Abstract
Over the last 30 years or so, human beings have been delegating the work of culture – the sorting, classifying and hierarchizing of people, places, objects and ideas – increasingly to computational processes. Such a shift significantly alters how the category culture has long been practiced, experienced and understood, giving rise to what, following Alexander Galloway, I am calling ‘algorithmic culture’. The purpose of this essay is to trace some of the conceptual conditions out of which algorithmic culture has emerged and, in doing so, to offer a preliminary treatment on what it is. In the vein of Raymond Williams’ Keywords, I single out three terms whose bearing on the meaning of the word culture seems to have been unusually strong during the period in question: information, crowd and algorithm. My claim is that the offloading of cultural work onto computers, databases and other types of digital technologies has prompted a reshuffling of some of the words most closely associated with culture, giving rise to new senses of the term that may be experientially available but have yet to be well named, documented or recorded. This essay, though largely historical, concludes by connecting the dots critically to the present day. What is at stake in algorithmic culture is the gradual abandonment of culture’s publicness and the emergence of a strange new breed of elite culture purporting to be its opposite.
Easter 2009 might well be remembered for Amazon.com’s having outshined Jesus Christ. That much was true on Twitter, at any rate, where, during that long weekend in April, a sudden influx of short missives about the online retailer propelled it to the Number 1 spot on Twitter’s trending topics list, unseating the Prince of Peace along the way (James, 2009b). As the Beatles learned back in 1966, however, ‘more popular than Jesus’ (as John Lennon had claimed of the band) is not necessarily an enviable position in which to find oneself. The hashtag to which the Twitterati directed tens of thousands of messages – #AmazonFail – indicated that something had gone terribly wrong with company. Why, they wondered, had Amazon apparently begun excluding gay and lesbian–themed books from its sales rankings, searches and bestseller lists?
Author Mark R Probst first brought the issue to widespread attention when, on Good Friday, he noticed that several gay romance books had lost their Amazon sales rankings, including his own novel, The Filly. Hoping the matter was a simple mistake, he wrote to Amazon customer service. The agent who emailed Probst explained that Amazon had a policy of filtering ‘adult’ material out of most product listings. Incensed, Probst (2009) posted an account of the incident on his blog in the wee hours of Easter Sunday morning, pointing out inconsistencies in the retailer’s policy. The story was subsequently picked up by major news outlets, who traced incidences of gay and lesbian titles disappearing from Amazon’s main product list back to February 2009 (Lavallee, 2009; see also Kellog, 2009; Rich, 2009).
In a press release issued on Monday afternoon, a spokesperson for Amazon attributed the fiasco to ‘an embarrassing and ham-fisted cataloging error’. More than 57,000 books had been affected in all, including not only those with gay and lesbian themes but also titles appearing under the headings ‘Health, Mind, Body, Reproductive and Sexual Medicine, and Erotica’ (quoted in James, 2009a; see also Rich, 2009). An Amazon technician working in France reportedly altered the value of a single database attribute – ‘adult’ – from false to true. The change then spread globally throughout the retailer’s network of online product catalogs, de-listing any books that had been tagged with the corresponding metadata (James, 2009b). This was not homophobia, Amazon insisted, but a slip-up resulting from human error amplified by the affordances of a technical system.
In the wake of the controversy, author and lesbian, gay, bisexual and transgender (LGBT) activist Larry Kramer observed: ‘We have to now keep a more diligent eye on Amazon and how they handle the world’s cultural heritage’ (quoted in Rich, 2009). Indeed, Amazon may have started as a retailer, but it has grown into an exemplar of the many ways human beings have been delegating the work of culture – the sorting, classifying and hierarchizing of people, places, objects and ideas – to data-intensive computational processes. 1 Amazon’s back-end data infrastructure is so vast, in fact, that in 2006 it began selling excess capacity to clients under the name Amazon Web Services. It also collects sensitive data about how people read through its Kindle e-book devices – which is to say nothing of how it profiles and then markets products to customers based on their browsing and purchasing patterns (Striphas, 2010). What one sees in Amazon, and in its kin Google, Facebook, Twitter, Netflix and many others, is the enfolding of human thought, conduct, organization and expression into the logic of big data and large-scale computation, a move that alters how the category culture has long been practiced, experienced and understood. This is the phenomenon I am calling, following Alexander R Galloway (2006), ‘algorithmic culture’. 2
The purpose of this essay is to trace one set of conditions out of which a data-driven algorithmic culture has developed and, in doing so, to offer a preliminary sense of what ‘it’ is. The overarching impulse here is historico-definitional, though there are many ways to execute such a project. One could focus on the propagation of ‘truthful’ statements (i.e. discourses) pertaining to algorithmic culture (Foucault, 1972), or map the sociological circuitry through which the concept has made its way through the world (Mannheim, 1955). Or instead, one could take an etymological tack in attempting to trace the origins of particular words, or adopt a philological thrust in trying to apprehend definitive usages of words in history.
