cultural studies of data mining: an introduction

16
European Journal of Cultural Studies 2015, Vol. 18(4-5) 379–394 © The Author(s) 2015 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/1367549415577395 ecs.sagepub.com EUROPEAN JOURNAL OF Cultural studies of data mining: Introduction Mark Andrejevic The University of Queensland, Australia Alison Hearn University of Western Ontario, Canada Helen Kennedy University of Sheffield, UK Over the past 2 years, the total amount of data about everything from the humidity of shipping crates, toilet flushes in shopping malls or tweets about Justin Beiber exceeded the total amount yet recorded in human history – equivalent to a zettabyte of data or sextillion bytes and growing (Shaw, 2014). Given this, it is now axiomatic to claim that we are in the ‘age of big data’ and are witnessing a quantitative (and perhaps qualitative) ‘revolution’ (Lohr, 2012) in human knowledge, driven by accompanying forms of data mining and analytics. New analytical methods and businesses seeking to monetize this explosion of data emerge daily. Often offered in black-boxed proprietary form, these companies and their analytic methods promise to help us gain insight into public opinion, mood, networks, behaviour patterns and relationships. Data analytics and machine learning are also ostensibly paving the way for a more intelligent Web 3.0, promising a more ‘productive and intuitive’ user/consumer experience. Data analytics involve far more than targeted advertising, however; they envision new strategies for forecasting, targeting and decision-making in a growing range of social realms, such as marketing, employment, education, health care, policing, urban planning and epidemiology. They also have the potential to usher in new, unaccountable and opaque forms of discrimination and social sorting based not on human-scale narratives but on incomprehensibly large, and continually growing, networks of interconnections. Corresponding author: Helen Kennedy, University of Sheffield, Sheffield, S10 2TU, UK. Email: [email protected] 577395ECS 0 0 10.1177/1367549415577395European Journal of Cultural StudiesAndrejevic et al. research-article 2015 Introduction by guest on June 17, 2015 ecs.sagepub.com Downloaded from

Upload: westernu

Post on 27-Apr-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

European Journal of Cultural Studies2015, Vol. 18(4-5) 379 –394

© The Author(s) 2015Reprints and permissions:

sagepub.co.uk/journalsPermissions.navDOI: 10.1177/1367549415577395

ecs.sagepub.com

e u r o p e a n j o u r n a l o f

Cultural studies of data mining: Introduction

Mark AndrejevicThe University of Queensland, Australia

Alison HearnUniversity of Western Ontario, Canada

Helen KennedyUniversity of Sheffield, UK

Over the past 2 years, the total amount of data about everything from the humidity of shipping crates, toilet flushes in shopping malls or tweets about Justin Beiber exceeded the total amount yet recorded in human history – equivalent to a zettabyte of data or sextillion bytes and growing (Shaw, 2014). Given this, it is now axiomatic to claim that we are in the ‘age of big data’ and are witnessing a quantitative (and perhaps qualitative) ‘revolution’ (Lohr, 2012) in human knowledge, driven by accompanying forms of data mining and analytics. New analytical methods and businesses seeking to monetize this explosion of data emerge daily. Often offered in black-boxed proprietary form, these companies and their analytic methods promise to help us gain insight into public opinion, mood, networks, behaviour patterns and relationships. Data analytics and machine learning are also ostensibly paving the way for a more intelligent Web 3.0, promising a more ‘productive and intuitive’ user/consumer experience. Data analytics involve far more than targeted advertising, however; they envision new strategies for forecasting, targeting and decision-making in a growing range of social realms, such as marketing, employment, education, health care, policing, urban planning and epidemiology. They also have the potential to usher in new, unaccountable and opaque forms of discrimination and social sorting based not on human-scale narratives but on incomprehensibly large, and continually growing, networks of interconnections.

Corresponding author:Helen Kennedy, University of Sheffield, Sheffield, S10 2TU, UK. Email: [email protected]

577395 ECS0010.1177/1367549415577395European Journal of Cultural StudiesAndrejevic et al.research-article2015

Introduction

by guest on June 17, 2015ecs.sagepub.comDownloaded from

380 European Journal of Cultural Studies 18(4-5)

A well-developed cultural studies approach has an important role to play in consider-ing the social and political consequences of data mining and analytics. When every move we make online is tracked by privately owned corporations and the state, advertisements follow us around in material retail spaces, and even our sleep patterns become fodder for self-tracking (to gain ‘self-knowledge’), we cannot afford to limit our thinking about data analysis technologies by approaching them solely as communication media. Instead, we must see them as techno-economic constructs whose operations have important implications for the management of populations and the formation of subjects. As Web 3.0 and the big data it generates move inexorably towards predictive analytics and the overt technocratic management of human sociality, questions need to be asked about what kinds of data are gathered, constructed and sold; how these processes are designed and implemented; to what ends data are deployed; who gets access to them; how their analysis is regulated (boyd and Crawford, 2012) and what, if any, possibilities for agency and better accountability data mining and analytics open up.

Although cultural studies has always been good at picking up on significant cultural trends and, in some cases, at anticipating them, databases and data processing practices do not fit neatly within the traditional ambit of the field and pose some challenges to it. Data mining challenges conventional forms of narrative and representation, promising to discern patterns that are so complex that they are beyond the reach of human percep-tion, and in some cases of any meaningful explanation or interpretation. Moreover, as a signifying practice, data mining is less accessible than other forms of cultural represen-tation, such as the cultural texts and everyday life practices that have been the tradi-tional focus of cultural studies research. Data mining is not only a highly technical practice, it also tends to be non-transparent in its applications, which are generally pri-vately owned and controlled. As a result, it can be extremely difficult for independent researchers to gain access to data sets and analytic methods in order to critically assess them. Most often, with so little insight available into their production, we are left only to theorize their effects.

