Читать книгу Being with Data - Nathaniel Tkacz - Страница 11

How will this series of investigations proceed? Often, I begin with a dashboard. For social science readers, this can be considered a grounding move. Dashboards will ground my words as they weave and loop through different times and places, acting as a centrifugal force. Sometimes the focus will be on dashboards themselves; at other times I will swirl around them, moving from format to the events that shape its arrangements, or to the contexts and situations that undergo formatting.

As this is a format that travels across times and spaces, through diverse settings and different medial configurations, there are many possible ways to initiate a dashboard-driven inquiry. For example, a dashboard may be realized through software; through numbers, colours and visualizations; through interfaces; and as part of larger systems. Dashboards are also used for specific things and exist in specific contexts (web analytics, fitness, city councils and so on). Each of the different components or features of a dashboard contains at least one related field of inquiry: software studies, the anthropology of numbers, studies of visual perception (cognitive science) and data visualization, human–computer interaction and user experience design, infrastructure studies, organization studies and so on. And each of these could easily be substituted with others, depending on the nature of the inquiry. For example, one could replace infrastructure studies with information systems, steering the inquiry more towards the practical requirements of business. My approach will be to traverse these terrains as a (mostly friendly) trespasser, pulling threads together as needed. It should not be mistaken for an arbitrary approach, however – not an interdisciplinary willy-nilly. I will attempt to stick close to and ‘follow the dashboard’, not in the same way the early multi-sited ethnographers employed a ‘follow the thing’⁵² methodology to track commodities (and other things) around the world in order to weave a tale about globalization or the logistics of capitalism. Instead, I follow the dashboard by picking up selected moments of its past or present that highlight significant aspects of it as a format. In these moments, I come into contact with different forms of expertise and practice, with some corresponding to disciplinary knowing and the coordinates of others more difficult to establish.

Along the way, I engage with the earliest motor cars; the horse and carriage; French industrial accounting and management; the field of decision support systems; business intelligence and analytics; the management of hospitals; commercial dashboard providers; situation rooms; many different types of data; and, of course, a Saturday morning jog. I draw upon equally diverse bodies of research, knowing well that I cannot hope to master all this terrain, or to contribute novel insights to experts in each respective area. I hope bringing them together allows me to say something worthwhile about how data formats are producing new ways of being and how we might study them.

The main part of the book is divided into three chapters. The first chapter, ‘Archaeology of Dashboards’, offers what I call a ‘format archaeology’. This sets the study apart from other historically minded research on the topic of data, which have tended to focus on numbers, statistics, facts and related methods, or on broader epistemological themes, such as Ian Hacking’s fascinating studies of chance and probability.⁵³ Indeed, for much of the first part of this study data are not in the picture at all or are only minor characters. What I aim to understand is the development of the dashboard as a format. Where do dashboards come from? How have they changed over time? What persists? What ways of relating to the world do they encourage? What cultural mythologies are they caught up in? And what forms of subjectivity do they foster?

There are many possible moments one could pull from the archive. I’ve tried to focus on ones that I think offer the most food for thought in the present because they say the most about how dashboards (and dashboard data) came to be, ways that I think could be overlooked if one were to focus only on the most famous historical moments. For example, whereas I could have focused on the well-known US military project SAGE (the Semi-Automatic Ground Environment) and positioned dashboards more strongly within a military context, I have instead focused on the lesser-known work of people such as Michael Scott Morton. His early research with Westinghouse was not on the scale of SAGE, but he was the first to propose something like a ‘dashboard relation’, an interactive data display for making decisions within a management context. His work would help establish the field of decision support systems (DSS) and today’s business dashboards can be traced more or less directly back to his early experiments. Thus, Scott Morton receives the lion’s share of my attention, whereas other well-known projects like SAGE, LEO or even Cybersyn are mostly left aside.⁵⁴

Beginning with the horse and carriage, the format archaeology moves to consider the motor car, pre-digital French managerial dashboards, the rise of DSS and related executive support systems and executive information systems, before ending with a discussion of business intelligence and the proliferation of dashboard analytics. Through this material, I am interested not only in how dashboards have changed over time and in different contexts, but also in what holds them together. Without elaborating too much, I argue that dashboards always facilitate an originary separation, a separation which the dashboard also bridges through its specific mediations. Dashboards also imply and rely upon a sense of motion. They exist in situations understood to be in motion and they in turn produce a sense of motion as part of their formatting. Through this separation and motion, dashboards encourage distinct forms of perception or ways of seeing data, which I refer to as driverly perception. Finally, dashboards configure the cognitive activity of their users as one relating to decision-making. The user of dashboards is a decision-maker and the context or setting they are in is recast around the dynamics of decision-making. I refer to this formatting work through the notion of a decision ontology.