While this essay combines elements of these approaches, it is inspired primarily by Raymond Williams’ (1983) work on keywords. This piece emphasizes moments of catachresis – instances of lexical ‘misuse’ that help concretize an alternative semantics for particular words and word clusters. These moments enable new or at least different ways of figuring reality through language, for example, in drawing what was long taken to be the conceptual sine qua non of qualitative human experience – culture – into the orbit of computational data processing (see, e.g. Kittler, 2006). It is a contention of this essay that the semantic dimensions of algorithmic culture (and also then of the related phenomena of big data, data mining and analytics, the themes of this special issue of European Journal of Cultural Studies) are at least as important as the technological ones, the latter, for perhaps obvious reasons, tending to command the spotlight. But as Williams (1983) noted, ‘some important social and historical processes occur within language’, giving rise to new existential territories that only later come to be populated by technical artifacts (p. 22; see also Striphas, 2014).
Moreover, a keywords approach is useful in apprehending latencies of sense and meaning that persist, insist and subsist in contemporary usage as ‘traces without … an inventory’ (Gramsci, 1971: 324; see also Seigworth, 2000: 237). Logging that inventory, as it were, allows one to not only situate algorithmic culture within a longer durée but also reflect on claims to objectivity and egalitarianism that are now made in its name. Beyond semantics, what is at stake in algorithmic culture is the gradual abandonment of culture’s publicness and thus the emergence of a new breed of elite culture purporting to be its opposite.
Keywords today
Gary Hall (2002) opens the final section of Culture in Bits with the line, ‘what if Richard Hoggart had had email?’ (p. 126). This is tantamount to asking, ‘what would the work of cultural studies’ canonical figures look like were it composed today, a time of ubiquitous digital computational technologies?’ Imagine, say, Raymond Williams (1958) were writing Culture and Society having to confront the #AmazonFail episode. How might he make sense of the entwining of culture, which he posited as a ‘court of human appeal’ (Williams, 1958: viii), and computational decision-making (see also Hallinan and Striphas, 2014)?
Williams’ (1983) original project was to show how culture, once a relatively obscure word in English-language usage, became ‘one of the two or three most complicated words’ by the start of the 20th century (p. 87). He did so by tracing semantic shifts across a network of terms, many of which formed the basis for his compendium, Keywords (Williams, 1976, 1983). The introduction to Culture and Society offers a more succinct version of the story, focusing on five words whose history and interconnection uniquely embodied ‘a general change in our characteristic ways of thinking about our common life’: industry, democracy, class, art and culture (Williams, 1958: xiii). With the first four, Williams established a set of semantic coordinates, which he then used to chart culture’s shifting meaning and importance: from a pre-modern understanding grounded in husbandry to a more capacious, modern view – ‘a thing in itself’, encompassing not only ‘the general body of the arts’ but also ‘a whole way of life, material, intellectual, and spiritual’ (p. xvi).
Spanning the years 1780–1950, Culture and Society is bookended by two major historical events, namely, the industrial revolution and the end of the Second World War. The latter helped precipitate another great transformation referred to variously as the computer revolution, the communications revolution, the cybernetics revolution and so on (Beniger, 1986: 4–5). Prescient as he was, it is doubtful Williams grasped the full significance of his endpoint. More likely, he chose 1950 because the date marked midcentury, the moment in which the symbolics of history and futurity mingle more or less freely. Still, one can see Williams (1958) grasping to understand new technological contexts in his reflections on communication appearing in the conclusion to Culture and Society (pp. 296, 300–304, 313–319). It was not until the publication of The Sociology of Culture, however, that Williams (1981) broached the relationship between culture, information and digital technologies – but then only in passing, in the work’s conclusion (pp. 231–232). 3 He may not have been able to work out a fully revised theory of culture per se, but he managed to lay important groundwork for assessing how the semantic – and hence practical and experiential – coordinates of culture had shifted since 1950.
