Ó American Sociological Association 2018 DOI: 10.1177/0094306118767649 http://cs.sagepub.com REVIEW ESSAYS

The Sociology of Quantification: Where Are We Now?

ELIZABETH POPP BERMAN University at Albany, SUNY [email protected]

1 DANIEL HIRSCHMAN Brown University [email protected]

Caring Capitalism: The Meaning and Weapons of Math Destruction: How Big Measure of Social Value,byEmily Data Increases Inequality and Threatens Barman. New York: Cambridge Democracy,byCathy O’Neil. New York: University Press, 2016. 226 pp. $99.99 Crown, 2016. 259 pp. $16.00 paper. cloth. ISBN: 9781107088153. ISBN: 9780553418835.

Engines of Anxiety: Academic Rankings, When Bad Policy Makes Good Politics: Reputation, and Accountability,byWendy Running the Numbers on Health Reform, Espeland and Michael Sauder. New by Robert Saldin. New York: Oxford York: Russell Sage Foundation, 2016. University Press, 2017. 168 pp. $31.95 281 pp. $35.00 paper. ISBN: paper. ISBN: 9780190255442. 9780871544278.

The Quantified Self,byDeborah Lupton. that ‘‘sociologists have generally been reluc- Malden, MA: Polity, 2016. 183 pp. tant to investigate [quantification] as a socio- $19.95 paper. ISBN: 9781509500604. logical phenomenon in its own right.’’ While accountants, anthropologists, and historians The Seductions of Quantification: had begun the reflexive study of numbers, Measuring Human Rights, Gender ‘‘sociologists have paid relatively little atten- Violence, and Sex Trafficking,bySally tion to the spread of quantification or the sig- Engle Merry. Chicago: University of nificance of new regimes of measurement’’ Chicago Press, 2016. 249 pp. $25.00 (Espeland and Stevens 2008:402). paper. ISBN: 9780226261287. That has clearly changed. While Google Scholar shows only nine results for the Quantified: Biosensing Technologies in phrase ‘‘sociology of quantification’’ through Everyday Life, edited by Dawn Nafus. 2007, the last decade returns 448. This Cambridge, MA: MIT Press, 2016. 243 proliferation of scholarship on numbers pp. $27.00 paper. ISBN: 9780262528757. goes hand in hand with a proliferation of numbers themselves. New technologies Self-Tracking,byGina Neff and Dawn have created a ‘‘quantified self,’’ and the Nafus. Cambridge, MA: MIT Press, explosion of the internet has produced ‘‘big 2016. 233 pp. $15.95 paper. ISBN: data.’’ As a senior sociologist recently 9780262529129. quipped to one of us, sociology has become quantitative researchers and qualitative

A decade ago, Wendy Espeland and Mitchell 1 We thank Diana Graizbord, Zach Griffen, Phi- Stevens published an essay titled ‘‘The Soci- lip Rocco, and Lisa Stampnitzky for comments ology of Quantification.’’ In it, they wrote on a previous version of this essay.

