Abstract
Recent work on Big Data and analytics reveals a tension between analyzing the role of emerging objects and processes in existing systems and using those same objects and processes to create new and purposeful forms of action. While the field of science and technology studies has had considerable success in pursuing the former goal, as Halford and Savage argue, there is an ongoing need to discover or invent ways to “do Big Data analytics differently.” In this commentary, I suggest that attempts to produce new ways of working with Big Data and analytics might be hindered by how science and technology studies-influenced scholars have conceptualized assemblages. While these scholars have foregrounded objects’ relations within existing assemblages, new materialist philosophers draw attention to properties of objects that transcend those relations and might indicate opportunities for more creative or generative uses of Big Data and analytics.
In his introduction to Deleuze and Guattari's A Thousand Plateaus, Massumi (2003: xv) declares, “A concept is a brick. It can be used to build the courthouse of reason. Or it can be thrown through the window.”
Massumi’s statement highlights a tension within science and technology studies (STS)-influenced work on Big Data and analytics—between analyzing the role of emerging objects and processes in existing systems and using those same objects and processes in order to create new and purposeful forms of action. Big Data and analytics are used to build various courthouses of reason, from the financial (Christiaens, 2016), to the scientific (Lowrie, 2017), to the legal (Mantello, 2016)—but, as scholars such as Halford and Savage (2017) point out, they might also be used to disrupt, tear down or provide exciting alternatives to those courthouses. To this point, however, efforts to analyze existing uses of Big Data and analytics have gained more traction than efforts to invent new, creative or lively ways of working with these objects.
It is revealing that Massumi’s statement on the dual nature of objects—that they not only form parts of existing systems but also have potentials that extend beyond those systems—comes in his commentary on Deleuze and Guattari’s work that initiated interest in assemblages. The concept of assemblages plays a central role in STS-influenced work on Big Data and analytics; however, STS’s uptake of this concept has largely come by way of actor-network theory (ANT) and, to a lesser extent, the work of Michel Foucault, engaging considerably less with Deleuze and Guattari’s earlier work or with the subsequent writings of new materialist philosophers such as Jane Bennett, Ian Bogost, and Manuel DeLanda. Consequently, STS scholars have tended to foreground some aspects of assemblages (focusing, for example, on heterogeneity and the agency of objects) while downplaying others (such as properties of objects that exceed their relations). While there is considerable overlap between STS-influenced scholars’ conceptions of assemblages and those of new materialists, the latter’s attempts to understand objects apart from their relations indicate methodological approaches that might lead to the kind of generative work on Big Data and analytics that Halford and Savage, among others, call for.
Characteristic of STS-influenced approaches, Kitchin’s often-cited The Data Revolution (2014a: 24–25) describes data assemblages as entangled and heterogenous, “composed of many apparatuses and elements that are thoroughly entwined, and develop and mutate over time and space [and] frame what is possible, desirable and expected of data.” Both data and analytic processes, Kitchin argues, are caught up in webs of relations so complex that it is impossible (and perhaps methodologically undesirable) to isolate an object and study it apart from its surroundings. Writing specifically of algorithms, for example, he notes that they combine original code with libraries and algorithms, forming relations with “potentially thousands of individuals, datasets, objects, apparatus, elements, protocols, standards, laws, etc. that frame their development” (Kitchin, 2014b: 15).
Focusing in this way on the complex relations that surround data and analytic processes has methodological consequences, including the encouragement of ethnographic studies that place the researcher in a position to observe how elements are made to cohere in existing or developing systems. Projects such as Ribes and Polk’s (2015) study of a long-running HIV study or Geiger’s (2014) study of Wikipedia assume that data cannot be understood in isolation. They also assume that the assemblages that matter most are those that are already formed—that societal consequences are not only a matter of data or analytic processes but, more importantly, the ways these are brought together with other elements to establish the bounds of what can be known or said. In line with these assumptions, tracing how objects—including data and analytic processes—are made to cohere in assemblages and critiquing the societal consequences that emerge has become a dominant methodology.