We are still living in the midst of this shift, although the nascent trends and tendencies Williams tried to make sense of in the early 1980s are more coherent today. The #AmazonFail episode illustrates this point, underscoring the degree to which shopping, merchandizing and a host of other everyday cultural activities are now data-driven activities subject to machine-based information processing (Striphas, 2009: 81–110). Indeed, the incident would not have been intelligible, much less possible, without a reshuffling of the terms surrounding the word, culture. Those that Williams (1958) identified in Culture and Society remain important, to be sure, but in recent decades a host of others have emerged. An extended list might include analog, application, cloud, code, control, convergence, copy, data, design, digital, format, free, friend, game, graph, hack, human, identity, machine, message, mobile, network, noise, peer, platform, protocol, search, security, server, share, social, status, web and many more. 4 Like Williams, however, I want to single out a small group of terms whose semantic twists and turns tell us something about senses and meanings of the word culture that are available today, and also then about the politics of big data, data mining and analytics. Williams identified the first one – information; the other two – crowd and algorithm – are my own. 5
Information
If culture’s usage is unusually ‘complicated’, then information’s is equally contradictory. John Durham Peters (1988) describes its etymology as ‘a history full of inversions and compromises’ (p. 10). Like a moody teenager, it swings from specificity to generality, and from the empirical to the abstract. Yet, this range is also what makes information intriguing from the standpoint of algorithmic culture, which channels an older sense of the word that the Oxford English Dictionary or OED describes as, ‘now rare’ (‘Information’, n., n.d.; see also Peters, 1988: 11; Gleick, 2011).
When information enters the English language sometime around the 12th or 13th century CE, chiefly from Latin, the tension at the heart of the word is already becoming manifest. At this early stage, it operates in two main semantic registers: religion and law. The use that the OED claims is ‘now rare’ is the religious one, although it might be more apt to describe it as spiritual, even deific. Here, information denotes ‘the giving of form or essential character to something; the act of imbuing with a particular quality; animation’ (‘Information’, n., n.d.). This definition posits an irreducible connection between the shaping of something and the endowment with character, substance or life.
Information’s juridical definition derives from the codes of ancient Roman legal procedure. Broadly, it refers to ‘the imparting of incriminating knowledge’, and more specifically, in US law, to ‘an accusation or criminal charge brought before a judge without a grand jury indictment’ (‘Information’, n., n.d.). Although information here depends on and passes among incarnate human agents, this definition differs from the more recent understanding, ‘knowledge communicated concerning some particular fact, subject, or event’ (‘Information’, n., n.d.). In the legal sense, one does not refer to information per se but to something narrower in scope – ‘an information’, or even ‘information’. The corresponding verb form, ‘laying of information’, is a special type of communication – a speech act – whose outcome is to transform the innocent into the accused and to set forth social rituals intended to restore order in the wake of some disturbance. Here the definition comes closest to the religious sense of character- or quality-giving, although now the information ‘source’ is secular interaction.
The influence of early modern empiricism and idealism on the word information must not be underestimated. The definition to which I referred in passing, ‘knowledge communicated concerning some particular fact, subject, or event’, is indicative of the term’s encounter with these crosscurrents of early modern thought, for it posits information not as intrinsic quality or character but as extrinsic sense data. This bit of semantic drift is significant, underscoring how far the locus of information has shifted from pre-modern through early modern times and beyond. Although it continued to refer to a kind of existential work, divine or worldly, gradually, a more object-oriented definition sidelined this sense of the word.
The passive voice construction ‘knowledge communicated’ is important to dwell on, moreover, because it indicates that information – now conceived as a thing – always emanates from some source external to one’s self. In the framework of Immanuel Kant, it belongs to the noumenon, or the realm of unmediated sense data. This marks a significant departure from the spiritual and legal definitions, both of which locate information in the body vis-a-vis its incarnations, godly or performative. The object-oriented definition, on the other hand, inaugurates a process of abstracting information from the body; instead of being vested there, information becomes a separate raw material that must be given order vis-a-vis our cognitive faculties (‘Information’, n., n.d.).
The 20th century information theorist Norbert Wiener famously quipped that the natural world consists of ‘a myriad of To Whom It May Concern messages’ (quoted in Rheingold, 1985: 113), drawing a line back to the work of his early-modern forebears. They similarly imagined a world bombarding us with sensory input. This was a broken, not a direct line, however, resulting in an even more diffuse meaning for the word. If information were akin to a ‘To Whom It May Concern Message’, then it need not be directed to anyone in particular. More to the point, in Wiener’s formulation, it need not be directed to anyone at all.