257 Contemporary Sociology 47, 3 258 Review Essays researchers studying quantification. Thus the avalanche appears to still be rolling along. moment seems ripe for revisiting the sociolo- Activists, investors, and individuals see in gy of quantification, looking at emerging numbers the power to reform states, firms, themes, and seeking signs that a new subfield and their very selves through the governance might be starting to consolidate. capacity of numbers. And yet, skeptics note Alas, the news is mixed. Lots of good work that many of the numbers that get produced is being done. The intellectual space is full of are promptly ignored, while other numbers ferment. Yet—and perhaps the fact that we may simply justify decisions that would are reviewing books by authors from at least have been made anyway. Much like the four different disciplines should have clued broader debate in political science and sociol- us in earlier—so far, we seem to be looking ogy around the causal power of ideas, it can at a genre, not a subfield. be difficult to tease out which kinds of num- Indeed, the very concept of quantification bers matter and when. splinters into fragments as one approaches. Third, how do we govern quantification? Are algorithms quantification? Big data? How should we govern quantification? Biosensors? And what are the differences Contemporary controversies about quanti- between quantification, classification, and fication range from concerns about the fair- commensuration? While there are common- ness of predictive algorithms to the objec- alities across all of these topics, and across tivity of budget estimates to privacy the eight books we read, the sociology of concerns surrounding the circulation of quantification is still very far from having our most intimate choices, captured and general claims or a common theoretical lan- transformed into big data. These controver- guage. The closest it comes, probably, is the sies raise new and old questions about gov- universal reference to Ted Porter’s ground- ernance. As with other technical arenas, breaking Trust in Numbers (1995). quantification often raises significant Yet all is not lost. While the sociology of barriers to democratic deliberation. But quantification may lack a well-defined object some of the numbers we discuss are shock- of study, shared theoretical concepts, and an ing in their simplicity—the calculations agreed-upon methodological toolkit, studies underlying the U.S. News & World Report that touch on quantification nevertheless rankings require no complex algorithms cluster around four broad questions, which and yet still have the capacity to reshape we use to orient our review. The space is cry- the field of higher education. ing out for consolidation and synthesis; and Fourth and finally, how should scholars while we only gesture in that direction, we study quantification? We identified at least hope that in doing so we point to some pro- three varieties of quantification studies. ductive ways of thinking about numbers. First, some authors focus on the effects of First, what shapes the production of num- a particular (often new) genre of quantifica- bers? Here, we are interested in the techno- tion. We examine three texts that look at political decision-making that guides meth- the ‘‘quantified self’’ movement as examples odological choices—who gets to decide of this approach. Second, some studies com- what we quantify and how we do it (on tech- pare different practices of quantification that nopolitics, see Hecht 1998). The project of share some common features or that are making numbers is itself sociological, with mobilized in the same empirical domain. some actors more influential than others These studies are especially valuable for pro- and some numbers easier than others to pro- viding comparative leverage on the ques- duce. A significant part of the sociology of tions of how numbers are produced and quantification is simply showing how social, when they matter. The third approach technical, and political factors interact to consists of case studies that situate a single make stable numbers. calculative practice inside a deep study of Second, when and how do numbers mat- a single field or decision-making context. ter? When does quantification make a differ- This research substitutes over-time compari- ence? Ian Hacking (1982) characterized the son and detailed process tracing for compar- early nineteenth century as experiencing an ative leverage in order to grapple with the ‘‘avalanche of printed numbers.’’ This power of numbers.

Contemporary Sociology 47, 3 Review Essays 259

Together, these books point to a vibrant against women, human trafficking, and com- conversation about quantification happen- pliance with human rights treaties. ing across many different fields. But such Through these detailed histories, she conversations have not yet coalesced into argues that indicators are produced by com- a coherent debate. Rather, these works are munities of shared expertise, embody theo- loosely connected: a genre with recurring ries of social change implying particular motifs, not an integrated literature with modes of action, and are shaped by power coherent terminology and a clear research inequalities among competing experts. She program. also emphasizes that history matters through ‘‘expertise inertia’’ and ‘‘data inertia’’—that is, past decisions about who counts as an What Shapes the Production of expert, how experts are trained, what kinds Numbers? of data are relevant, and what data have actu- ally been collected shape the potential for Making numbers requires a lot of work. Soci- developing new indicators or reforming ologists are intimately familiar with the day- existing ones. In Merry’s context, as in to-day challenges confronting would-be many comparable cases, experts were rela- quantifiers, from crafting question wordings tively limited in their capacity to collect to defining sampling frames, handling miss- new data; and thus existing data sources ing variables, constructing indices, and all of became major resources, and serious the other labor involved in making quantita- constraints, on the creation of global human tive knowledge claims. Researchers interest- rights indicators. ed in the power of quantification foreground Barman examines six cases of efforts to that potentially invisible labor and investi- measure the social value of activities for gate how and why we came to have the purposes of investment and finds that ‘‘value numbers we have. In particular, researchers entrepreneurs’’ draw on their own forms of focus on the experts, politics, and technolo- expertise (which suggest what to measure), gies that shape the production of numbers. in conjunction with their ‘‘communicative Sally Engle Merry’s The Seductions of Quan- goals’’ (do they want the valuation device tification and Emily Barman’s Caring Capital- to establish legitimacy? show conformity? ism tackle this question most directly, though change behavior? justify a field?), to produce most books in this genre touch on these a particular valuation device. The project is issues because it is difficult to trace the organized around understanding how for- effects of quantification without under- profit or nonprofit status and market- or mis- standing its sources. The politics of produc- sion-orientation affect the valuation process. tion contain within-case counterfactuals For our purposes, though, what is most that showcase how a number that may interesting about the book is its analysis of seem solid could have been different and how value entrepreneurs draw on their thus can help to identify the consequences own form of expertise and their communica- of particular decisions. tive goals to build valuation devices. The six Merry and Barman approach the question case chapters, which cover valuation projects with slightly different theoretical orienta- from outcome measurement to socially tions and language, but with similar responsible investing to corporate responsi- methods and very compatible findings. bility, draw on interviews and documents Both books compare multiple cases in a single to show how these come together to produce domain: global human rights indicators for forms of valuation that embody specific Merry; measures of social value for Barman. ways of seeing the world and are tailored to As an anthropologist, Merry approaches the specific audiences. creation of human rights indicators through Both Merry and Barman emphasize the comparative ethnographic work sited in the role of expertise in suggesting what is worth United Nations and U.S. State Department. quantifying and in the political process After a helpful theoretical introduction, Mer- through which a number stabilizes (or fails ry analyzes the creation and circulation of to do so). Put differently, the politics of quan- indicators that attempt to measure violence tification are not open and democratic, but