In his 1963 inaugural address at the University of Birmingham, Richard Hoggart cap-tured cultural studies’ original intentions when he observed that his motives (and by association, those of the emerging discipline) ‘for attending to the contemporary and especially “commercial” culture were to find ways of understanding the languages of developing popular media forms: television, magazines, and, especially, advertising’ (as paraphrased by Gray et al., 2007: 5). Tellingly, Hoggart’s speech also critically diag-nosed the forms of categorization and targeting that have since become a staple of com-mercial media research and marketing: ‘So much language is used not as exploration but as persuasion and manipulation’ he observed – manipulation that treats individuals ‘as bits, bits who belong to large lopsided blocks – as citizens, as consumers, as middle-aged middle-class parents … as voters or viewers, as religiously inclined car owners, as typi-cal teenagers or redbrick professors’ (Gray et al., 2007: 5). It is a formulation that antici-pates a tendency that has become increasingly pronounced throughout the last decades of the 20th century, as we have moved into the era of mass customization, targeting and personalization. Indeed, this trend has subsequently developed to the point that the man-ufacture of the atomized ‘bits’ that are the targets of customized messaging has become as significant as the messages themselves. So, as we tweet, post, like, share and Google

by guest on June 17, 2015ecs.sagepub.comDownloaded from

Andrejevic et al. 381

to generate meaning, related platforms and their analytics generate us as ‘bits’ in turn and deploy our communicative efforts to their own advantage.

Micro-targeting via data analytics and metricization is ‘the message’ in the current conjuncture. This development is reflected in the heightened popular attention to media technologies that function in a different register than the content-driven mass media: the rise of a fascination not just with the Internet and the mobile phone, but also with data-bases, their algorithms and analysts. Drawing on John Durham Peters (2013), we might describe this development as the rise of ‘logistical media’ – media whose content is not so much narratival or representational as organizational (p. 40). Peters (2013) refers to media such as ‘data processors’ broadly construed: media that ‘arrange people and prop-erty into time and space’ (p. 40). It is a formulation that captures the allocative character of data mining and the processes of segmentation, sorting and categorizing that emerge from the search for useful patterns.

To observe that the database as a research object fits more comfortably into the cate-gory of logistical media than traditional mass media is not to suggest that databases are ‘contentless’, but rather to note the shift away from interpretive approaches and mean-ing-making practices towards the project of arranging and sorting people (and things) in time and space. Consider, for example, Google’s oft-repeated response to concerns about its data mining of personal messages on its free email service: ‘no humans read your email’. The process of reading proper is not what matters in this approach to the medium. That is, content is not approached as content but as ‘metadata’ – information about the content that can be used to analyse and categorize individuals and groups of people, places and things. Messages are mined not to understand or interpret their intended or received content but to arrange and sort people and their interactions according to the priorities set by Google’s business model.

We might describe this process of ‘metadatification’ – whereby a message is reconfig-ured into data about itself – as the post-ideological or post-textual moment taken to its logical conclusion. That is, once the notion of a containable, interpretable and transmis-sible ‘dominant’ meaning is deconstructed beyond recognition, what we are left with is the circulation of affects, and eventually, from an instrumental point of view, their effects. This trajectory was anticipated, perhaps inadvertently, by John Fiske’s move towards a version of post-textualism in which, he states, ‘there are no texts, no audiences. There is only an instance of the process of making and circulating meanings and pleasures’ (Turner, 1992: 123). Intimations of these developments can also be found within semi-otic criticism itself, for example, in early discussions about promotionalism by Andrew Wernick. In his work, echoing Barthes, the logics of capitalist instrumentality prey para-sitically on modes of human representation, turning them towards politically interested ends. Anticipating the reductive instrumentality of metadatification and its predictive uses, Wernick (1990) argued that a promotional message is an ‘anticipatory advocate’ of its original referent, marked ‘not by what it says but by what it does’ (p. 184).

Perhaps not coincidentally, recent forms of social and cultural theory mirror develop-ments in big data analytics; new materialism, object-oriented ontology, post-humanism and new medium theory – all of which are coming to play an important role in digital media studies – de-centre the human and her attendant political and cultural concerns in favour of a ‘flat’ ontology wherein humans are but one node, and perhaps not the most

by guest on June 17, 2015ecs.sagepub.comDownloaded from

382 European Journal of Cultural Studies 18(4-5)

important, in complex networks of interactions and assemblages. Thus, analysis of the circulation of affects and effects rather than of meanings, content or representations, con-nected as they are to human-centred forms of meaning-making, has become a dominant trope in some influential current approaches to media. Such analyses tend to fashion themselves as anti-discursive in their rejection of a focus on representation and cognition and their turn towards bodies and things in their materiality (rather than their significa-tion). For example, Karen Barad (2003), whose theory of ‘agential realism’ has been influential in the development of ‘new materialist’ approaches recently taken up in media studies (and digital media studies in particular), argues that

Language has been granted too much power. The linguistic turn, the semiotic turn, the interpretive turn, the cultural turn: it seems that at every turn lately every ‘thing’ – even materiality – is turned into a matter of language or some other form of cultural representation. (p. 804)

Google, with its attempt to distance itself from reading, might well agree. By contrast, Graeme Turner (1992) argues that cultural studies is a formation in which ‘language looms as the most essential of concepts, either in its own right or through being appropri-ated as a model for understanding other cultural systems’ (p. 13). To line up the ‘cultural’ turn alongside the linguistic or semiotic one, as Barad does, is not to suggest that cultural approaches limit themselves solely to questions of textual representation, but that when they turn their attention to audiences, institutions, artefacts, economies and activities, they remain within the horizon of interpretation, explanation and narrative. We suggest that this adherence to the horizon of meaning is a strategic critical resource in the face of theoretical tendencies that reproduce the correlational logic of the database by focusing on patterns and effects rather than on interpretations or explanations.

A related tendency is manifest in the growing interest in German-inflected medium theory, as seen in the growing influence of Friedrich Kittler. As Jeremy Packer (2013) puts it, Kittler’s work represents a ‘cold-turkey cure for hermeneutic and ideological fixation’ (p. 295). In John Durham Peters’ words, the German theorist ‘has no use for the category of “the human” or “experience”’ (Kittler, 2010: 5). For him, the conditions of the latter are subordinated to core media processes in a particular historical period: namely, the collection, storage and processing of data. The virtue of such an approach, according to Packer (2013), is that it addresses a shift in the way digital media operate, a shift that he describes in distinctly logistical terms: ‘digital media power is first and fore-most epistemological, not ideological. It is computational’ (p. 297). Packer’s words pro-vide a theoretical echo of former Wired magazine editor Chris Anderson’s (2008) post-hermeneutic paean to ‘the end of theory’ ushered in by the advent of big data:

Out with every theory of human behavior, from linguistics to sociology … Who knows why people do what they do? The point is they do it, and we can track and measure it with unprecedented fidelity. With enough data, the numbers speak for themselves. (n.p.)