Apropos, the stars of Wiener’s two major books on cybernetics and information are neither the brain nor the cognitive structures that purportedly allow people to make our way in the world. They are, instead, photoelectric cells and antiaircraft guns, and more utilitarian things like automatic door openers and thermostats (Wiener, 1954, 1961). In contrast to wind-up clocks and other simple mechanical devices, which function in a manner more or less unattuned to environmental conditions, these machines ‘must be en rapport with the world by sense organs’ and adjust their behavior according to the information they receive (Wiener, 1954: 33; see also pp. 21–22). In 1944, the physicist Erwin Schrödinger (1967 [1944]) argued that life ‘feeds on negative entropy’, meaning that life is nothing more and nothing less than a small pocket of order within a world abuzz with information (p. 70). Four years later, Wiener told a similar story but threw in a major plot twist. If machines possessed an appetite for information, then apparently information was not particular to human beings.
From the Second World War on, then, machines begin being seen not merely as useful things but as custodians of orderliness. Critical to their work was information, which Gregory Bateson (2000 [1971]) defined as ‘a difference which makes a difference’ (p. 315). Bateson, like Wiener, identified as a cyberneticist, so in one sense it should not surprise to find him defining information in terms of bits, or simple yes–no decisions. But in another sense, his definition may surprise. Bateson was a trained anthropologist and spouse of Margaret Mead, to whom he was married for 14 years. They had one child together, Mary Catherine, who also became a noted anthropologist. In a family so thick with interest in people and culture, it is telling that Bateson never bothered with the question ‘to whom?’ when he called information ‘a difference which makes a difference’. By the early 1970s, information was only residually the process by which people and things were endowed with substance, trait or character – in-formed, as it were. It had become, instead, a counter-anthropological leveler, smoothing over longstanding differences between humans and machines: Inform-uniform. James Gleick (2011) puts the matter succinctly: ‘it’s all one problem’ (p. 280).
In 1966, Michel Foucault concluded The Order of Things (1971 [1970]) by claiming that ‘man is an invention of recent date … [a]nd one perhaps nearing its end’ (p. 387). Six years later, Gilles Deleuze and Félix Guattari (1983) opened Anti-Oedipus by proclaiming that ‘everything is a machine’ – plant life, animal life, mechanical devices, electronic goods, economic activities, celestial bodies and more (p. 2). Sandwiched between them was Bateson, the an-anthropic anthropologist for whom cultural life becomes one type of information processing task among many. One can also see emerging the sense of cultural objects, practices and preferences as comprising a corpus of data (from the Latin, ‘something given’), albeit data that exceed the traditional view of the human sciences in the agnosticism toward the intended recipient. No longer would human beings hold exclusive rights as cultural producers, arbiters, curators or interpreters – a welcome development, perhaps, given the shame, disrespect and brutality elites have long exacted in the name of cultural difference. But what if the apparent uniformity between people and machines resulted in cultural practices and decision-making that were no better in-formed?
Crowd
The etymology of the word crowd is, like that of information, a study in polarity reversal. It entered the English language around the 15th century CE as an adaptation of verbs extant in Dutch, German and Frisian denoting pressuring or pushing. The English verb form ‘to crowd’ preserves this early meaning of the word, although in some contexts the element of physical force may be figurative rather than literal. The OED mentions that crowd was ‘comparatively rare down to 1600’, which means its rise roughly coincides with early modernity (‘Crowd’, n.d.). The noun form of the word has often been used interchangeably with mass, mob, multitude and throng to refer to large gatherings of people, generally in public, especially in urban settings. Frequently, it denotes impedance, inefficiency and frustration, as in the expressions ‘fighting the crowds’ and ‘three’s a crowd’. It also conveys individual anonymity and engaged inaction, as in the phrase, ‘a crowd of onlookers’. 6 For these reasons crowd has, until recently, harbored almost exclusively pejorative connotations.