Contemporary Sociology 47, 3 260 Review Essays closed and technocratic. Experts thus figure breaks down (e.g., Star 1999). To the extent prominently in the central challenges of that indicators and measures serve as infra- commensuration. structure of government and market, they Both authors also show that producing do so best when their own epistemological a number means little unless you convince infrastructures function invisibly. others to use it. Thinking across the books Somewhat surprisingly, neither Merry nor helps to illustrate the usefulness of their the- Barman attend much to the technologies oretical insights. Expertise and data inertia through which quantification is built. (Merry’s concepts) seem clearly relevant for Although Merry usefully names the tenden- understanding the production of numbers cy for ‘‘data inertia’’ to occur and Barman in Barman’s cases. For example, ‘‘Inclusive talks about valuation ‘‘devices,’’ neither Business’’ seeks to address poverty in devel- book emphasizes the bureaucratic or compu- oping countries by including the poor as tational technologies that make possible the employees, customers, suppliers, or distribu- measures and indicators under discussion. tors and by stimulating economic growth. It And perhaps because they are both measures impact by drawing on the numbers focused on somewhat older modes of quanti- already produced by development organiza- fication, neither book addresses the techno- tions, which focus on various forms of social logical advances that promise a continual change. It does not try to measure the finan- revolution in our capacities to measure. cial value of its favored practices for the busi- Cathy O’Neil, Gina Neff and Dawn Nafus, nesses who participate, although advocates and Deborah Lupton, by contrast, all focus see the potential of being able to do so, their gaze on the coming storm. We discuss because data inertia means that those num- these books in more detail in the next section, bers are not easily accessible. but here we note that studies of the quanti- Similarly, we could ask of Merry’s cases: fied self movement and big data-fueled are there ‘‘rights entrepreneurs’’ to parallel decision-making algorithms in arenas like Barman’s ‘‘value entrepreneurs’’? Merry finance and employment all highlight the shows, for example, that although a United immense new quantities of data being pro- Nations office put a great deal of work into duced as parts of other technological and producing human rights indicators, the bureaucratic transformations. ‘‘project was a technology in search of an ‘‘Data inertia’’ may keep human rights audience, a tool kit without a clientele’’ (p. advocates from pinning down their pre- 181). Without entrepreneurs to make the ferred measures of gender violence or link between producing the numbers and human trafficking, but the opposite clear communicative goals for the use of process—a ‘‘data avalanche,’’ say?—is open- those numbers, they were left floating in ing up space for a cornucopia of new forms of midair—formed, but never effective. quantification. Without tackling the question We are also struck by how both Barman directly, Merry and Barman’s work helps us and Merry’s work shows how the political think through how this data avalanche will and epistemological choices that are so be channeled into particular measures and explicit during struggles to establish a num- stabilized (or not) around the political proj- ber disappear once it stabilizes and travels ects of particular expert communities and out from its context of production. Occasion- entrepreneurs. ally, those choices are brought to the front again, as when (to turn to Espeland and Sauder) U.S. News tweaks its What Does Quantification Do? influential law school rankings, but success- The forces shaping the production and stabi- ful quantification projects tend to hide their lization of new forms of quantification mat- assumptions—and the extent to which those ter because quantification itself matters. assumptions remain hidden is perhaps itself That is, there would be little reason to care a measure of the success of the project. Sci- about how numbers are produced absent ence studies scholars have long identified evidence that such numbers had the poten- infrastructure through its tendency to recede tial to powerfully alter the trajectories of into the background unless and until it individuals, organizations, and fields. We