The various strands of theory referenced above are diverse movements on their own, and there are significant differences among them, but they partake to various degrees in countering an emphasis on the discursive. It is not difficult to discern certain affinities between tendencies in these approaches and a ‘datalogical’ turn more generally: that is,

by guest on June 17, 2015ecs.sagepub.comDownloaded from

Andrejevic et al. 383

a growing interest in what automated forms of information processing can reveal about the object formerly known as ‘content’. This affinity, in turn, raises questions regarding the critical purchase of such approaches, including whether they are adequate to the task of critiquing digital media practices and the social relations they rely upon and/or repro-duce. Or, on the contrary, are these theoretical tools prone to buy in to the logics of metadatification? Google may not ‘measure ideology’, but does this mean that ideology and the need for a critical political analysis of it cease to exist? More pointedly, as Alex Galloway (2013) asks,

Why, within the current renaissance of research in continental philosophy, is there a coincidence between the structure of ontological systems and the structure of the most highly evolved technologies of post-Fordist capitalism? […] Why, in short, is there a coincidence between today’s ontologies and the software of big business? (p. 347)

Taken in the spirit of a needed corrective to a perceived over-emphasis on discourse and language, the re-assertion of the extra-discursive serves as a potentially useful prov-ocation. Elizabeth Grosz (2004) (whose work has also been influential in the formation of ‘new materialism’), for example, issues a call for

a remembrance of what we have forgotten – not just the body, but that which makes it possible and which limits its actions: the precarious, accidental, contingent, expedient, striving, dynamic status of life in a messy, complicated, resistant, brute world of materiality. (p. 2)

But, while it is true that the tendency towards textualism or constructivism is not for-eign to cultural studies, neither is its critique. Almost a quarter century ago, Graeme Turner (1992) noted in response to the celebration of the possibilities of textual poly-semy (as a form of liberation from the alleged tyranny of a text’s ‘dominant’ meaning):

The obvious limitation to the progressive effect of this multitude of textual possibilities is that … making over the meaning of a television program may be much easier than climbing out of a ghetto, changing the color of one’s skin, changing one’s gender … (p. 122)

Indeed, in part to address the limitations of textualism, cultural studies has always incorporated a much broader scope of research objects, including cultural institutions, audiences, subcultures and the practices of everyday life.

We argue that cultural studies is well positioned to challenge the dangers that can come from an over-correction of a focus on discourse (whether in the realm of audiences, artefacts or texts). In promising to push beyond narrowly human-centred concerns, for example, these recent theoretical positions threaten to adopt a ‘view from nowhere’ in which the goal of comprehensiveness (the inclusion of all components of an endless network of inter-relations) tends towards a politically inert process of specification in which structures of power and influence dissipate into networks and assemblages. As in the case of data mining, infinite specification results in a deadlock of interpretation. The various parties to an interaction become subsumed to its overall effects so that all we can say, in the end, is, ‘it’s complicated’. Jane Bennett’s (2009) version of ‘vibrant’ material-ism, for example, displaces simple narratives of causality with the recognition of the constantly unfolding network of contributors to any event or outcome. Likewise, in data

by guest on June 17, 2015ecs.sagepub.comDownloaded from

384 European Journal of Cultural Studies 18(4-5)

mining, humans take their place within the growing menagerie of data about inanimate objects, forces, interactions, flora and fauna. Data become not so much symbols to be understood and differentiated as inputs to be sorted. The multiplication of variables and the automaticity of the algorithm displace the roles of interpretation, comprehension and, most significantly, politics. As Galloway (2012) has also argued, these theories remove ‘the point of decision from the people (demos) to the object world at large’. And, because there is no way of generating a dynamics of contestation and argument in a flat ontology of ever proliferating relations or objects, these theories leave little room for judgment – obviously a crucial element in any political project.

Our commitment to cultural studies’ focus on politics and power (and relatedly, dis-course and representation) is not meant as a ‘retro’ or rear-guard action counter-poised to approaches that have forsaken or surpassed the cultural, but rather a (re-)assertion of its unsurpassability. This is not to say that there is nothing beyond culture, but that the sig-nificance of the assertion of such a beyond is inescapably caught up in cultural logics. A critical analysis of data mining, then, must involve an interrogation of the embrace of a post-cultural imaginary within contemporary media theory – that is, such an analysis must step back and explicitly situate data mining culturally, politically and economically. Such an approach presses against the notion that it would be either possible or desirable to permanently bracket the distinctive character of the cultural (and associated concerns with the political implications of representation, narration, explanation and interpreta-tion) in the complementary fantasies of automated tracking and sense-making that dis-pense with the need for comprehension and the theoretical analysis of assemblages in their infinite complexity. In the end, the cultural, representational and political cannot be surpassed by theory or algorithm, no matter how complex, unknowable or compelling they might be. Indeed, complex and compelling theories and algorithms are products of history and subject to cultural logics, no matter how vociferously they might claim it to be otherwise. If cultural studies is ‘intellectual political work’ or ‘a practice which always thinks about its intervention in a world in which it would make some difference, in which it would have some effect’ (Hall, 1996: 275), then the cultural studies of data mining and data analytics must attend to questions of power, subjectivity, governance, autonomy, representation, control and resistance which have always been central to enquiry in this field. The articles in this special issue do just that.

One of the pressing tasks of such an endeavour must necessarily be an engagement with the social relations that render data mining itself opaque. In this regard, our con-cerns align with those strands of cultural studies that locate their roots in a critique of political economy. As boyd and Crawford (2012) have argued, the creation of huge data sets ushers in a new digital divide between those with access to data (and the resources for making sense of it) and those without, as elite commercial companies like Google, Facebook and Amazon have the best access to data, as well as the best tools and methods to make sense of them (Williamson, 2014). These companies (and others like them, such as OKCupid) can experiment constantly for commercial purposes on their (sometimes very large) user populations without notifying them, whereas those supposed to be oper-ating in the public interest are effectively locked out or provided only limited access. This state of affairs seems unlikely to change any time soon, given the economic and competitive incentive to maintain the proprietary character of data mining. Data mining

by guest on June 17, 2015ecs.sagepub.comDownloaded from

Andrejevic et al. 385

is capital-intensive insofar as it relies on building the networks and platforms to collect data, the server farms to store it and the computing power to analyse it. If one of the casu-alties of a fascination with the ‘virtual’ character of software, applications, code and data has been a shift away from political-economic approaches to the media, the reliance of data mining on capital-intensive infrastructure reminds us of the ongoing salience of political-economic concerns. These debates inform the articles in this collection.