Semantically, crowd comes fully into its own in the 19th century, becoming a mainstay of journalistic and scholarly attention. 7 Charles Mackay’s Extraordinary Popular Delusions and the Madness of Crowds, first published in Britain in 1841, is a key text in this regard. The book chronicles incidents in which, as Mackay (2001 [1841]) puts it, ‘whole communities suddenly fix their minds upon one object, and go mad in pursuit of it’ (p. ix). The list ranges from stock bubbles to hairstyles, catch phrases, slow poisoning, dueling, occult practices and the mania for tulips in 17th century Holland. It is a capacious book, yet one that offers surprising little in the way of explicit insight into crowds – at least beyond their guilt by association with practices and events that, for Mackay, evidenced the collective abandonment of reason. Instead, he seems to play to the conventional wisdom of the time: ‘Men, it has been well said, think in herds; it will be seen that they go mad in herds, while they only recover their senses slowly, and one by one’ (Mackay, 2001 [1841]: x).
But Mackay does not only play to the conventional wisdom – he also plays upon it. Preceding his statement about ‘thinking in herds’ is a passing reference to an analogous concept: the ‘popular mind’ (Mackay, 2001 [1841]: x). Terminologically, it is a small difference, but semantically it is a bait-and-switch. The verb phrase ‘thinking in herds’ would seem to designate an active, living process, albeit one in which any individual contribution registers diffusely. The noun form ‘popular mind’ largely elides that process, positing some overarching thing referring to everyone in general and no one in particular. And in this way, crowd’s etymology closely parallels that of information, which follows the term’s divestiture from the human body, its transformation into an immaterial object and its dispersal into the world.
This way of conceiving of crowds culminates in Gustave Le Bon’s (2002 [1895]) The Crowd: A Study of the Popular Mind. Le Bon offers something like the explanatory framework absent in Mackay. Le Bon’s book is occasioned by ‘the entry of the popular classes into political life’, whom he paints a vicious, unthinking horde: ‘History tells us that from the moment when the moral forces on which a civilisation rested have lost their strength, its final dissolution is brought about by those unconscious and brutal crowds known, justifiably enough, as barbarians’ (Le Bon, 2002 [1895]: xii–xiii; see also Arnold, 1993 [1869]).
Le Bon’s book has been read, understandably, as an attack on crowds (see, for example, Milgram and Toch, 2010; Surowiecki, 2004). It is one, to be sure, and an elegy for the decline of privileged minority rule akin to Edmund Burke’s (1999 [1790]) Reflections on the Revolution in France. Yet, there is a tone of resignation evident in Le Bon’s prose, suggesting a kind of begrudging acceptance of the emerging political realities of the time: ‘The age we are about to enter will in truth be the
Whether by default or by design, Le Bon was drawing on a subterranean line of thinking about crowds. This line developed in the overlap of Classical Liberalism and the Scottish Enlightenment and received its most enduring expression in the work of Adam Smith. It was Smith (1977 [1776]) who, in An Inquiry into the Causes of the Wealth of Nations, struggled to make sense of apparently spontaneous economic activities whose outcome was – in Le Bon’s words – ‘admirably organised production’. Yet, the figure of the crowd is noticeably absent from Smith. In fact, the word crowd appears only four times in his 375,000 word magnum opus, and only then in verb form. His figure is a different one, and of a different kind, although it performs rhetorical work comparable to Le Bon’s ‘genius’ crowd. This is the famous ‘invisible hand’, which, in Smith’s (1977 [1776]) view, aligns the interests of individual economic actors with the needs of a society as a whole (p. 477).
Mysterious, ghostlike, the ‘invisible hand’ is essentially a deus ex machina of economic activity, and in this regard it is not too far removed from the spiritual sense of information mentioned earlier. In the 20th century, Friedrich A Hayek would make the link more explicit, helping to bolster the more affirmative view of crowds nascent in both Smith and Le Bon. The key work here is Hayek’s (2007 [1944]) Road to Serfdom, published in 1944, arguably the strong state’s high-water mark in both Europe and the United States. Hayek believed there ought to be some force to which was assigned the task of holding the state in check; for him, that force was the economic sphere. Hence, his desire to strip the state of the responsibility of economic planning and to leave the task of coordinating economic activities up to individual actors dispersed far and wide (Hayek, 2007 [1944]: 232). Instead of positing that coordination resulted from the arcane workings of an invisible hand, Hayek stressed the crucial role that information – his word – played in choreographing this intricate group dance, particularly through the price system (Hayek, 2007 [1944]: 95).
