If...Then Algorithmic Power and Politics

The first step in knowing algorithms is not to regard the “impossibility of seeing inside the black box” as an epistemological limit that interrupts any “futile atempts at knowledge acquisition” (von Hilgers,  2011: 43). As Ashby recognized:“In our daily lives we are confronted at every turn with systems whose internal mechanisms are not fully open to inspection, and which must be treated by the methods appropriate to the Black Box” (1999: 86). Because opacity, secrecy, and invisibility are not epistemic anomalies but a basic condition of human life, the black box is not something to be feared but something that “corresponds to new insights” (von Hilgers, 2011: 32) When confronted with a black box, the appropriate task for the experimenter, as Ashby saw it, is not necessarily to know exactly what is inside the box but to ask, instead, which properties can actually be discovered and which remain undiscoverable (Ashby, 1999). Because what maters is not the thing but what it does, the cybernetic lead does not ask us to reveal the exact content of the box but to experiment with its inputs and outputs.

What can be discovered and described about algorithms difers from context to context with varying degrees of access and availability of information. However, even in the case of seemingly closed and hidden systems such as Facebook or Google, there are plenty of things that can be known. Speculative experimentation and playing around with algorithms to figure out how they work are not just reserved for hackers, gamers, spammers and search engine optimizers (SEO). In the spirit of reverse engineering—“the process of extracting the knowledge or design blueprints from anything man-made” (Eilam  2005: 3)—we might want to approach algorithms from the question of how they work and their general “operational logics” (Wardrip-Fruin,  2009). There are already some very instructive examples of “reverse engineering” algorithms within the domain of journalism and journalism studies (Angwin et al.,  2016; Diakopoulos,  2014) and in related academic calls for “algorithm audits” (Hamilton et al, 2014; Sandvig, et al., 2014).

Mythologizing the workings of machines, however, does not help, nor should we think of algorithmic logics as somehow more hidden and black-boxed than the human mind. While we cannot ask the algorithm in the same way we may ask humans about their beliefs and values, we may indeed atempt to find other ways of making it “speak.” Similar to the way ethnographers map people’s values and beliefs, I think of mapping the operational logics of algorithms in terms of technography.24 As Latour puts it, “specific tricks have to be invented to make them [technology] talk, that is, to ofer descriptions of themselves, to produce scripts of what they are making others—humans or non-humans—do” (2005: 79). Technography, as I use the term, is a way of describing and observing the workings of technology in order to examine the interplay between a diverse set of actors (both human and nonhuman). While the ethnographer seeks to understand culture primarily through the meanings atached to the world by people, the technographic inquiry starts by asking what algorithms are suggestive of. Although he does not use the term himself, I think Bernhard Rieder’s (2017) way of scrutinizing diverse and general algorithmic techniques could be thought of as an atempt to describe the “worldviews of algorithms” I have in mind when I use the term “technography.” Rieder ofers a particularly instructive account of how diferent technical logics—in this case, the Bayes classifier—entail certain values and assumptions that inevitably have consequences for the operational logics of specific systems.

60

I F . . . T H E N

Following Ashby’s cybernetic lead, what is at stake in a technographic inquiry is not to reveal some hidden truth about the exact workings of software or to unveil the precise formula of an algorithm. Instead the aim is to develop a critical understanding of the mechanisms and operational logic of software. As Galloway states, the question is “how it works” and “who it works for” (2004: xiii). Just as the ethnographer observes, takes notes, and asks people about their beliefs and values, Ashby’s observer and the technographer describe what they see and what they think they see. The researcher confronted with the black box algorithm does not necessarily need to know much about code or programming (although it is certainly an advantage). As Ashby points out, “no skill is called for! We are assuming, remember, that nothing is known about the Box” (1999: 89).