While developments in data collection and processing appear complex, opaque, ine-luctable and unassailable, it is crucial to remember that they comprise just another phase in human technological development, constituted by and constitutive of dominant eco-nomic and cultural systems. This, of course, is the insight that cultural studies approaches bring to bear on the study of the data mining practices and discourses. These practices may promote and exploit the dream of total knowledge and future prediction, but such initiatives (at least in their proprietary form) call out to be examined in their mundane materiality, in their contexts of production and use, for the ways they are helping to inau-gurate new monopolies of knowledge, systems of governance, forms of subjectivity, and social and cultural divisions and stratifications. No matter the kinds or numbers of machines we have wrought, in the end, we remain humans attempting to make meaning with whatever forms of communication we have available. The collection of essays pre-sented here takes up these concerns by developing critical approaches to emerging tech-nologies and practices of data mining and analysis; they historicize, probe, complexify, humanize, identify contradictions and modes of resistance, and trace the ways our cul-tural worlds are changing in the wake of data analytics. As such, they address themes central to both academic discussions of digital media and public concerns about emerg-ing forms of data-driven sorting, prediction and decision-making.

The first group of essays provides historical and macro-assessments of the cultural worlds we are making as a result of big data and data analytics. Each essay challenges the celebratory presentism with which many of these developments are often met and highlights the specific kinds of inversions and contradictions the new regime of ‘algo-rithmic culture’ can engender. They remind us that technological forms, and the rhetorics and analytic practices that accompany them, do not come from nowhere – they have histories, which shape and condition them, and inevitably bear the marks of the cultural, social and political conditions surrounding their production and implementation. As such, and most crucially, these essays remind us that the development of big data analyt-ics conditions new forms of political domination and resistance.

In the first of these essays, ‘Algorithmic Culture’, Ted Striphas examines the ways in which computational processes of sorting, classifying and hierarchizing of people, places, objects and ideas (e.g. on sites like Amazon and Netflix or the results generated by Google, Bing and others) have profoundly altered the way ‘culture’, as a category of experience, is now practised, experienced and understood. Following Frederic Jameson’s famous dictum ‘always historicize’, and in the spirit of Raymond Williams’ Culture and Society (1958), Striphas examines the conceptual ‘conditions out of which algorithmic culture has developed’ by addressing a small group of terms whose bearing on the mean-ing of the word culture has become unusually strong in recent years. Williams identified the first one – information – in his later work, The Sociology of Culture (1981); the other two – crowd and algorithm – are Striphas’ own. Arguing that the ‘semantic dimensions

by guest on June 17, 2015ecs.sagepub.comDownloaded from

386 European Journal of Cultural Studies 18(4-5)

of algorithmic culture … are at least as important as the technological ones’, Striphas takes us through the histories of these contemporary keywords, highlighting their defini-tional contradictions, expansions and contractions, and identifying the ways the circula-tion of these concepts supports the privatization of cultural processes: the consolidation of the work of culture into the hands of a few powerful companies and their ability to use a variety of legal instruments to hide the how and why of this work from the public eye. At stake, he argues, is the gradual abandonment of culture’s publicness and, ultimately, the emergence of a strange new breed of elite culture purporting to be its opposite. In an era of information surfeit, the information organizers are the new gatekeepers and their algorithms the contemporary analogues of ‘objectivity’ – as if the results they yield are not already shaped by the imperatives baked into the algorithms. In this regard, cultural concerns return us to questions regarding control over informational resources in the digital era.

Robert Gehl picks up the theme of instrumentalized ‘sharing’ in the digital age but historicizes and individualizes it in his essay ‘Sharing, Knowledge Management, and Big Data: A Partial Genealogy of the Data Scientist’. Following Foucault’s genealogical method, Gehl unpacks the discourse of sharing as it emerged specifically in the corporate practices of ‘knowledge management’ and the emergence of the knowledge worker in the 1990s. At this time, a firm’s strategic advantage was seen to lie in their employees’ ‘tacit’ knowledge, and the goal of management strategy was to compel it to be shared. This was the beginning of the highly flexible, precarious contemporary knowledge worker of today, whose primary asset is communicative skill and whose primary role is to turn knowledge into information, all under the aegis of ‘innovation’, ‘creativity’ and ‘play’. Tracing the connections between the rise of the knowledge economy in the 1960s, knowledge management strategies and the knowledge worker in the 1990s, and the Facebook ‘like’ button in the 2000s, Gehl then focuses on the worker whose job it is to make sense of all the information being produced: the data scientist. While touted as being the ‘sexiest job in the 21st century’, Gehl argues that the systems designed by data scientists will inevitably fold back on them, subjecting these workers to instrumental logics of their own design. Ultimately, data scientists will feel the problematic material effects of informationalization along with the rest of us: precarity, low pay and exploita-tive working conditions. As Gehl writes, ‘what big data firms crave from their data sci-entists is precisely what they enjoy with their data: cheapness and ubiquity’.

Moving from the data scientist to the actual techniques of data mining and predictive analytics, Adrian Mackenzie similarly looks to problematize the evolution of our current digital moment in his essay ‘The production of prediction: what does machine learning want?’ Mackenzie draws on the Foucauldian concept of ‘problematization’, insisting that in order to assess the ways in which material life is being transformed, we need to exam-ine the conceptual components and techniques of machine learning and prediction that lie behind data analytics. Mackenzie draws our attention to the fact that these techniques are generic and ‘indifferent’ to the specific data in question, insisting that we must rec-ognize conceptual tools, such as decision trees, perceptrons, logistic and linear regres-sion models, association rules and Naïve Bayes classifiers, as methods that have histories, cultural contexts and biases. Mackenzie traces the styles of reasoning and the narratives embedded in these different techniques of machine learning and data analysis, and argues

by guest on June 17, 2015ecs.sagepub.comDownloaded from

Andrejevic et al. 387

that, in these forms of reasoning, we can see foundational assumptions that are now working to condition the contours of our lives. Mackenzie argues that these abstract techniques are performative, comprising forms of material action in the world. Like Gehl, he contends that these predictive techniques are recursive and will inevitably fold back on themselves, working ‘to normalize the situations in which they themselves are entangled’. But, while Gehl insists that the data scientist will eventually be displaced by the very automation he helped to design, Mackenzie concludes by highlighting the ‘trans-individual co-operative potential’ of the techniques themselves – reminding us to be attentive to the specific political and economic contexts of these technological devel-opments and their liberatory possibilities.