However, while the majority of STS-influenced work on Big Data and analytics falls within this broad framework, a noticeable strand of work expresses a desire to not only trace and critique but to also imagine or develop alternative data assemblages that would better align with scholars’ values and goals. These calls respond to two, related threats: first, that the growing prominence of data and analytics challenges sociologists’ claim to speak or make claims about the world and, second, that existing data assemblages have harmful effects that scholars could directly address.
Speaking to the first threat—that data and analytics challenge sociologists’ relevance—Burrows and Savage (2014; see also Savage and Burrows, 2007, 2009) express concern that data scientists and others working outside of academic institutions are able to produce knowledge from emerging forms of transactional, digital data in ways that diminish the authority of sociologists. Halford and Savage (2017: 2) similarly worry that sociologists are “marginalis[ed] […] from a new and increasingly powerful data assemblage that is coming to play an important part in the production of information and knowledge in the 21st century.” While researchers at social media companies, for example, analyze activity traces at very high volumes using sophisticated machine learning algorithms, academic researchers struggle not only to keep up but also to articulate new uses of technologies that would better align with their goals and values.
In addressing the second threat—that existing data assemblages have harmful societal consequences—scholars have focused critical attention on specific domains. Currie et al. (2016: 2), for example, consider the politics of police officer-involved homicide data and argue that official statistics are produced in incomplete and misleading ways. Their project begins with the tracing of current systems—a process that the authors note is informed by recent work in critical data studies—but they also ask “how we can move beyond dissecting and analyzing POIH data towards understanding it as a lever of political action” (2). Similarly, Kennedy and Moss (2015: 1) begin their project by pointing to existing concerns that the ability to access and produce knowledge from digital data is unevenly distributed in ways that disadvantage the public but also note increasing “calls to do data mining differently and democratize data power.” While they concede that the alternative uses of data and analytics they propose are “far from being realised in practice,” they strongly argue for the value of “imagining what could be” (2).
These calls imply a considerable shift in scholars’ goals. Rather than tracing the development of systems, critiquing their consequences or even participating in their production and rollout, they suggest that STS scholars might more actively steer the development of systems in novel directions. It is not surprising that, from a methodological perspective, this transition is still being worked out, as studying the dominant assemblages in which Big Data and analytics are entangled does not necessarily entail that scholars are in a good position to imagine new assemblages. Critics of ANT, largely coming from the new materialist tradition, have articulated similar concerns, arguing that the theory’s focus on relationships leads to methodologies that produce endless descriptions of what is but that lack the ability to criticize or to generate new ideas. As an alternative, new materialist critics point to a conception of assemblages that draws more heavily on the work Deleuze and Guattari and, they argue, leads to more generative methodologies. Bear (2013), for example, argues, “While an actor-network approach might favour a forensic examination of a particular event or process, Deleuze and Guattari are more ‘anticipatory’ and concerned with continuing trajectories and future possibilities or becomings” (24). Similarly, Müller (2015) describes ANT as an “empirical sister-in-arms” of Deleuze and Guattari’s assemblage thinking—while both approaches see heterogeneous objects as coming together to form assemblages, ANT has the goal of looking out at existing assemblages while Deleuze and Guattari emphasize what those assemblages might become or how the objects that comprise them might participate in other assemblages.
These criticisms do not apply to all STS-influenced work on Big Data and analytics. Most glaringly, studies in this area are consistently critical, and projects such as The Programmable City, led by Rob Kitchin, do attempt to guide the development of new systems. Critiques of ANT do, however, establish useful guideposts for thinking about how conceptions of assemblages align with various goals. The most useful of these guideposts is the distinction between seeing objects as defined by their relations within an assemblage and seeing them as exceeding those relations. This is largely a matter of emphasis, as both ANT and new materialism recognize the agency of objects and the heterogeneity of assemblages. However, where scholars influenced by ANT emphasize relations and how they are made to cohere, those influenced by new materialism emphasize the objects themselves. Bogost (2012: 7) provides a clear example of this difference in emphasis: In the networks of actor-network theory, things remain in motion far more than they do at rest. As a result, entities are de-emphasized in favor of their couplings and decouplings. Alliances take center stage, and things move to the wings. As Latour says, “Actors do not stand still long enough to take a group photo.” But yet they do, even as they also assemble and disband from their networks. The scoria cone and the green chile remain, even as they partake of systems of plate tectonics, enchiladas, tourism, or digestion.
