In this post, I want to explore television programming. Not the kind of programming we might typically associate with TV (i.e. the content, or the act of commissioning and scheduling) but the sort of programming that happens behind-the-digital-scenes; the computer programming, or coding, of the interfaces, databases and algorithms which increasingly frame our experiences of television today. Whether we are tuning into broadcast TV, bingeing on a DVD box set, or streaming content through services such as the iPlayer or Netflix, each of these encounters with television is structured by code.

Mr Robot

Mr. Robot (USA, 2015): The culture of coding enters the mainstream (again).

Understandably, television programmers (in the more traditional sense of the term) have received far greater critical attention than their much more recent namesakes (who, in order to avoid ambiguity, I refer to as coders from here on in). Nevertheless, there is a growing body of literature concerned with the work of coders. Indeed, in recent years there has been a wide range of analyses of the interfaces, algorithms and databases that are produced by coders and which are commonly associated with the new breed of more data-driven television services such as Netflix, Amazon Instant Prime, and even the streaming outlets of broadcasters such as the iPlayer (BBC) and All 4 (Channel 4). Daniel Chamberlain (2011), for instance, uses the geographical concept of the “scripted space” to describe the way in which television interfaces are often designed to guide viewers along particular predetermined paths (and more often than not, to promote specific commercial interests). In a somewhat similar critical vein, Hye-Jin Lee and Mark Andrejevic (2014) have considered the way that second-screen devices, and their various TV-oriented apps, potentially offer the ‘promise of the “democratic surround”: [through which] users will be provided with greater degrees of freedom and more meaningful forms of participation in constructing their mediated information environments’ (52). But despite their initial optimism, Lee and Andrejevic ultimately conclude that the design of these technologies and their interfaces more closely approximates a “digital enclosure” than a “democratic surround”.

Like other studies of television interfaces, both of these accounts address a series of important issues connected to the work of coders and the digitization of television more broadly – in particular, they engage with issues of control, freedom, commodification and surveillance. Yet both accounts are primarily concerned with the moment at which the viewer encounters the product of the coder’s labour. In other words, this material focuses on the fruits of their labour, rather than the process of labour itself. As such, they tell us little about the culture, values, communities and/or working conditions of this particular profession. Anything they do tell us in this regard can only be gleaned through inference.

The need to examine the labour of coders is necessary as there is a tendency in popular and critical discourses to treat interfaces, algorithms and databases as sentient mechanisms that organise, select and disseminate culture – entirely independent of human interference. This attitude is especially evident in a recent article in Ad Age, the by-line of which reassures us that, contrary to claims of technophobes, ‘algorithms won’t replace humans any time soon’ (Kaye, 2013). But the added qualification of “any time soon” has a slightly disconcerting ring of Skynet about it. On a more serious note, the article seems to overlook the fact that algorithms themselves are the products of human labour. Like interfaces, they are designed to achieve specific (almost always economic) goals [1]. Algorithms may displace certain forms of human labour but this is nothing new, nor does it mean that machines will entirely replace humans when it comes to making creative decisions.

For television scholars and TV viewers, the interface is perhaps the most tangible product of the coder’s labour. In contrast, databases and algorithms are far more discreet – performing a crucial if unseen function in the background that drives the delivery of media content. Yet much of the literature on algorithms (such as, Havens 2014; Striphas 2015) and databases (such as Manovich, 2001) represent attempts to either conceptualise these emerging phenomena or to explore their wider cultural implications. For instance, as the Ad Age article indicates, there is a recurring theme in much of the recent critical discourse on algorithms, in which it is regularly argued that human agency is increasingly diminished by the efficiency and efficacy of the algorithm. While there is certainly some truth to these claims, I’m interested in what happens before we even reach that point. Surely that’s worth exploring?

Will Program for Food

And so, after a rather convoluted but necessary critical detour, I finally arrive at the point of this post. Since attending a recent industry training event designed as an introduction to the field of “big data” (a topic that I explored in my previous contribution to CST and one that underpins the types of television experiences discussed here), I have become fascinated by TV’s “below-the-line” digital workers (as John T. Caldwell might describe them): specifically coders and the open-source communities to which many of them belong. My introduction to this particular profession has resulted in some interesting and significant discoveries. Perhaps the most important of these discoveries is that many, if not most, digital television platforms are largely based on different iterations of open-source software (hence the connection of television’s coders to the open-source community). Initially, this reliance on “free” [2], open-source software and its affiliated communities may seem somewhat ironic given the highly commercial, protective and litigious nature of the media industries. Yet open source software, and its wider community, constitute an integral part of the success of companies such as Netflix and Amazon. Not only do they comprise a significant part of their (often unpaid) workforce, but they are also the producers of many of the tools and software solutions adopted by the television industry.

Netflix are particularly interesting in the context of this post as they have a history of utilizing free, or at best, partially compensated labour. For instance, in 2006, the streaming service launched a competition called the Netflix Prize, which challenged programmers to improve the efficiency of their collaborative filtering algorithm by at least 10% (this is the main algorithm responsible for organising the user interface and driving the company’s recommendation engine). In exchange for providing a non-exclusive license to Netflix, the winner would receive $1m; meanwhile Netflix would retain the rights to the successful code / algorithm.