The next group of papers develop and particularize the issues and contradictions that accompany the implementation of data analytics already identified – issues involving power, capitalism, labour, representation, materiality, surveillance and subjectivity. They do so by examining specific examples of the use and deployment of data mining and analysis across a range of social and cultural phenomena. These papers undertake the hard work of identifying and analysing new kinds of socio/cultural industry workers – the ‘infomediaries’, consumer data miners and reputation managers who have arisen to curate, shape and predict our tastes and desires in the 21st century.

In the first of these essays, ‘Curation by code: infomediaries and the data-mining of taste’, Jeremy Morris examines automated music recommendation systems, arguing that they occupy an increasingly central position in the circulation of media and cultural products. Far from being neutral purveyors of predictions, these recommendation sys-tems exert a logistical power that shapes the ways audiences discover, use and experi-ence informational and cultural content (Gillespie, 2014). The paper examines the specific case of The Echo Nest, a music intelligence service that provides a massive database, which services other recommendation systems, media measurement compa-nies, labels and other arms of the music industries. Largely a business-to-business affair, the company is a paradigmatic ‘infomediary’, providing not just content but also estab-lishing the databases, relational connections and audience intelligence that underpin the recommendations. Like ratings and measurement companies before them, Morris argues, an emerging constellation of infomediaries simultaneously measure and manufacture audiences. By looking at the shift from intermediation to infomediation, and through a critical interpretive reading of the Echo Nest database and some of its algorithms, the paper probes how cultural content itself is being readied, or ‘pre-configured’, for algo-rithmic analysis and examines what effects this has on its production. Picking up on the themes of inversion and recursivity highlighted by Gehl and Mackenzie and echoing Striphas’ claims, Morris underscores the ways in which the culture industries and our own cultural tastes are being re-shaped as a result of what Kate Crawford (2013) has called ‘data fundamentalism’ (n.p.).

While data analytics re-shape musical consumption and production, they are also busy transforming retail environments. In their paper, ‘Making Data Mining a Natural Part of Life: Physical Retailing, Customer Surveillance, and the 21st Century Social Imaginary’ Joseph Turow, Lee McGuigan and Elena Rosa Maris examine current efforts to reorganize retail environments around the data capturing and processing affordances of digital media. Using Charles Taylor’s concept of the social imaginary, Turow,

by guest on June 17, 2015ecs.sagepub.comDownloaded from

388 European Journal of Cultural Studies 18(4-5)

McGuigan and Maris contend that new institutional actors in the retail industry are using predictive analytics to re-shape lifestyles and identities and are instantiating new forms of social discrimination as a result. The authors focus specifically on loyalty programmes in retail stores, tracing their history from green stamps to branded charge cards and from frequent flyer programmes to the rise of Internet shopping, as they work to attract con-sumers and collect highly personalized data about them at the same time. Given that all manner of mobile apps and games are ‘venues for performing loyalty and accumulating rewards’, individual consumers now leave ever-growing trails of ‘data breadcrumbs’ for retailers. These analytics are deployed in turn on shop floors via mechanisms like dynamic pricing and digital price displays, or via loyalty rewards apps like Shopkick, which explicitly rewards consumers for ‘checking in’ with stores or scanning items using their smart phones. In this way, the authors argue, storefronts literally become ‘factories’ for the generation of consumer data, and function, simultaneously, to produce new sub-jectivities where ‘who you actually are is determined by where you spend time, and which things you buy’. Not all consumers and their data are created equal, however; as these new tracking technologies work to naturalize and reward the practices of surveil-lance, they develop new forms of social discrimination predicated entirely on consumer behaviour. In the wake of these developments, data-driven discrimination increases and we see a lessening of democratic relations in all areas of life.

Kate Crawford, Tero Karppi and Jessa Lingel take up similar concerns about the extraction of value from all aspects of human life, specifically sleep, in their essay ‘Our Metrics, Ourselves: One Hundred Years of Self-Tracking from the Weight Scale to the Wrist Wearable Device’. Attending to the ‘commodification and knowledge-making that occur between bodies and data’, the authors focus on the emergence of contemporary self-tracking apps and devices like the Fitbit Flex, Jawbone UP or the Nike FuelBand SE and analyse them by way of a comparison with a historical precedent – the weight scale. Initially positioned as a public machine, which would reward users with candy along with their weight when they emerged in the later 1900s, the weight scale gradually became a private home-based device, tied to normalizing body types but sold under the guise of increasing self-knowledge and social power. Indeed, the message of self-knowl-edge through measurement can be found reiterated in much contemporary advertising for new self-tracking apps and devices and in the rise of the current ‘Quantified Self’ move-ment. These devices reinforce the view that individuals are not the most authoritative source of data about themselves (Bossewitch and Sinnreich, 2013) but, rather, need the help of a quantifying machine to know themselves better. Taking issue with Jonathan Crary’s (2013) claim that, in an era of 24/7 activity, sleep is one of the only areas of life yet to be colonized by the market, Crawford et al. point out that extracting monetary value from sleep is exactly what devices like Fitbit and Jawbone UP now seek to do. While users extract value from the measures provided, companies collect data from indi-vidual users that can then be used or sold in a myriad of ways. In addition to the moneti-zation of user data, the authors argue that there is an implicit form of normalization of bodies and standards at work in the operations of these devices, which remain opaque to users. These new apps and devices help to create more efficient, normalized gendered bodies; function to naturalize forms of implicit participation in regimes of data surveil-lance, monitoring and extraction; generate data sets for all manner and means of

by guest on June 17, 2015ecs.sagepub.comDownloaded from

Andrejevic et al. 389

capitalist industry and, most troubling, extract value from our desire for self-understand-ing, all while we sleep.