DeLanda (2006) presents a cohesive review of Deleuze and Guattari’s assemblage theory, and his articulation of “relations of exteriority” seems especially relevant for those attempting to imagine new uses for data and analytic processes. Relations of exteriority, for DeLanda, are a key feature of assemblages and mark a point of distinction between theoretical approaches that focus on “organic totalities” that must cohere and those that focus on assemblages of objects that exceed their relations. As DeLanda argues: [Relations of exteriority] imply, first of all, that a component part of an assemblage may be detached from it and plugged into a different assemblage in which its interactions are different. In other words, the exteriority of relations implies a certain autonomy for the terms they relate, or as Deleuze puts it, it implies that ‘a relation may change without the terms changing.’ (10–11)
Bates et al. (2016) situate their data journeys approach as diverging from conventional ANT and STS methodologies. Rather than “produc[ing] detailed accounts of specific knowledge or data infrastructures and the intra-network politics of their development,” data journeys focus attention on the movement of data between sites (2). As Bates et al. argue, focusing on the movement of data emphasizes moments of transition or times at which the data are used to achieve different goals by different actors. In their study of the journeys of meteorological data, for example, they highlight the mutability of data and ways that this quality contributes to potential reuse. In part the data journeys approach is valuable for drawing attention to how actors at different sites work to adapt data to align with their distinct needs, but it is also valuable for prompting questions about data themselves and their material properties. Not all data are mutable in the same way, and focusing on the journeys they take might be one way to begin to ask about how specific storage formats, for example, make data mutable in ways that might be valuable for generative work.
Devendorf et al. (2016) draw on many of the new materialist philosophers cited above but work in the area of human–computer interaction and design, where they advocate a morphogenetic approach, or one that would allow the intrinsic properties of materials to guide their use in technical processes. In addition to seeing objects as having agency, this approach specifically sees that agency as rooted in properties of objects or materials before they enter into assemblages. Redeform, a prototype developed by Devendorf and Ryokai, translates G-Code instructions—conventionally executed by a 3D printer—into a visual form that humans can follow using a variety of materials, from charcoal to Cheez Whiz. In using Redeform to create various art projects, users note a growing awareness of materials. A property of charcoal, for example, is that it smears, and Devendorf describes coming to understand how this property might be used to produce different effects. Similar, materials-led approaches in HCI attend to the material properties of yarn (Klefeker and Devendorf, 2018) and silver (Tsaknaki et al., 2017) in order to generate new prototypes and to reflect on design processes, and Gross’s (2013) “datamoshing” approach illustrates how related, material-focused processes can apply to digital objects and lead to creative new uses. In these contexts, new materialists’ incitement to focus on or stay with objects leads to methodologies that allow materials to lead and encourage researchers to attend to unexpected consequences as opportunities to learn and innovate.
These projects suggest new ways of thinking about and working with data and analysis processes—ways that do not necessarily conflict with existing STS-influenced approaches but that do, perhaps, provide useful indications of how scholars might work with these objects in more generative ways. In line with the spirit of Deleuze and Guattari’s work, the suggestion that scholars interested in Big Data and analytics experiment with new materialist conceptualizations of assemblages is pragmatic. It is less concerned with questioning what Big Data and analytics are and more with seeking new ways to use them to affect the world. Massumi, continuing the discussion of bricks and courthouses with which I began this commentary, describes this pragmatic goal in ways that align well with scholars’ recent desires to use Big Data and analytics in new, creative and inventive ways: “The question is not: is it true? But: does it work? What new thoughts does it make it possible to think? What new emotions does it make it possible to feel? What new sensations and perceptions does it open in the body?” (xv)
Footnotes
Acknowledgments
This commentary benefited considerably from the generous feedback of the reviewers and editors at Big Data and Society.
Declaration of conflicting interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.