ABSTRACT
The current, widespread dissemination of algorithms represents a double challenge for both our society and the social sciences tasked with studying and making sense of them. Algorithms have expanded and woven their logic into the very fabric of all social processes, interactions and experiences that increasingly hinge on computation to unfold; they now populate our everyday life, from the sorting of information in search engines and news feeds, to the prediction of personal preferences and desires for online retailers, to the encryption of personal information in credit cards, and the calculation of the shortest paths in our navigational devices. In fact, the list of things they can accomplish is rapidly growing, to the point where no area of human experience is untouched by them-whether the way we conduct war through ballistic missile algorithms and drones, or the manner in which we navigate our love lives via dating apps, or the way we choose how to dress by looking at weather forecasts. Algorithms make all of this possible in a way that initially appears disarmingly simple. One way to approach algorithms is through Kowalski’s now classic definition: “Algorithm = Logic + Control” (1979). Using both simple and complex sorting mechanisms at the same time, they combine high-level description, an embedded command structure, and mathematical formulae that can be written in various programming languages. A wide variety of problems can be broken down into a set of steps and then reassembled and executed or processed by different algorithms. Hence, it is their versatility that constitutes their core capability and power, which extends far beyond the mathematical and computer sciences. According to Scott Lash, for instance, “a society of ubiquitous media means a society in which power is increasingly in the algorithms” (2007, 71), an idea echoed by Galloway when he states that “the point of power today resides in networks, computers, algorithms, information and data” (2012, 92). Yet, it is imperative to remain cautious with such formulations, and their tendency to be too critical, too quickly. While it may capture important challenges that society faces with ‘the rise of the algorithm,’ it can also provide something of a teleological or deterministic “seductive drama,” as Zietwitz has recently warned us (2016, 5). Algorithms can actually be considered less sovereign than mundane in this regard-that is, again, deeply rooted in the fabric of society. Rather than being omnipotent, they are oftentimes ambiguous and quite messy. What is crucial,
then, is to bring into question how, and especially why, the apparent simplicity of algorithms is in fact inseparable from their complexity, in terms of their deployment and multiple, interrelated ramifications. These are epistemological as well as ontological interrogations, confronting not only the social sciences but society at large. As both a known unknown and an unknown known, the sorting mechanism that is the algorithm still needs some sorting out. This introduction is certainly not the first to stress the inherent difficulty of shedding light on algorithms. Seaver, for instance, observes how they “are tricky objects to know” (2014, 2), while Sandvig insists on “the complexity of representing algorithms” (2015, 1; see also Introna 2016; Barocas et al. 2013). Conceptually perspicacious as they are, these arguments do not, however, foreclose the need to understand the extent of such invisibility and inscrutability. On the surface, it is often the ‘black box’ nature of the algorithms that is first evoked, namely that they are incredibly valuable patented trade secrets for companies such as Amazon, Google, Facebook, and the like. If they were revealed to noninsiders, they would eo ipso be ruined. Or at least so we are told by numerous technical, economic, legal, and political experts (Pascale 2015). This is where things noticeably start to get more serious and profound. There is not one box, but multiple boxes. The opacity of algorithms is more precisely expressed in different forms of opacity, all of which, in specific ways, are contingent on the inbetweenness of a plethora of actors, both human and non-human. While a few commentators have remarked upon the plural character of such opacity (Burrell 2016; Morris 2015), the fact remains that each and every algorithm can only exist in rich and dense, if not tense, environments. This is the inherently messy, vivid, and dynamic nature of algorithms, which explains why they are ultimately so challenging to study. As Kitchin puts it, “creating an algorithm unfolds in context through processes such as trial and error, play, collaboration and negotiation” (2014, 10). The latter term is of particular interest here: “negotiation” refers to the very condition of possibility/difficulty of algorithms. On the most fundamental level, they are what one can call anthropologically entrenched in us, their creators and users. In other words, there is a “constitutive entanglement” where “it is not only us that make them, they also make us” (Introna and Hayes 2011, 108). Indeed, the problem with such mutual imbrication is that algorithms cannot be fully ‘revealed,’ but only unpacked to a certain extent. What is more, they always find themselves temporally entrenched, so to speak. They come to life with their own rhythm, or, to use Shintaro Miyazaki’s description in this volume, “they need unfolding, and thus they embody time” (p. 129). Another metaphor that proves useful in this regard is Latour’s idea of the cascade (1986, 15-16): algorithms follow a non- linear course, caught in constant changes, fluctuations, and deviations both large and small. Such changes may very well be hard to follow or may even be imperceptible from time to time. The most important point to make here is how practical and mundane they are. Again, they unfold in a state of incessant negotiation and in-betweenness; for all algorithms, as Seaver has noticed, there are “hundreds of hands reaching into them, tweaking and tuning, swapping out parts and experiencing with new arrangements” (2014, 10).