Music production and consumption, the retail shopping experience and even our sleep patterns are increasingly conditioned and determined by new forms of data extraction and analysis, and all of these tools, devices and systems incorporate some aspect of geolocation to enable their functioning. In ‘Platform specificity and the politics of location data extrac-tion’, Rowan Wilken and Carlos Barreneche examine the growing commercial significance of location data and argue for a ‘medium-specific analysis’ of the differences between dominant techniques of geolocation data extraction and their political-economic arrange-ments. To this end, the authors offer a comparison between the ways in which Foursquare and Google each extract and use geocoded user data. Foursquare has moved from a gami-fied form of user check-in into a geospecific kind of user retail recommendation and rank-ing platform. Indeed, Foursquare would fit neatly with the kinds of retail shopping apps and devices reviewed by Turow et al., as it puts its geolocation data to work to produce a retail consumer database. The ultimate effect, the authors argue, is the generation of ‘geode-mographic profiling’, which can then be used for predictive purposes. Google, on the other hand, deploys its own users as sensors to drive mobility flows. From Global Positioning System (GPS) signals in mobile phones, to WiFi access points, to check-ins on sites like Facebook and Twitter, Google collects user data, claiming to anonymize it. The truth is, however, that many Android cell phones transmit geolocation data whether the phone is running or not, and Google only needs a small number of anonymous spatio-temporal mobility traces to uniquely identify most of us. As with so many other forms of data analyt-ics, Google’s ultimate goal is to be able to predict user behaviour in order to ‘convert data traffic into foot traffic for local retailers’. Insofar as they do not simply collect information about humans and their environment, but reflexively interact with, and adjust to them, modulating flows of populations of users, the authors argue that we must come to see these kinds of geolocation as environmental technologies. Recalling Crawford et al.’s claims about the production of the ideal worker subject via self-measuring devices and apps, Wilken and Barreneche argue that geolocation techniques facilitate ‘technocratic forms of … urban mobilities, aligning “the rhythms of the city” with the “rhythms of the commod-ity”’. The authors caution us to be mindful that these forms of consumer tracking are also underpinned by efforts to securitize mobility and, thereby, contribute to the construction of ‘distinct “place ontologies” – that is, “ways of categorizing the world”’ that threaten to re-form our very notion of the ‘public’.

Mark Davies expands on the themes of surveillance, consumer tracking and ‘neolib-eral digital capitalism’ in his essay ‘Ebooks in the global information economy’. Noting that little scholarly attention has generally been paid to the ebook, Davies points out that Amazon, the world’s largest purveyor of ebooks, is also one of a group of large digital media corporations currently vying for dominance in the global information economy. As such, Davies argues, we must understand the ebook as a ‘node in the wider discourse of neoliberal digital informationism’. In other words, ebooks are exemplary objects in the current alignment of neoliberal economic doctrine and celebratory forms of informa-tion capitalism. As Davies shows, reviewing the work of Babbage, Taylor, Mumford, Turing, Peter Drucker and Daniel Bell, discourses of information and market freedom have gone hand in hand for many decades and have culminated in the rise of

by guest on June 17, 2015ecs.sagepub.comDownloaded from

390 European Journal of Cultural Studies 18(4-5)

contemporary neoliberal digital capitalism and ‘Silicon Valley ideology’. Davies then turns to the ebook, arguing that it works to commodify reading and readers, coopts users’ labour and serves as a strategic site for the ‘corporate struggle over the digital commons’. For example, Amazon’s proprietary ‘Whispersync’ technology works to collect informa-tion about Kindle eBook users, such as what they read, how fast they read and what pas-sages they underline, and then adds this to its consumer database. Indeed, Amazon’s business strategy depends on selling books at cost while generating valuable user data, and, in turn, Amazon exerts its distributing power by delisting publishers who refuse to participate in these practices. Ebook publishers also use reader data to design their next publishing projects, down to preferred word length or type of protagonist. In this way, the ebook as data-generating device serves to re-shape what kinds of books and information actually get published, producing a new ‘information ecology’ driven by market infor-mation and surveillance, and signalling ‘the retreat of public duty notions of book pub-lishing’. Focusing specifically on Amazon’s Kindle, Davies positions the ebook as a strategic commodity, a crucial element in corporate competition ‘that has little to do with book publishing, and everything to do with beating competitors Apple and Google in the race to dominate the global information economy’. Under the aegis of ‘creative disrup-tion’, global network powers, like Amazon, naturalize the practices of digital enclosure and digital capitalism, and place citizens in a democratic double-bind, where the poten-tial for enhanced participation provided by digital media must be traded off against new forms of corporate surveillance. Davis concludes by making a plea to step up efforts to popularize a ‘progressive critique of digital network media’ which, paradoxically, depends on ‘currently out-of-fashion notions of a critical public culture’.

In his essay ‘Open source intelligence, social media and law enforcement: visions, con-straints and critiques’, Dan Trottier continues a focus on questions of surveillance and big data analytics by examining the adoption of open-source intelligence (OSINT) and social media monitoring in European law enforcement and intelligence agencies. He argues that OSINT practices are techno-economic constructs, working to ‘augment the visibility and exploitation of social life’. Positioning the current use of OSINT in relation to CCTV monitoring of the 1980s and 1990s, Trottier goes on to examine the implementation of open-source social media intelligence tools, such as NiceTrack, in policing contexts in Europe. While police agencies tend to describe social media information as a pure, constantly flowing source of information and social media users as, alternatively, customers or criminals, the actual implementation of OSINT in policing contexts brings to the foreground a far more complex set of issues. Drawing from interviews with 19 police and intelligence officials and 14 privacy advocates from 13 European Union (EU) member states, Trottier highlights respondents’ motivations to adopt social media monitoring technologies in their jurisdictions; these include the desire to identify terror threats, child exploitation and copyright infringement. Trottier goes on to highlight the ways in which the use of OSINT is severely limited and conditioned by already-existing legal frameworks, budgetary constraints, an unresponsive culture of policing and the circulation of deliberate misinformation online, and enumerates the liabilities and social costs attached to the use of OSINT. These include an increase in forms of categorical discrimination, the gradual expansion of monitoring and surveillance to more and more areas of life or ‘function creep’, the criminalization of online spaces and the demarcation of every individual online as a pos-sible suspect and an increased risk that the public will see OSINT as yet another opaque form

by guest on June 17, 2015ecs.sagepub.comDownloaded from

Andrejevic et al. 391

of police surveillance. From his analysis, Trottier notes that the use of OSINT collapses social and institutional contexts, remediating personal content offered on privately owned platforms into public evidence that can be used in a variety of troubling and unpredictable ways. These new uses of social media data, in turn, can come to have a chilling effect on users’ online speech, rendering us all more passive in the long run.