In general, we might say that a good way to start is by confronting known unknowns in terms of one’s own limits of knowledge. What, for example, is there to know about key concepts in computer science, mathematics, or social sciences that would be useful for understanding a specific algorithmic context? Then, we also have the option of tracing the many “semiotic systems that cluster around technical artefacts and ensembles” (Mackenzie, 2002: 211): patent applications and similar documents that detail and lay out technical specifications, press releases, conference papers on machine learning techniques, recorded documents from developer and engineering conferences, company briefs, media reports, blog posts, Facebook’s IPO filing, and so on. Finally, we might want to experiment with systems as best we can or even code them ourselves. As the next chapter will show, a technography of algorithms need not imply elaborate experimentation or detailed technical knowledge but, above all, a readiness to engage in unknown knowns, seeing the black box not as an epistemological obstacle but as a playful challenge that can be described in some ways (but not all).

PHENOMENOLOGICAL ENCOUNTERS WITH UNKNOWN KNOWNS

The second tactic suggests another way of knowing algorithms, which is to bring sense to what we do not know but know nonetheless. From a phenomenological perspective, approaching algorithms is about being atentive to the ways in which social actors develop more or less reflexive relationships to the systems they are using and how those encounters, in turn, shape online experience. Drawing on concepts such as “tacit knowledge,” “nonknowledge,” and “practical knowledge,” a phenomenological approach to algorithms is concerned with understanding how algorithms are perceived and made sense of by the actors in a given situation. Positing the algorithm as  an unknown known in this case is meant to highlight the productive force of personal forms of knowledge, knowledge that is gained through experience and practical engagement with one’s lived-in environments (Ingold, 2000). Although algorithms, in the strict technical and mathematical sense of the term, may remain unknowns, there are tacit forms of knowledge that linger in the background and which may have as much of an impact on the ways in which life with algorithms take shape as the coded instructions themselves. When long-lost friends suddenly show up on your Facebook news feed or a hotel that you have been looking at on a booking site materializes as an ad and seems to follow you wherever else you go on the Web, or Netflix just seems to be “reading your mind” in terms of the kind of movie you are in the mood to watch, algorithms become locally available for analysis in terms of the kinds of experiences and afects they engender.

Neither Black nor Box

61

Here, the notion of the black box is tackled by claiming that what there is to know about algorithms may not even be inside the box. As we have seen, for Ashby and the early cyberneticians, the black box was not an obstacle but a means for play and exploratory experimentation (Kaerlein, 2013: 659). The cybernetic notion of the black box as a basic human condition bears strong resemblance to the ways in which the psychologist Jean Piaget (originally, a zoologist) theorized the cognitive development of children. According to Piaget’s notion of playful learning, “children build up a picture of the world through making their own sense of inputs to the black box of the mind, converting experience to personal understanding of concrete objects” (Murray, 2012: 902). Instead of talking about knowledge in a definite sense, the emphasis is put on knowing as a continued and playful process. As biologists Maturana and Varela (1987) later put it, knowledge does not reflect a true world but should, rather, be judged by the construction of meaning from experience.

The focus is on how life takes shape and gains expression through encounters with algorithms. A phenomenological approach also considers the ways in which algorithms are “already embedded in particular circumstances” (Berry, 2011: 121). For Alfred Schütz, a pioneer in bringing phenomenology into sociological analysis, experience and the knowledge people derive from it is situated and contextual. It depends as much on various realms of relevance as it does on various social roles. Not everybody needs to know what algorithms are or how they work, but the “fact that we do not understand the Why and the How of their working and that we do not know anything of their origin does not hinder us from dealing undisturbed with situations, things, and persons” (Schütz, 1946: 463). Using “the most complicated gadgets prepared by a very advanced technology without knowing how the contrivances work” (Schütz, 1946: 463) is just a natural part of human life, but there are distinctions as to how relevant it is for people to know or atempt to know. A phenomenological analysis seeks out those slight transformations of relevance for an understanding of how phenomena change from being taken for granted to becoming a field of further inquiry. For example, Nick Couldry and his colleagues (2016) recently proposed the notion of “social analytics” as the phenomenological study of how social actors use “analytics” to reflect upon and adjust their online presence. As they see it, a “social analytics approach makes a distinctively qualitative contribution to the expansion of sociological methods in a digital age” (Couldry et al., 2016: 120). In a world in which large-scale data analysis takes on a greater hold, it is important to investigate the ways in which actors themselves engage with increased quantification. In Schütz’s (1946) terms, we might think of such actors as “well-informed citizens” who seek to “arrive at reasonably founded opinions” about their whereabouts in a fundamentally uncertain world.

