the algorithmic media of which news organizations are a part. This applies in particular to their relationship to powerful competitive players such as Facebook and Google.

One aspect of this has to do with the way the figure of the algorithm problematizes existing boundaries of what journalism is and ought to be, an issue to which I will soon return. Another aspect has to do with the way news media are increasingly caught up in powerful “algorithmic configurations” (Callon & Muniesa, 2005) that involve other calculative agencies such as Facebook that help to organize their business. Of course, social media platforms have long been important to established news media. As Jens Nicolaisen at Jyllandsposten points out that, from a business perspective, their biggest competitors are not the other major Danish newspapers but social media platforms. “Facebook has had a tremendously greater impact on our business than Berlingske and Politiken,” he says (Interview 3, February 2015). However, research on the rise of social media platforms and their impact on the existing mass media environment show how social media do not replace but, rather, supplement and challenge more established forms of media and media use (Chadwick,  2013; Hermida et al., 2012; Kleis Nielsen & Schrøder, 2014). As Nicolaisen puts it, news media and social media seem to have entered into a marriage of convenience of sorts. In the “hybrid media system,” the news information cycle is no longer defined by the mass media alone but needs to be understood, instead, as an assemblage composed of many diferent media forms, actors, and interests in mutual relations of co-dependence (Chadwick, 2013). While we know that many news media sites atract a fair share of web trafic from social media—in particular, from Facebook (Kleis Nielsen & Schrøder, 2014), the notion of hybridization would also suggest that Facebook is dependent on the news media. Indeed, news is crucial for Facebook. People seem to share less of their personal lives on platforms like Facebook and Twiter than they used to (Grifith, 2016) and interact more with news shared by professional content producers. According to Will Cathcart (2016), who is vice-president of product at Facebook, 600 million people see news stories on Facebook on a weekly basis. Moreover, the company claims that a big part of why people come to Facebook in the first place is to catch up on the news and talk with their friends about it. As personal updates seems to be in decline, Facebook clearly counts on the news media and external content producers to help atract those much-needed clicks, likes, and engagement metrics that fuel its advertising business.

Programming the News

143

The hybrid media system extends well beyond web trafic and “likes,” encompassing the operational logics of platforms and the players involved. Similar to how users reorient their social media behavior to accommodate the algorithmic logics of platforms, as was discussed in the previous chapter, the news professionals in this chapter note how algorithmic logics increasingly also inform journalistic practices. Ingeborg Volan at Adresseavisen observes how journalists in many ways have to orient themselves toward the algorithmic logic of external players like Facebook. Volan emphasizes how vital it has become to be good at playing Facebook’s game. Becoming more atuned to the algorithmic logic of Facebook is now a necessary ingredient in reaching as many existing and new readers as possible. This does not mean, however, that you simply “dismiss all your journalistic principles and just follow Facebook blindly,” says Volan. The question is whether you can find ways of “gaming Facebook’s algorithms to serve your own journalistic purpose.” The tricky part, Volan adds, is finding good ways of adhering to the “things we know Facebook will reward but without compromising journalistic principles.” If you want your audience to come to your own site to watch a particular video, you do not put the whole video up on Facebook. What you do, instead, is share a “teaser” on Facebook, which directs people to your own site, according to Volan. However, producing videos also makes sense from the perspective of the operational logics of platforms. Given that Facebook’s algorithms privilege videos over text, Volan suggests, they make sure to publish more videos on Facebook in order to be prioritized by the algorithm (Interview 4, August 2016).