Following in the cultural studies tradition of identifying spaces for agency in the face of structures of power, the final three papers of the volume pose distinct challenges to accounts of big data that see it as ineluctably oriented towards power asymmetries and domination. These papers locate possibilities for resistance, engagement and enrichment via the use of new means of computation and big data analytics. In ‘Plenty as a response to austerity: big data expertise, cultures and communities’, Caroline Bassett focuses on the links between data analytics and governance. Echoing Mark Davis, she situates the explosion of big data within the trends of advanced neoliberal digital capitalism, specifically its recent obsession with austerity. But, drawing on research into the use of big data analytics in small communities in the United Kingdom, she argues that these analytical tools are not ‘intrinsically “virtuous” or terminally evil’. Indeed, Bassett contends that there can be ‘good’ uses and forms of big data analytics. In an attempt to address the ‘lag’ that occurs between the introduction of new technological tools and the development of expertise on the ground to make use of these tools, Bassett begins from first principles, working to redefine both the terms ‘big data’ and ‘expertise’. Bassett picks up themes explored by Adrian Mckenzie as she draws on Bernard Steigler’s claims about the ‘co-construction’ of humans and technological forms, and the pharmacological nature of technolo-gies insofar as they can have both curative and toxic dimensions. Bassett redefines big data as ‘a process, involving humans and machines, hermeneutic and algorithmic operations … engag-ing specific … locations and conditions, and producing forms of knowledge that … tend … to present themselves as entirely computational’. She goes on to reassess definitions of expertise, arguing that, as big data challenges the very ways in which we define and think about expertise, it simultaneously is shaped by, and has embedded within it, various kinds of already-existing distributed human expertise. If we see big data as constitutive of and constituted by forms of human expertise, Bassett argues, then we can see it as open, malleable and amenable to alterna-tive uses. Given this corrective to our understanding of big data analytics as inherently socio-technical forms of operation, the question becomes, ‘Who has access to these tools and how are they being used?’ If, as Bassett hopes, we can realign our understanding of Big Data analytics away from celebratory views of it as a kind of post-interpretive computational knowing, and towards a view of it as co-constitutive of humans and their social conditions, then we might ‘find ways to use Big Data circuits to enable particular forms of life to flourish’.

Dan McQuillan takes Edward Snowden’s revelations as a starting point to interrogate the continued salience of our established modes of critique in contemporary algorithmic culture. In ‘Algorithmic States of Exception’, McQuillan takes up Bassett’s claim that our critical focus should not be on big data per se, but on the ‘nature of the material-political apparatus that connects data to decision-making and governance’. Tracing changes in the conceptual apparatus that buttress and support certain kinds of data analytics, McQuillan specifically examines the shift from the use of relational database management systems to the use of NoSQL – a form of data storage that does not look for relations between data but stores all manner and means of data in a ‘schema-less’ system. McQuillan argues that this kind of data storage facilitates forms of data mining that stress prediction and

by guest on June 17, 2015ecs.sagepub.comDownloaded from

392 European Journal of Cultural Studies 18(4-5)

correlation over relationships and causation. These predictions are then deployed as forms of algorithmic regulation, prescribing corrective measures and new modes of subjectivity. McQuillan, drawing on Agamben, asserts that these trends towards prediction in all fields of life are ‘colliding with our assumptions about political and judicial fairness’ and are producing a perpetual ‘state of exception’; they increasingly operate with coercive force on populations, but without any substantive controls or legal consequences; these ‘new operations have the potential to create social consequences that are unaddressed in law’. McQuillan asks how we might create counter-movements to these new forms of exclu-sionary power and suggests that a potentially fruitful way forward involves making cor-relations between historical and contemporary forms of resistance and assertions of the commons. For example, McQuillan traces the ways 13th century Christian antinomianism resonates in the work of hacker group Anonymous; both, he argues, sought to disrupt apparatuses of control ‘without themselves engaging in the cycle of lawmaking and law-preserving’. He also notes that 18th century food riots in Britain share much in common with contemporary DIY cryptoparties and festivals; both work to intervene in situations of excess and unfairness, using whatever tools they have at hand, in a collectivist mode and without legal cover. These historical examples, McQuillan argues, can give us clues as to how to re-shape the dominant privatized big data apparatus; the way forward does not include discarding our technological affordances, but in ‘learning against’ their current forms and logics, and reconfiguring them to enable and advance social struggles.

Finally, Fenwick McKelvey, Matthew Tiessen and Luke Simcoe argue, like McQuillan, that concerns about publicness and privacy no longer obtain in contemporary algorithmic culture. Rather, they insist that our digital lives are now working to produce a vast simulated reality. Inspired by the 1964 science fiction classic Simulacron-3, which envisages an entire city run by computers and mined by scientists as a giant exercise in market research, ‘A Consensual Hallucination No More? The Internet as Simulation Machine’ explores the oper-ations and implications of today’s Internet, run as it is on the fuel of the individual impulse to share and communicate. In a kind of matrix-like inversion, the authors claim that humans’ attempts to share, connect and making meaning have been reduced to ‘standing reserves’ of information, owned and used overwhelmingly by those who wield global power and control: banks, corporations, governments. The Internet as simulation machine intensifies existing power asymmetries and social and financial exclusion in addition to producing new ones; it creates new digital divides between those who do not choose to take part in life online and those who do, those who have the means to curate their online identity and those who do not, and those who do and do not have access to the ‘firehose’ of data flowing in through sites such as Twitter and Facebook. McKelvey et al. explore possibilities for resistance under cir-cumstances where traditional ‘logical’ forms of antagonistic communication are also inevita-bly reduced to feeding bots making stock market decisions. Focusing on the work of 4Chan and the Deterritorial Support Group, they ask whether it might not make more sense to make no sense at all in this day and age. Is non-communication or ‘idiocy’ a new form of resistance, or will communicative tricksters be used to train the next generation of simulation machines?