62

I F . . . T H E N

A phenomenological approach to algorithms, then, is concerned with excavating the meaning-making capacities that emerge as people have “strange encounters” (Ahmed, 2000) with algorithms. As chapter 5 will show, people make sense of algorithms despite not knowing exactly what they are or how they work. When the outcomes of algorithmic processing do not feel right, surprise, or come across as strangely amusing, people who find themselves afected by these outcomes start to turn their awareness toward algorithmic mechanisms and evaluate it. As with the social analytics described by Couldry et al. (2016), when modes of appearance or senses of identity are at stake, actors may reflect at length on how to influence such operational logics; and, in doing so, they performatively participate in changing the algorithmic models themselves, a key reason it is important to study actors’ own experiences of the afective landscape of algorithms.

INTERROGATING THE CONFIGURATIONS OF STRATEGIC UNKNOWNS

The final methodological tactic that I want to address here focuses on examining algorithms as configurations in situated practices. Donna Haraway’s notion of figuration is particularly instructive in this regard as it points to the individuation of figures through co-constitutive relationships. Throughout her work, Haraway develops the notion of figures and figuration as a means of talking about the ways in which diferent elements come together, both materially and discursively, to form the appearance of more or less coherent relations. The cyborg, which is probably Haraway’s most well-known figure, is to be understood, above all, as a coalition across boundaries of identity. As a figure, the cyborg is not marked by a stable identity, but by the interaction of diferent and often conflicting concepts. Figurations, then, hold together “contradictions that do not resolve into larger wholes” because each idea, although conflicting, might be “necessary and true” (Haraway 1991: 149).

By conceptualizing algorithms as particular figurations that comprise and suggest conflicting ideas, we might be able to study the becoming of algorithms and the diferent ways in which algorithms come to mater in specific situations. As figurations, the identity of algorithms is variously made and unmade, and it is the analyst’s task to study when the algorithm becomes particularly pertinent and to challenge the conceptual separation between, for example, the social and the technical. Interrogating the configuration of algorithms begins with tracing out its specific cultural, historical, and political appearances and the practices through which these appearances come into being (Suchman, 2004). In this chapter, I have described this process as one of moving from a question of where the agency of algorithm is located to when it is mobilized, by whom and for what purpose.

Neither Black nor Box

63

What is at stake here, then, is not the black box as a thing but, rather, the process of “blackboxing,” of making it appear as if there is a stable box in the first place. If the metaphor of the black box is used too readily as a way of critiquing algorithms, Latour’s notion of blackboxing reminds us that we might want to scrutinize critically the ways diferent actors have a vested interest in figuring the algorithm as a black box in the first place. As discussed in this chapter, the alleged unknowability of algorithms is not always seen as problematic. It can also be strategically used to cultivate an ignorance that is sometimes more advantageous to the actors involved than knowing. By referring to the notion of strategic unknowns as part of this third methodological option, the intention is to point out the deeply political work that goes into the figurations of algorithms as particular epistemic phenomena. In chapter 6, I turn to these questions anew by looking at how algorithms variously become enlisted, imagined and configured as part of journalistic practices and products.

Concluding Remarks

Let this be the general conclusion: For every epistemological challenge the seemingly black-boxed algorithm poses, another productive methodological route may open. The complex and messy nature of social reality is not the problem. Just as algorithms constitute but one specific solution to a computational problem, we cannot expect a single answer to the problem of how to know algorithms. Borrowing from Law, “one thing is sure: if we want to think about the messes of reality at all then we’re going to have to teach ourselves to think, to practice, to relate, and to know in new ways” (2004a: 2). In this chapter the black box was used as a heuristic device to deal with this mess—not by making the world less messy, but by redirecting atention to the messiness that the notion of the black box helps to hide.