144

I F . . . T H E N

The case of Omni further underscores the fact that the journalistic practice in the social media age cannot be reduced to making the news “algorithm ready,” blindly following the dictations of Facebook. Here, it becomes apparent that algorithms also problematize the practice and meaning of doing journalism. In contrast to previous research on journalists’ responses to new technologies in the newsroom, which is “nearly unanimous in concluding that journalists react defensively in the face of such boundary intrusions on their professional turf” (Lewis & Usher, 2016: 15), many of the editors working at Omni see new technologies as part of the atraction of working there in the first place. Although, of all the organizations studied, Omni has, perhaps, gone furthest in making the algorithm an essential component of the editorial workflow itself, we can conclude that the algorithm by no means deskills the workforce, as critics often worry. Instead, the inclusion of an algorithm into the workflow repositions and displaces the work in new ways. As the notion of problematization is meant to highlight, one cannot look at algorithms as something new and detachable that has been added to a preexisting journalistic domain. Such a notion falsely presupposes that algorithms and journalism are discrete and freestanding domains that suddenly have been forced to interact. Conceived instead as “eventful,” algorithms are defined by their capacity to produce new environments. This alternative view is much more nuanced analytically, providing greater explanatory force to the cases I have been considering in this chapter. As a consequence, it renders the question of whether algorithms are responsible for deskilling the workforce superfluous. Instead, it urges us to ask how forms of algorithmic interventions are productive of new journalistic realities, and to what efect. Simply claiming that algorithms make people redundant will not do. While it seems safe to say that algorithms do something that, in some cases, may imply reducing the need for human labor, we should be wary of claims that simply end there.

What, then, do algorithms problematize? In the remainder of this chapter, I will focus on two possible answers, aware of there being no exhaustive explanation for a question like this. First, algorithms contest and transform how journalism is performed. As the interviews suggest, algorithms do not eliminate the need for human judgment and know-how in news work; they displace, redistribute, and shape new ways of being a news worker. Similarly, seeing machines as lacking instincts and, therefore, ill-suited to replace journalists and their “gut feelings” is a too simple a view. What journalists may have a “gut feeling” about expands and transforms as well. The field observations and conversations I had with staf at Omni revealed how editors in particular have developed an instinct for how the machine works. As “Clare” tellingly suggests, it is not just a mater of having an instinct for news at Omni, it is also about “developing a feeling for the algorithm” (Interview 11, March 2015). In contrast to much recent discourse on computational journalism, which hinges on the notion of “computational thinking,” the case of Omni suggests that not all “computational thinking” requires journalists to learn to think more like computer scientists. Whereas the notion of “computational thinking,” as originally conceived, refers to an approach to problem solving that draws on concepts fundamental to computer science such as abstraction, automation, and recursivity (Wing, 2006), editors at Omni do not necessarily think of their work in computational terms. Rather, they think like journalists who have learned to optimize their product and practice not just for an outside reader but also with an algorithm in mind. In contesting existing professional boundaries, the algorithm contributes to the emergence of diferent readers, diferent journalists, and, ultimately, diferent news.15

Programming the News

145

Second, as I have already alluded to, algorithms problematize existing boundaries of what journalism is and ought to be. The interviews reveal how the inclusion of algorithms in journalistic practices and products fuel important discussions about the nature of news. A central theme in Gillinger’s discussion of developing an algorithm to curate and rank socially engaging content at Sveriges Radio is the way in which the process itself brought about important reflections and discussions on the significance of journalism today. As he says, “In my 20 years of working in various editorial teams, I have witnessed a lot of implicit ideas about what constitutes news values.” Talking about what an algorithm is supposed to reflect and how it ought to work has forced people to reflect on these implicit and often unspoken believes and ideas about news, Gillinger says (Interview 17, May 2016).

Moreover, algorithms problematize journalism through “boundary-drawing practices—specific material (re)configurings of the world—which come to mater” (Barad, 2007: 206). As a contested mater, algorithms have become the focus of much public discussion surrounding journalism, making important marks upon the journalistic institution itself. As “pharmacological” objects of sorts, algorithms make themselves known as both remedy and harm.16 The question, however, is not so much whether algorithms are poison or cure but what is at stake in their diferen tial configuration as one or the other. As Suchman writes, “Objects achieve recognition within a matrix of historically and culturally constituted familiar, intelligible possibilities [in which] we might understand ‘things’ or objects as materializations of more and less contested, normative identifications of mater” (2007: 10). As the interviews and various controversies surrounding algorithms suggest, the news media are by no means a neutral component in the diferential mobilization of algorithms. In line with the broader argument of the book, this chapter reveals how algorithms do not always mater (as much), but are rather made to mater in certain ways. We can think of this form of diferential matering as a form of ontological politics, insofar as algorithms are carved out as entities with specific features and “normative variability” (Marres, 2013). That is, algorithms do not always carry the same moral or normative capacities. Sometimes they are rendered bad or inferior, do the wrong things, as indicated in the discussions over filter bubbles. At other times, algorithms do the right thing, for example, by helping journalists do a beter and more efective job. What the algorithm does and how it is made meaningful is not set in stone. This variability suggests that the power and politics of algorithms cannot necessarily be found in the algorithm, but rather, as it was also suggested in chapter 3, is more a question of when the agency of algorithms is mobilized and on whose behalf.