Like Smith, Hayek had little to say about crowds per se. His understanding of the individual, however, harkened back to the earliest English-language sense of crowd as the exertion of force on others. And with this, he helped to usher the idea of the intelligent, constructive crowd more fully into view. He was not alone in this endeavor. In 1965, the economist Mancur Olson (1971), a friend of Hayek, refuted the claim that groups were intrinsically stupid and irrational by describing the hidden ‘logic’ underlying collective action. 8 So, too, with sociologist Stanley Milgram, whose early work on obedience to authority was given subtlety and dimension in his later work on crowds, where he dismantled the view that crowds caused otherwise mindful people to become deluded (Milgram, 2010; Milgram and Toch, 2010). Finally, inasmuch as he was Hayek’s ideological opposite, we must nonetheless reckon with the contributions Raymond Williams made to the redemption of crowds. The conclusion to Culture and Society is an extended critique of the notion of masses-as-mob, culminating in the insight, ‘there are in fact no masses; there are only ways of seeing people as masses’ (Williams, 1958: 300). The alternatives Williams proposes – ‘community’ or ‘common culture’ – bear an uncanny resemblance to the sense of crowd I have been tracing here: a ‘complex organization, requiring continual adjustment and redrawing’; denying the individual the possibility of ‘full participation’, while still granting her or him a modicum of effect or influence; and incapable of being ‘fully conscious of itself’ (Williams, 1958: 333–334).
It is this set of positive connotations that crystallizes into contemporary terms like ‘crowdsourcing’, ‘crowd wisdom’ and a host of cognates, all of which have entered popular usage over the last two decades or so: ‘hive mind’, ‘collective intelligence’, ‘smart mobs’, ‘group genius’ and more (see, for example, Howe, 2008; Jenkins, 2006; Kelly, 1995; Levy, 1999; Rheingold, 2002; Sawyer, 2007; Shirky, 2008; Surowiecki, 2004; Tapscott and Williams, 2006). The translation between then and now is hardly perfect, owing to the diverse traditions out of which this vision of crowds has emerged, which is to say nothing of the technological transformations that have occurred over the last century. Indeed, when Williams (1958) wrote about the ‘solidarity’ necessary to sustain a ‘common culture’ (pp. 332–338), could he have anticipated the degree to which, today, that solidarity would be forged computationally? And what then to make of the redemption of crowds, when proprietary computer platforms have become the major hubs for interaction online? 9
Algorithm
Compared to information and crowd, algorithm is a less obvious keyword by means of which to make sense of culture today. If the former two terms could be considered dominant, or prevalent, as judged by their popular usage, then the latter would best be described as emergent, or restricted, though tending in the direction of conventionality. Yet, as James Gleick (2011) puts it in The Information, ‘[t]he twentieth century gave algorithms a central role’ (p. 206). 10
Algorithm comes to modern English from Arabic by way of Greek, medieval Latin, Old French and Middle English. Historically, it has maintained close ties to the Greek word for number, arithmós (αριθμός), from which the English form arithmetic is derived. Algorithm’s most common contemporary meaning – a formal process or set of step-by-step procedures, often expressed mathematically – flows from this connection, although the OED insists it is a point of etymological ‘perversion’ (‘Algorism’, n., n.d.). In fact, the word is a ‘mangled transliteration’ of the surname of a 9th century mathematician, Abū Jafar Muḥammad ibn Mūsā al-
The confusion stems mainly from two key mathematics texts attributed to al-
The other key work is al-
Still, this is not the only or most interesting sense of the term. Algorism’s semantic context includes a range of secondary meanings that are key to making sense of algorithmic culture. Among the most important is its close association with zero (Smith and Karpinski, 1911: 58). The word zero comes from śūnya, Sanskrit for ‘void’, which migrates into Arabic as ṣifr, meaning ‘empty’, the root from which the modern English language form cypher derives (Smith and Karpinski, 1911: 56–57). Thus, it is no coincidence that the phrase ‘cypher in algorism’ was long used interchangeably with the word zero; sometimes cypher would be used to designate any of the Arabic numerals, making it synonymous with algorism (‘Algorism’, n., n.d., ‘Cipher, Cypher’, n., n.d.). Moreover, until the middle of the 19th century, cypher, like zero, could refer to a placeholder – often in a derogatory sense, indicating a ‘worthless’ person (‘Cipher, Cypher’, n., n.d.). This was alongside what has emerged today as cypher’s more commonplace definition, namely, a secret code or the key by means of which to crack it.
So, on the one hand, we have algorithms – a set of mathematical procedures whose purpose is to expose some truth or tendency about the world. On the other hand, we have algorisms – coding systems that might reveal, but that are equally if not more likely to conceal. The one boasts of providing access to the real; the other, like an understudy, holds its place. Why in the early 20th century did algorithm become preferred over algorism, so much so that the latter form is now all but an archaism?