Not to be taken as a definitive, exhaustive list of well-meant advice, I ofered three steps to consider when researching algorithms. First, do not regard the “impossibility of seeing inside the black box” as an epistemological limit that impinges any “futile atempts at knowledge acquisition” (von Hilgers 2011: 43). Ask instead what parts can and cannot be known and how, in each particular case, you may find ways to make the algorithm talk. Second, instead of expecting the truth to come out from behind the curtain or to lay there in the box just waiting for our hands to take the lid of, take those beliefs, values, and imaginings that the algorithm solicits as point of departure. Third, keep in mind that the black box is not as seamless as it may seem. Various actors and stakeholders once composed black boxes in a specific historical context for a specific purpose. Importantly, they evolve, have histories, change, and afect and are afected by what they are articulated to. While we often talk about algorithms as if they were single stable artifacts, they are boxed to precisely appear that way.

64

I F . . . T H E N

Neither Black nor Box

65

Life at the Top

Engineering Participation

Over a decade ago, a geeky Harvard computer science undergraduate, together with a couple of friends, founded the world’s largest social networking site to date. With a claimed 1.4 billion daily active users, 25,105 employees, and a market value of \$530 billion as of December 2017, Facebook is not just a site through which users can “stay connected with friends and family.”1 Facebook is also a multibillion dollar business, engineering company, sales operation, news site, computational platform, and infrastructure. As the feature film The Social Network (2010) nicely shows, Facebook is also a myth, an entrepreneurial dream, a geeky fairy tale. Far from just a Hollywood story with its founder Mark Zuckerberg in the leading role, Facebook is a carefully crafted brand, which over the past decade has been marketed as a technology company comprising a young, vibrant, and creative class of geeks and hackers. Its mission: to make the world more open and connected.

Indeed, research suggests that, in many parts of the world, “Facebook is the Internet.” As one recent news article reported, “Millions of Facebook users have no idea they’re using the Internet” (Mirani, 2015). When people in countries such as Indonesia, Nigeria, Brazil, and India were surveyed, many reported they did not use the Internet while, at the same time, indicating that they were avid users of Facebook. Despite some media reports about a teenage flight from Facebook to other social media platforms such as Snapchat, the sheer scale and volume of active users and content shared via Facebook remains historically unprecedented. In the American context, Facebook remains the most popular social media site with a large user base that continues to be very active. According to a Pew Research report on social media usage for 2014, fully 70% of Facebook users engage with the site on a daily basis, a significant increase from the 63% who did so in 2013 (Duggan et al., 2015). It is probably no exaggeration to state that, for many, Facebook has transformed the ways in which they live their lives—the way they communicate and coordinate with their friends and family, receive and read news, find jobs and are fired from jobs. Facebook is constitutive of what Deuze calls media life, shaping “the ways in which we experience, makes sense of and act upon the world” (2012: 5). As I argue in this book, our lived experiences are not just permeated by media in general but, increasingly and more specifically, by the principle of the algorithm. Algorithmic life is a form of life whose fabric is interwoven by algorithms—of what can and cannot be  adapted, translated, or incorporated into algorithmic expressions and logics. Facebook, as we will see, sits right at the center of this algorithmic fabric, making intelligible and valuable that which can be quantified, aggregated, and proceduralized. This fabric is the theme of this chapter, and the question I explore is, what kind of togetherness does software invent? If, as van Dijck (2012) suggests, Facebook is wrapping relationships in code by configuring connections algorithmically, the question is in what ways.