146

I F . . . T H E N

Concluding Remarks

Algorithms are not simply algorithms, as it were. Their ontology is up for grabs. Let’s briefly return to the notion of fake news introduced at the beginning of the chapter. While Facebook should not be blamed for creating a misinformed public, or for determining the outcome of elections, the debate over fake news and post-truth underscores the general point I make in this chapter about the power and politics of the algorithmic media landscape. If we want to understand the ways in which algorithms mater, we need to pay atention to the ways in which they are made to mater, as well as the ways in which they are made to mater diferently in diferent situations and for diferent purposes. Going beyond viewing algorithms as problems solvers, as the standard computer science definition would have it, or creating new problems and concerns, a trope on which the social science perspective on algorithms tends to focus, I have made a case for atending to the eventfulness of algorithms that constitutes them as particular forms of problematization. Investigating the ways in which algorithms become relevant, I discussed how relations of sameness or diference are enacted on given occasions, with given discursive and material consequences.

As I showed, the “truth” of algorithms is “played” out in material-discursive practices. Constitutive of what Foucault called “games of truth,” the narratives of news media professionals show how the matering of algorithms is predicted on establishing certain norms and values of what is desirable/problematic, good/bad, true/false. Similar to how truth games were set up in the organization of medical knowledge and the designation of madness in Foucault’s writings (1988; 1997), the truth games surrounding algorithms have to be connected to a whole series of socioeconomic processes and professional practice. As Foucault makes sure to highlight, a game “is not a game in the sense of an amusement,” but rather, “a set of rules by which truth is produced” so that something can be “considered valid or invalid, winning or losing” (Foucault & Rabinow, 1997: 297). These rules, however, are never static. While implicit and explicit “rules,” such as journalistic objectivity and professional ethics endure, the meaning of these “rules” change. In an algorithmic age, news media are not merely adapting to the presence of algorithms or simply using algorithmic systems as part of journalistic practices. What it means to be a journalist and publisher changes, as do definitions of what news and newsworthiness is and should be. News selection only moderately falls to a “gut feeling.” The “hard facts” of metrics, trafic data and new media logics help transform feeling from gut to byte. The way news professionals talk about developing “a feeling for the algorithm,” or “gaming the

Programming the News

147

Facebook algorithm” to their advantage, is indicative of the operation of politics. Whether we are speaking of the “internal” algorithm developed inside a specific newsroom or the “external” ones of Facebook, they seem to have the capacity to disturb and to compose new feelings, instincts, and sensibilities. In the hybrid media system where traditional publishers increasingly rely on trafic from social media platforms, people’s engagement with the news is configured by the algorithmic systems of increasingly interdependent, yet professionally and ideologically distinct, information providers. News happens on diferent platforms, it migrates from publishers to social media and back again, and in the process gets caught up in complex sociomaterial dynamics of ranking, sharing, linking, and clicking. It reconfigures how journalism is apprehended, and what counts as news at all. Such is the ontological politics of algorithms, shaped in the encounters and conflicts that make them mater.