$1m is no small sum, but Netflix profited from this one off payment in a number of signifanct ways. Firstly, the company were able to improve their primary recommendation algorithm by 10.09%. This, we have to assume, must have had some kind of positive economic effective. Unfortunately, the precise financial implications of this refined algorithm are, somewhat ironically, difficult to determine. Secondly, Netflix were able to assemble and capitalise upon a truly enormous workforce. It was reported that by June 2007 more than 20,000 different teams had registered to participate. While it’s difficult to determine exactly how many individuals were involved, or how many hours of labour they contributed (presumably, many of these were working on the project in their spare time), the prize nevertheless carried on for several years and thus generated a significant amount of labour. Suddenly, $1m doesn’t sound like so much. Furthermore, had nobody managed to succeed, the company wouldn’t have lost any money (aside from the very small “progress” funds they periodically awarded). Thirdly, the prize functioned as a form of promotion. On the one hand, the difficulty of the challenge drew attention to the existing efficiency of the Netflix algorithm, whilst on the other hand, it eventually demonstrated small but economically significant improvements to its already well-renowned system. In short, it was a win, win, win scenario for the streaming giant.

Netflix prize-winners 2009

The winners of the Netflix Prize in 2009, pictured here holding an obligatory oversized novelty cheque

Importantly, the prize also helped to establish and nurture a relationship with coders and the open-source communities to which many of them belong. This is especially important for Netflix, whose entire digital infrastructure is based on various versions of Apache’s open-source software. As an example of digital labour, the Netflix Prize is now somewhat outdated. It began almost a decade ago and concluded by 2009. Nevertheless, it does draw attention to the way in which the media industries (and television distributors in particular) are increasingly reliant on coders and, by extension, the open-source community. In the example above, it’s fair to say that collaboration was absolutely integral to the eventual development of a more effective algorithm. Indeed, it would have been difficult if not impossible for Netflix to achieve this level of algorithmic improvement through the more traditional means of selecting and recruiting staff on a permanent or project-by-project basis. Collaboration with the coders and the open-source community is therefore integral to the Netflix business model.

As the example of the Netflix Prize indicates, the streaming service is acutely aware of its reliance upon outsourced and open-sourced forms of labour. As such, Netflix has sought to develop a relationship with these communities through a variety of strategies, of which the Netflix Prize was just one such attempt. For example, the company also regularly courts collaboration through their tech blog, where, as per the expectations and customs of the open-source community (and, perhaps more importantly, as per the conditions of the Apache Licence), they regularly publish software based on the Apache code, outline its potential applications, and openly invite feedback from those working in the wider community. Quite often, they even use this platform to solicit job applications.

Another example of this digital outsourcing of labour can be seen in the Netflix Hack Day – a regular event in which coders from across the globe can pitch their ideas for new Netflix features. Importantly, the event is designed to not only to solicit ideas that might be commodified and turned into actual Netflix features (and presumably protected/controlled through the enforcement of intellectual property rights), but participants themselves also write the code. In other words, these coders do most of the hard labour. It isn’t entirely clear what their reward is; therefore more research needs to examine the labour dynamics of this particular arrangement. The key point to emphasise here is that the Netflix Hack Day is distinct from the way in which other software developers traditionally solicit feedback from users or software programmers (not to mention the ways in which they may or may not be compensated for their labour).

Netflix Earth: A concept developed for a recent Netflix Hack Day which provides an interactive visualisation of global Netflix usage.

Netflix Sleep Tracker: A slightly more sinister example of a user-generated Netflix feature, this time from the August 2014 Hack Day.

All of this matters because the labour and culture of coders has significant implications for the television industry more broadly. The commercial and global success of Netflix, for instance, has positioned the company as a forerunner in the development of streaming technologies. Through various in-house projects and external collaborations, they have developed a reputation as algorithmic pioneers, and as the producers of tools that can and often are used more widely across the industry: tools that, to varying degrees, determine the structure, delivery and commissioning of television content. Given the open-source nature of these platforms, it’s reasonable to assume that a coder for the BBC might utilise code or solutions developed, and openly published, by Netflix (of course, the reverse is also possible). Even if the BBC prefer to code in-house, the collaborative nature of the open-source community and open-source software will surely produce a kind of leakiness of ideas, resources, and approaches to television that may therefore transcend specific institutional styles and practices.

I’ve only really scratched the surface of the interface here, but in doing it’s clear that beneath lies an exciting and interesting culture of coding through which, I would argue, we can learn a great deal about the mechanics and development of television today. [3]

Notes

[1] Having said that, my colleague James Bennett recently contributed to the ‘100 ideas for the BBC’ project in which he proposed that algorithms might also be designed in a way that fulfils public service obligations. Specifically, a public service algorithm could introduce viewers to content that they might not normally discover. This approach is distinct from the more conventional (and commercial) application of algorithms which tailor programming (and other commodities) based on an often relatively niche set of taste preferences. Instead, Bennett suggests that a public service algorithm might mimic the diversity of the television schedule – specifically the way that watching broadcast television often results in a chance encounter with a programme that we might not necessarily discover through a commercial algorithm. In a way, Bennett’s proposal speaks to both types of programming/programmers discussed in this post: the scheduler and the coder. You can see the complete list of ideas for the BBC here: https://opendemocracy.net/100ideasforthebbc/.

[2] It should be noted that the idea of “free” in the context of coding and the open-source/copyleft movement is a rather complicated one. For more on this, see Berry, David M. (2008) Copy, Rip, Burn: The Politics of Copyleft and Open Source. London: Pluto Press.

[3] Tiziana Terranova’s work on labour in the digital economy might offer a more systematic and rigorous way in which we can think through some of these dynamics in relation to the labour of television’s new programmers. See Terranova, T. (2000) ‘Free Labor: Producing Culture for the Digital Economy’, Social Text 63, Vol. 18(2): pp-33-58.

 

JP Kelly is a lecturer in film and television at Royal Holloway, University of London. He has published work on the emerging economies of online TV in Ephemeral Media (BFI, 2011) and on television seriality in Time in Television Narrative (Mississippi University Press, 2012). His current research explores a number of interrelated issues including the development of narrative form in television, issues around digital memory and digital preservation, and the relationship between TV and “big data”.