66

To examine this, I discuss the case of Facebook’s news feed, focusing on the ways in which the feed is fundamentally governed and organized by an algorithmic logic. If Facebook is the computational platform that facilitates other things to be built on top of it, an architectural model for communication, then the news feed is the actualization of this model as a communication channel, a designed space in which communicative action can take place. Despite its purported discourse of openness, the Facebook news feed difers in fundamental ways from the ideal public sphere envisioned by Jürgen Habermas and other political theorists based in the idea of mutual deliberation and argumentation. While the intention is not to suggest compatibility between the communicative space of the news feed and the public sphere model, it is not without its warrants. The news feed, I want to suggest, is political insofar as it is exercising a position of governance. Here, I am not referring to politics in the conventional notion of parliamentary politics. Rather, politics is understood in a much broader sense as the encounter and conflict between diferent ways of being in the world (see also chapter 1). As the French philosopher Jacques Ranciére (2004) has theorized, politics concerns the reconfiguration of the distribution of the sensible. In this sense, politics is productive of “the set of horizons and modalities of what is visible and audible as well as what can be said, thought, made, or done” (Ranciere,  2004: 85). It is a way of informing experience that has no proper place. In this sense, power and politics are not limited to parliamentary governments or ruling bodies. Whereas political theorists such as  Habermas considered politics mainly as discursive forms of power produced through arguments in an informal public sphere (Flynn, 2004), I propound a view on politics that is sensitive to its material dimensions as well. In his seminal studies of the Long Island Bridge and the mechanical tomato harvester, Winner (1986) argued that technological arrangements impose a social order prior to their specific use. If taken to the realm of algorithms, we may start to consider them as political devices in the sense that they represent certain design decisions about how the world is to be ordered. After all, as Mackenzie suggests, algorithms “select and reinforce one ordering at the expense of others” (Mackenzie, 2006: 44). It is the work of ordering and distributing the sensible that I am concerned with in this chapter.

Life at the Top

67

Building on a Foucauldian sensibility toward the “microphysics of power” the question is not whom power most obviously belongs or where it is located but how and when power is exercised: those algorithmic techniques and practices that animate it and give it efect.

Before I get to a more detailed examination of the news feed, it is important to acknowledge that “there is an undeniable cultural dimension to computation” (Berry, 2011: 17). Software, code, and algorithms, while mathematical, logical, and mechanical in their materiality, are always made, maintained, and unmade. As computer historian Michael Mahoney says, software is always “constructed with action in mind; the programmer aims to make something happen” (1988: 121). As I discussed in the previous chapter, it can be dificult to study what actions programmers have in mind when the object of study is a commercial social media platform such as Facebook, since the motivations and intentions of software engineers often remain as obscure as the companies they work for. While cultural anthropologist Nick Seaver is right to criticize the tendency among media scholars to focus on figuring out the technical details of popular platforms at the expense of trying to figure out what their engineers think and do, access to the interiors of many of these companies for most researchers remains closed. Whether the scant literature in social media production studies is a result of closed access or merely a disproportionate interest in the study of users and consumption practices is not clear. As Nick Couldry has pointed out with regards to cultural and media studies, “production is often downplayed because it is not regarded as a place where everyday practice can be found” (quoted in Mackenzie, 2006: 14). While this may be true of much previous social media research, a notable exception is Alice Marwick’s ethnography of the Silicon Valley social media “technology scene.” In her book Status Update (2013), Marwick shows how software can embody the values of a group of users or creators—specifically, how software features may reflect the technology scene’s cultural beliefs about how the world is ordered and organized.