In a word, information. The touchstones in this regard are two landmark papers, both written by engineers who worked at Bell Laboratories in the United States. The first, Ralph Hartley’s ‘Transmission of Information’, appeared in 1928. The second, Claude E Shannon’s ‘Mathematical Theory of Communication’, appeared in 1948. Hartley’s paper was noteworthy for many reasons, chiefly technical, but perhaps his most audacious move was to subsume communication under the rubric of information. He states, ‘In any given communication the sender mentally selects a particular symbol … As the selections proceed more and more possible symbol sequences are eliminated, and we say that the information becomes more precise’ (Hartley, 1928: 536). Hartley thus conceived of communication as a procedural activity – a game of chance in which the stake was information, or the likelihood of achieving identity of message within and across a specific context of interaction. He was followed in his work by Shannon, who raised the stakes on Hartley’s theory.
One of the differences between Hartley and Shannon’s papers was that Shannon placed significantly less faith in the process of communication. For Hartley, it was a relatively placid affair in which the revelation of symbols led to understanding and order. For Shannon, communication was a tumultuous affair consisting of such a complex ensemble of determinations, both past and present, that disorder was the state to which it naturally tended. Communication thus occurred within a context rife with uncertainty, or in the language of information theory, entropy; order could not be taken for granted but instead needed to be engineered. Shannon’s problem was thus to figure out how to parse signal and noise, and thereby increase the odds that the system would achieve a sufficient degree of order. Hence, he needed to devise a set of procedures – an algorithm, if you will, although he did not use the term specifically – capable of dealing with the cascade of determinations that governed communicative encounters. Shannon may have believed he was developing a ‘Mathematical Theory of Communication’, but in fact he produced among the first algorithmic theories of information.
It is worth mentioning that Shannon was not just a talented electrical engineer, he was also a world-class cryptographer, having worked on several government-sponsored ‘secrecy’ projects at Bell Labs throughout the Second World War. During that time, he produced a lesser-known paper, originally classified, entitled ‘A Mathematical Theory of Cryptography’ (Shannon, 1945). Shannon operated, in other words, at the junction point of algorithms and algorisms. Or, as he described his work on communication and cryptography many years later, ‘they were so close together you couldn’t separate them’ (quoted in Kahn, 1967: 744; see also Gleick, 2011: 216–218). Indeed for Shannon, communication in the ordinary sense of the term was nothing other than a special, simpler case of cryptography, or of ciphering and deciphering. Both in his view consisted of signals and noise stuck in a dizzying, entropic dance, along with telling redundancies that, if exploited using the right mathematics, could mitigate much of the turmoil and thereby point the way toward order (Rheingold, 1985: 119). What Shannon was essentially proposing in his work, then, was the use of algorithms to attenuate algorisms.
Conclusion
I have tried my best to connect as many of the dots between the words information, crowd and algorithm as possible. I realize, of course, that there are a great many dots left to be connected – something that is inevitable in a synoptic piece such as this, which marshals a version of what Franco Moretti (2005) has called a method of ‘distant reading’ (p. 1). I am well aware of the prejudices of this research, moreover, particularly the privileging of the work and experiences mostly of White men of European descent. The purpose in doing so is not to exalt them. Instead, I am trying to tell a story about word-worlds or ‘universes of reference’ they helped call into being by using specific terms unconventionally (Guattari, 1995: 9). Having said that, I want to say a few more words about how the conceptual history I have presented here connects up with culture, and also then big data, data mining and analytics.
Matthew Arnold’s (1993 [1869]) Culture and Anarchy is infamous for having defined culture in elite terms, as ‘the best which has been thought and said’ (p. 190). Elsewhere in the book Arnold refers to culture as ‘sweetness and light’, thereby defining a model disposition for the small group of apostles whom he imagined would, like him, spread the gospel of culture. Yet, there is a third sense of culture present in the book which, while hardly overlooked, is overshadowed by these two more quotable definitions. He refers to culture as ‘a principle of authority, to counteract the tendency to anarchy which seems to be threatening us’ (Arnold, 1993 [1869]: 89). For Arnold, culture as authoritative principle meant a selective tradition of national – specifically English – art and literature that would, in his view, create a basis for national unity and moral uplift at a time when heightening class antagonism was threatening English society.