The overarching themes that emerged from the interviews with Scandinavian news media professionals revealed how algorithms are entangled in and emerge as contested maters in various processes of drawing boundaries. That is, algorithms are materialized as they become invested with certain social, political and ethical capacities. A “good algorithm,” for example, is one that reflects the right kind of values—in this case, journalistic values. By the same token, the “bad algorithm” is one that takes no editorial responsibility. But the good and the bad vary too. The “bad” or inferior algorithms of social media become the condition of possibility for the “good” journalistic algorithm. Facebook’s problematic algorithm simultaneously represents an opportunity for news sites. As Aftenposten’s censorship allegations against Facebook in the “Terror of War” case usefully suggests, algorithms are not inherently bad or good, but can be used more or less strategically to make such claims. While the “bad algorithm” of Facebook may instigate worries about declining conditions for democracy, it may also be mobilized as a productive tool insofar as it allows the news media to strengthen their authority as public watchdogs. As media scholar Charlie Becket (2017) provokingly suggests, “fake news” may be the best thing that has happened to journalism in recent years, as it “gives mainstream quality journalism the opportunity to show that it has value based on expertise, ethics, engagement and experience.” Conversely, the journalistic algorithm is manifested as the solution to a problem that social media made more or less conceivable. According to such a view, there is no such thing as a problematic algorithm in and of itself, only algorithms that become problematic as an efect of diferentially atributed forms of agency. With regard to the allegations of the algorithmic propagation of fake news on Facebook and its alleged efects of a misinformed public, it is my claim that any fruitful explanation needs to move beyond the algorithm per se. Although it certainly maters that Facebook’s news feed algorithms are optimized for shares and engagements, how and when it maters may mater even more.

148

I F . . . T H E N

Conclusion

Algorithmic Life

I began this book by asking readers to consider the following scenario: Copenhagen on a rainy November day, where a semester is about to finish. The festive season is around the corner, but it’s otherwise a normal day like any other, filled with the habits of everyday life. Many of these involve media, such as checking Facebook, reading online news, searching Goggle for information, writing emails, tweeting a link, buying a Christmas present from Amazon, watching an episode of “House of Cards” on Netflix. While there is nothing particularly eye-opening about these moments, that was exactly the point. It describes life lived with, in and through the media. Or more precisely, a life fundamentally intertwined and entangled with algorithmic media of all sorts. As I am writing this concluding chapter it so happens to be a remarkably similar day in November, albeit two years later.

Of course, nothing ever stays exactly the same. This time around it is sunny, not raining, and I have much less time to check Facebook or watch Netflix, as I am trying to finish this book. The platforms and algorithms have changed too. While we might not always notice, the calculative devices of contemporary media constantly work to give us more of what we seemingly want. Atuned to users’ clicks, shares, and likes, algorithms are constantly updated, revised, and tweaked to make the flow of information seem more relevant and timely. As Wendy Chun suggests, “New media live and die by the update: the end of the update, the end of the object” (2016: 2).1 This condition of the perpetual update means that I am never guaranteed to see the same content on Facebook or Twiter as my neighbor or friend, due to network logics and algorithmic processes of personalization. Every day, or even several times a day, users are implicitly confronted with the question of “what will I see this time”?

It used to be that one could expect a certain sense of chronological order to the ways in which information was presented online. Today, chronos seems increasingly to have been replaced by kairos, the right or opportune time to say and do something. As evident in news feeds of all kinds, time is no longer about the linear, continuous flow, but rather about the very punctuation of linear time itself. As it was argued in chapter 4, kairos is what constitutes the temporal regime of algorithmic media. After all, as Facebook tellingly suggests, the goal of the news feed “is to deliver the right content to the right people at the right time so they don’t miss the stories that are important to them” (Backstrom, 2013). In the past few years, many more media platforms have followed suit. Instagram and Twiter are only some of the more prominent platforms that have started to punctuate the flow of the “realtime” feed by highlighting “right time” content instead. The previous chapter shows how the logic of “the right content, to the right people, at the right time” is increasingly becoming part of traditional news media as well. Faced with economic hardship, news media are reorienting themselves to the new digital realities. One consequence is the way news organizations adapt to the algorithmic media landscape by deploying algorithms to produce, distribute and present news in ways that readers are already familiar with from being in social media spaces. But even more significance is carried by the fact that traditional news media are becoming algorithmically atuned.

7

149