From a “values in design” perspective (Flanagan and Nissenbaum, 2014), the cultural beliefs and values held by programmers, designers, and creators of software mater. If we want to understand how software and algorithms are imbued in the fabric of everyday life, acknowledging and taking seriously the notion that these artefacts are far from neutral is imperative. By way of introducing the reader to Facebook and situating the object of analysis discussed in this chapter—the news feed—I first turn to the culture of engineering at Facebook. I show how materiality and culture are not separate but inherently entangled. While subscribing to the view that technology and practices need to be seen as co-constitutive, the emphasis in this chapter is on the material side of the sociotechnical framework.2 This should not be taken as a statement about cause and efect. Rather, it is about acknowledging that, despite purporting a relational view on the social and material, scholarship tends to “lean” toward one side or the other for analytical and empirical purposes. While the next two chapters “lean” toward the social side in terms of looking at social media users and news practitioners’ oral histories respectively, this chapter leans toward diferent aspects of materiality—to spaces and technology. Next, I draw on Michel Foucault’s architectural diagram of power introduced in Discipline and Punish (1977) to establish an analytical framework through which the news feed can be understood as an example of how sociality is programmed in algorithmic media.

68

I F . . . T H E N

Facebook and the Hacker Way

It is no coincidence that we only see Mark Zuckerberg wearing a washed out gray T-shirt or college sweater when he appears in public. Whatever imagined gap there might have been between users and the mythical founder figure, the T-shirt does its job of leting the world know that he is just a regular guy. He is not in it for the money, for the fancy designer clothes, for the frills. He just wants to do his job, “serving the community” the best way he can, as he himself said when confronted at the company’s first in-person public Q&A session with the question of why he always wears a T-shirt.3 For Mark Zuckerberg, the T-shirt is about eliminating unnecessary decisions. Decisions “make you tired and consume your energy,” he claims. Knowing exactly what to wear everyday leaves him more time to “build the best product and services.” The T-shirt, Zuckerberg says, helps him dedicate all his time and energy to “reach the company’s goal and achieve the mission of helping to connect everyone in the world and giving them the ability to stay connected with the people that they love and care about” (Facebook, 2014). Bolstering his role as the hard-working CEO, the T-shirt comes to solidify Mark Zuckerberg’s image as the Samaritan of code and connectedness. Not simply the uniform of the dedicated but the uniform of a particular kind of dedication—the hacker.

Facebook famously cultivates a culture and management approach they call “the hacker way.” As described in a leter penned by Mark Zuckerberg and included in Facebook’s IPO registration statement, the hacker way is an approach to building that “involves continuous improvement and iteration” (Facebook, 2012: 69). The hacker, Zuckerberg suggests, has typically been portrayed unfairly in the media. In reality, “hacking just means building something quickly or testing the boundaries of what can be done.” As Facebook already suggests, there are diferent versions and  meanings of the term “hacker.” The Facebook hacker is not like the university hackers of the 1960s described by Steve Levy (2010) in his seminal book called Hackers, nor anything like the underground hackers described more recently in Gabriela Coleman’s (2014) work on Anonymous. The Facebook hackers specifically “believe that something can always be beter, and that nothing is ever complete. They just have to go fix it—often in the face of people who say it’s impossible or are content with the status quo”. The hacker way is characterized by a number of core values, carefully crafted through slogans and popular sayings that pervade not only the company’s discourse but also its physical setting and work place. The word “hack” dominates the facades of one of Facebook’s two ofice buildings in Menlo Park, and the ofice spaces are “tatooed with slogans that inculcate through repetition the value of speed and iterative improvement: ‘Move Fast and Break Things;’ ‘This Journey Is 1% Finished;’ ‘Done Is Beter than Perfect’” (Fatal, 2012: 940). After moving headquarters in December 2011, Facebook is now even located at “the Hacker Way,” and Zuckerberg’s ofice faces onto “Hacker Square.” In a fascinating NYT article, journalist Quentin Hardy (2014) details the way in which the interiors of the ofice spaces at Facebook have been carefully designed to reflect the company’s culture of openness and change: Open ofice plans, of course, where meetings often happen on the fly, where no one has secrets, and common areas that without a warning may have the couches replaced in order to create a space of perpetual change. “Similarly, design changes to Facebook’s home page are known as ‘moving the furniture around’ ” (Hardy, 2014). The hacking culture is such an integral part of Facebook that it has become the principal way in which the company now presents itself, internally as well as externally.

Life at the Top

69