This latter definition – culture as authoritative principle – is the one that is chiefly operative in and around algorithmic culture. Today, however, it is not culture per se that is a ‘principle of authority’ but increasingly the algorithms to which are delegated the task of driving out entropy, or in Arnold’s language, ‘anarchy’. You might even say that culture is fast becoming – in domains ranging from retail to rental, search to social networking, and well beyond – the positive remainder resulting from specific information processing tasks, especially as they relate to the informatics of crowds. And in this sense, algorithms have significantly taken on what, at least since Arnold, has been one of culture’s chief responsibilities, namely, the task of ‘reassembling the social’, as Bruno Latour (2005) puts it – here, though, using an array of analytical tools to discover statistical correlations within sprawling corpuses of data, correlations that would appear to unite otherwise disparate and dispersed aggregates of people (see, e.g., Hallinan and Striphas, 2014).
I suggested at the outset of this article and at points along the way that part of what is at stake in algorithmic culture is the privatization of process: that is, the forms of decision-making and contestation that comprise the ongoing struggle to determine the values, practices and artifacts – the culture, as it were – of specific social groups. Tarleton Gillespie (2011) has explored this issue in relationship to Twitter’s trending topics, noting how the company’s black box approach creates all sorts of mystifications about how it adduces topical importance. ‘The interesting question is not whether Twitter is censoring its Trends list’, he writes. ‘The interesting question is, what do we think the Trends list is … that we can presume to hold it accountable when we think it is “wrong”’ (Gillespie, 2011). His point is that Twitter and its kin bandy about in what one might call the algorithmic real, where placeholders for trending topics and the like are presented as if they were faithful renderings of reality. But the issue is even more complex than this. Gillespie (2011) adds that ‘[w]e don’t have a sufficient vocabulary for assessing the algorithmic intervention in a tool like Trends’, an observation that underscores just how deeply entangled are questions of language, technology, big data, analytics and political economy. This is all the more reason to broach the issue of the privatization of cultural decision-making only after having explored the semantic context, or keywords, that frame the issue in the first place.
In brief, consider the product recommendations one sees on Amazon. These, says the retailer, are the result of one’s browsing and purchasing histories, which are correlated with those of Amazon’s millions of other customers – a crowd – to determine whose buying patterns are similar to one’s own. You, too, might like what this select group has bought, and vice-versa – a process Amazon calls, ‘collaborative filtering’. Google reportedly works in a similar way. Although the company has moved far beyond its original ‘PageRank’ algorithm, which measured the number of links incoming to a website to determine its relative importance, it still leverages crowd wisdom to determine what is significant on the web. As Wired magazine explained in 2010, PageRank has been celebrated as instituting a measure of populism into search engines: the democracy of millions of people deciding what to link to on the Web. But Google’s engineers … are exploiting another democracy – the hundreds of millions who search on Google, using this huge mass of collected data to bolster its algorithm. (Levy, 2010: 99–100)
All this makes algorithmic culture sound as if it were the ultimate achievement of democratic public culture. Now anyone with an Internet connection gets to have a role in determining ‘the best that has been thought and said’! I am tempted to follow here by saying, ‘much to the chagrin of Matthew Arnold’, but I am not convinced that algorithmic culture is all that far removed – in spirit, if not execution – from a kind of Arnoldian project. Despite the populist rhetoric, I believe we are returning to something like his apostolic vision for culture. The Wired magazine story from which I just quoted also says this: ‘You may think [Google’s] algorithm is little more than a search engine, but wait until you get under the hood and see what this baby can do’ (Levy, 2010: 98). The problem is, thanks to trade secret law, nondisclosure agreements and noncompete clauses, virtually none of us will ever know what is ‘under the hood’ at Amazon, Google, Facebook or any number of other leading tech firms. As Gillespie (2007) is fond of saying, you cannot look under a hood that has been ‘welded shut’ (p. 222).
All this harkens back to the oldest sense of information – where some mysterious entity is responsible for imbuing people and objects with shape, quality or character. I do not mean to downplay the role that crowds play in generating raw data. Yet, it seems to me that ‘crowd wisdom’ is largely just a stand-in – a placeholder, an algorism – for algorithmic data processing, which is increasingly becoming a private, exclusive and indeed profitable affair. This is why, in our time, I believe that algorithms are becoming decisive, and why companies like Amazon, Google and Facebook are fast becoming, despite their populist rhetoric, the new apostles of culture.
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Notes
Biographical note
.