Our goal with this volume is to ask cultural studies scholars to take seriously the prolif-erating processes of data mining and analytics. As we have argued, we recognize that these developments can prove challenging as a focus of study. This is so not just because what they ‘are’ and how they operate are so opaque and relatively new, or because they are so

by guest on June 17, 2015ecs.sagepub.comDownloaded from

Andrejevic et al. 393

different from other kinds of cultural objects, or even because they are predicated on deny-ing that they are cultural objects in the first place, but because of the real material barriers that exist for us to access their operations, logics and forms of implementation. These barri-ers alone are reason enough to bring questions of power, ownership, signification and sub-jectivity back into the forefront of their critique. To be sure, the essays we have assembled only begin to scratch the surface of these developments in big data and data analytics. Much more remains to be done. We need to develop new methodologies and new intellectual and critical competencies to tackle the embedded assumptions buried in the code and their politi-cal and cultural implications. Our ability to accomplish these things will require more than isolated scholarly effort; collaborative, politically engaged activist sensibilities will no doubt be required in order to push past the privatized digital enclosures and open up access to the algorithms, analytics, distributive regimes and infrastructural monopolies that are increas-ingly coming to condition the contours and substance of our daily lives.

ReferencesAnderson C (2008) The end of theory: The data deluge makes the scientific method obsolete.

Wired Magazine, 23 June. Available at: http://www.wired.com/science/discoveries/maga-zine/16-07/pb_theory (accessed 30 August 2013).

Barad K (2003) Posthumanist performativity: Toward an understanding of how matter comes to matter. Signs 28(3): 801–831.

Bennett J (2009) Vibrant Matter: A Political Ecology of Things. Durham, NC: Duke University Press.

Bossewitch J and Sinnreich A (2013) The end of forgetting: Strategic agency beyond the panopti-con. New Media & Society 15(2): 224–242.

boyd d and Crawford K (2012) Critical questions for big data: Provocations for a cultural, techno-logical, and scholarly phenomenon. Information, Communication & Society 15(5): 662–679.

Crary J (2013) 24/7: Late Capitalism and The Ends of Sleep. New York: Verso.Crawford K (2013) The hidden biases in big data. Harvard Business Review, 1 April. Available at:

http://blogs.hbr.org/2013/04/the-hidden-biases-in-big-data/Galloway AR (2012) A response to Graham Harman’s ‘Marginalia on Radical Thinking’. An und

für sich, 3 June. Available at: https://itself.wordpress.com/2012/06/03/a-response-to-graham-harmans-marginalia-on-radical-thinking/

Galloway AR (2013) The poverty of philosophy: Realism and post-fordism. Critical Inquiry 39(2): 347–366.

Gillespie T (2014) The Relevance of algorithms. In: Gillespie T, Boczkowski P and Foot K (eds) Media Technologies: Essays on Communication, Materiality, and Society. Cambridge, MA: MIT Press, pp. 167–193.

Gray A, Campbell J, Erickson M, et al. (2007) CCCS Selected Working Papers, vol. 1. New York: Routledge.

Griffin (2014) Incredible machines: trying to negotiate in a zettabyte world. The Vancouver Sun, 3 March. Available at: http://blogs.vancouversun.com/2014/03/03/incredible-machines-trying-to-negotiate-in-a-zettabyte-world/ (accessed 25 March 2015).

Grosz E (2004) The Nick of Time: Politics, Evolution, and the Untimely. Durham, NC: Duke University Press.

Hall S (1992) Cultural studies and its theoretical legacies. In: Grossberg L, Nelson C and Treichler P (eds) Cultural Studies. New York and London: Routledge.

by guest on June 17, 2015ecs.sagepub.comDownloaded from

394 European Journal of Cultural Studies 18(4-5)

Kittler F (2010) Optical Media. Cambridge: Polity Press.Lohr S (2012) The age of big data. The New York Times, 11 February. Available at: http://www.

nytimes.com/2012/02/12/sunday-review/big-datas-impact-in-the-world.html?_r=0 (accessed 10 August 2012).

Packer J (2013) Epistemology not ideology OR why we need new Germans. Communication and Critical/Cultural Studies 10(2–3): 295–300.

Peters JD (2013) Calendar, clock, tower. In: Stolow J (ed.) Deus in Machina: Religion and Technology in Historical Perspective. New York: Fordham University Press, pp.25–42.

Turner G (1992) British Cultural Studies. London and New York: Routledge.Wernick A (1990) Promotional Culture: Advertising, Ideology and Symbolic Expression. London,

Thousand Oaks, CA, New Dehli: Sage.Williamson B (2014) The death of the theorist and the emergence of data and algorithms in digital

social research. Impact of Social Sciences, LSE Blog, 10 February. Available at: http://blogs.lse.ac.uk/impactofsocialsciences/2014/02/10/the-death-of-the-theorist-in-digital-social-research/

Williams R (1958) Culture and Society, 1780-1950. New York: Columbia University Press.Williams R (1981) The Sociology of Culture. Chicago: University of Chicago Press.

Biographical notes

Mark Andrejevic is Associate Professor of Media Studies at Pomona College in the US. His latest book, Infoglut: How Too Much Information Is Changing the Way We Think and Know (2013), explores the social, cultural, and theoretical implications of data mining and predictive analytics. His work has appeared in a edited collections and in academic journals including Television and New Media; New Media and Society; Critical Studies in Media Communication; Theory, Culture & Society; Surveillance & Society; The International Journal of Communication; Cultural Studies; The Communication Review, and the Canadian Journal of Communication. His current work explores the logic of automated surveillance, sensing, and response associated with drones.

Alison Hearn is an Associate Professor at the University of Western Ontario in Canada. Her research focuses on the intersections of promotional culture, new media, self-presentation, and new forms of labour and economic value. She also writes on the university as a cultural and politi-cal site. She has published widely in such journals as Continuum, Journal of Consumer Culture, Journal of Communication Inquiry, and Topia: Canadian Journal of Cultural Studies, and in edited volumes including The Media and Social Theory, Blowing Up the Brand, and The Routledge Companion to Advertising and Promotional Culture. She is co-author, with Liora Salter, of Outside the Lines: Issues in Interdisciplinary Research (McGill-Queens University Press, 1997).

Helen Kennedy is Professor of Digital Society at the University of Sheffield. She has been researching new and digital media since they came into existence and has published widely in this field. She is author of Net Work: ethics and values in web design (Palgrave Macmillan, 2011). Her work has been published in various journals including New Media and Society; Information, Communication and Society; Media, Culture and Society; convergence; Ephemera; The Information Society. She is currently researching ordinary forms of social media data mining, and how ordinary people interact with data visualisations.

by guest on June 17, 2015ecs.sagepub.comDownloaded from