The remainder of the book has a contemporary focus, moving from what ‘persists’ to ‘lived experience’, drawing on two main case studies. I arrived at these by taking advantage of opportunities as they came over a period of roughly six years. Over this time, I interviewed or informally discussed dashboards with many users and developers and across a range of sites. I have interviewed the creators of an ‘activism’ dashboard; worked with researchers in a London-based think tank while they ‘dashboarded’ a televised election debate; discussed dashboards with a team of ‘smart city’ public servants in Melbourne, Australia, as they pondered what kind of dashboard was best for their city’s needs. I studied civil servants in the UK’s Government Digital Service team as they rolled out almost 900 performance dashboards as part of a larger digital transformation strategy. On an individual level, I have burned through six wearable fitness devices while researching and writing this book, each with its own dashboard configuration, and I have tested other non-fitness-related apps that feature dashboards (personal finance, mood, app usage analytics etc.), while also undergoing training in the use of business analytics dashboards. I have kept a keen eye on how dashboards have spread, and I have paid close attention to new and interesting developments, such as the Dublin Dashboard, led by Rob Kitchin and his team of researchers. Eventually, I settled on a case involving the use of commercial dashboard software in the context of hospitals and another based in a situation room which monitors for natural hazards and disasters. I realized along the way that there was no ideal setting for studying dashboards, no case that could hold the burden of acting as an ideal. Although there were obvious similarities that enabled me to recognize the dashboard format as such,⁵⁵ each empirical instance had its divergences and idiosyncrasies. In the wild, dashboards are put to all kinds of uses. My attempts to speak of the format in general, or to outline distinctive characteristics, is made with the recognition that there are always counterexamples and counter-tendencies. I settled on two cases partly because I had the opportunity to observe them in the most sustained way, but also because I consider each exemplary for particular aspects of the format that I wanted to cover in detail.

Chapter 2, ‘Formatting Cognition’, unfolds through a consideration of dashboards in hospitals. The chapter is partly narrated through an interview with a hospital manager, but it also tells the story of Qlik, the business analytics provider I introduced earlier. I explore Qlik’s dashboards in greater detail, drawing on interviews with a number of Qlik employees, documents and observations gathered during visits to the company’s UK headquarters, and web materials, and weave this back into the context of hospital life where Qlik’s dashboards are used. I use this material, however, to explore the cognitive elements of dashboards (recalling that dashboards involve a ‘cognitive function’). Dashboards force us to think about data explicitly in relation to cognition and to consider data as cognitive actors.

The chapter begins with an overview of distributed theories of cognition, focusing on anthropological, philosophical and mediatheoretical contributions. I find these theories of cognition compelling in general, but they are especially well suited as a conceptual backdrop from which to consider the cognitive elements of dashboards, since dashboards are more-or-less conscious attempts to extend cognition. I come to focus on the work of N. Katherine Hayles and what she has called the rise of the technological ‘cognitive nonconscious’ (briefly: things that do cognitive work that is not quite ‘thinking’).⁵⁶ Hayles has done important work formally differentiating between different types of cognition with regard to recent technological developments. I aim to supplement and extend her work by stressing the importance of specific distributions and, especially, types of cognition – that is, all thought is not the same. I offer an empirically and historically informed discussion of cognition in hospital life, formatted through dashboards.

Chapter 3, ‘Formatting Data’, is set in an environmental hazard and disaster situation room in Brazil, where teams of specialists monitor weather-related data and make decisions about issuing warning reports. The chapter explores the composition of the room and what lies beyond it, from displays, infrastructures and specialists, to the types of cognition found within it, and it does so in order to advance an argument about the formatting of data. While a situation room is a very specific site, it is well suited to explore how data are used for monitoring in an ongoing way in the context of making decisions. I use this setting to draw out what is specific about ‘dashboarded’ data, and place this in dialogue with recent debates about the transformative role of data in the production of knowledge. Dashboarded data require a rethinking of how data relate to knowledge, facts and truth. The chapter makes three general claims about data formatted through dashboards. First, they participate in fundamentally ‘uncertain’ ways of knowing. While any data element may increase or decrease certainty, understood in terms of the capacity to make decisions, dashboard data are always in relation to other data and they are always in motion, part of a format in motion. Such motion and relationality, I suggest, produce a constitutive uncertainty that forms the basis of any possible decision-making to follow. Second, dashboarded data prioritize what I call ‘time-value’. It is not that data’s truth-value is abandoned; rather, these data are prioritized in terms of their temporality. Third, and following on, these data are selected, arranged, compared or disregarded according to their capacity to contribute to making decisions. Since time is often crucial to making decisions, time-value is important here as well, but a more general ‘decision-value’ is also observable. The chapter concludes with a discussion of situations and ‘situationness’, which I define as the general dynamics or time-spaces produced by data formatted through dashboards. I use this discussion to further explore the specificity of dashboarded data in contrast to other sites and ways of knowing.