Contemporary Sociology 47, 3 Review Essays 261 consider the range of effects identified by numbers liberating or disciplining? The authors, starting from the self and proceed- Quantified Self movement assumes the for- ing up in scale. mer, and in some cases there seems to be little The books Self-Tracking and The Quantified question that self-quantification improves Self directly tackle the question of how quan- quality of life. Nafus, who began her project tification of everyday behaviors changes our as a self-quantification skeptic, describes subjective experiences, as do many of the her chastened reaction when a colleague authors in the edited volume Quantified: pointed out that knowing one’s glucose level Biosensing Technologies in Everyday Life. Self- can mean life or death for a diabetic. Like Tracking is the ideal entrance into this world Nafus, sociologists are likely to start with for those unfamiliar with the wide array of some doubt—with fear that the technologies new tools that have become available for that we use to monitor ourselves for fun or monitoring our steps, our meals, and our out of curiosity produce data that companies very heartbeats. Authors Gina Neff and use to sell us products and that employers Dawn Nafus usefully distinguish the explicit use to constrain our actions. And of course ‘‘Quantified Self’’ movement (which has all self-tracking enables Foucauldian self- the features of an emergent community or discipline of new intensity, as we map our social movement, from local meet-ups to use of work time minute by minute or exer- conferences) from the broader terrain of per- cise like our health insurance discount sonal data collection (lower-case ‘‘quantified depends on it. self’’). With chapters on the politics of data, The interesting question these books touch how people use data, the self-tracking indus- on, though they cannot fully answer it, is try, self-tracking and medicine, and the how we encourage more of the liberating future, they provide a broad and readable use of quantification and less of the control- overview of phenomena from step counters ling. In a Quantified chapter on health priva- to glucose monitors to activity trackers. cy, philosopher Helen Nissenbaum and cog- Deborah Lupton’s Quantified Self covers nitive scientist Heather Patterson propose similar territory but takes a more theoretical the concept of ‘‘contextual integrity’’: a new approach, framing self-quantification in set of standards in architecture, law, and pol- terms of actor-network theory and Foucaul- icy that differentiates between contexts in dian governmentality. Quantified, a collection which one should expect privacy and others edited by Nafus, is less synthetic than the in which the sharing and sale of data can be other two but more deeply empirical, with assumed or conducted with permission. a number of interesting case studies of how While we agree that shared standards ‘‘biosensing technologies’’ (a phrase Nafus around our right to privacy and use of our consciously deploys over ‘‘quantified self’’) numbers would be very welcome, it is hard are created, governed, and used. to see how we get there from here, as gov- As all three of these books are focused on ernment decisions happen slowly and tech- understanding a particular family of new nological change is rapid. Perhaps a Bill of quantification technologies, they share Data Rights could help us envision such a focus on an overlapping set of questions. standards. One is how self-tracking affects the self. Of course, it is also possible that self- Neff and Nafus show that people use self- quantification produces a subject who is nei- tracking to achieve a variety of goals—to ther liberated nor enslaved, but simply differ- monitor and evaluate themselves (was I pro- ent. The solipsism of constant attention to ductive today?), to elicit sensations (how do I one’s numbers creates a self-involved self— feel at a particular glucose level?), to satisfy one who engages less with the external aesthetic curiosity (what patterns can I see world because of expanded access to the in a map of my bike rides?), to debug a prob- internal one. Anthropologist Jamie Sherman, lem (which foods trigger my migraines?), in another of Quantified’s chapters, asks and to cultivate habits (can I hit 10,000 steps whether self-quantification changes the per- a day?). ception of experience in a way analogous to More deeply, though, a recurring question the one Walter Benjamin described in The runs throughout these books: are the Work of Art in the Age of Mechanical

Contemporary Sociology 47, 3 262 Review Essays

Reproduction. We stop observing experiences and reshape cognitive maps. Law school in their uniqueness, situated in time and deans, professors, alumni, current students, place, and begin seeing them as abstractions applicants—all consume the same rankings, represented first and foremost through the come to understand their professional world data they produce. This may be too much through the rankings’ lens, and make deci- weight to give to the app that calculates sions accordingly. Some of those consumers how far one has run, but the analogy is (applicants) use the numbers as a guide to interesting. action but have little recourse over what Self-quantification is generally a freely they are; others (deans), however, spend chosen activity—at least until your life insur- much of their energy trying to change ance provider requires it. But, as individuals, them. For the deans and others who live we also interact with, and react to, many and die by the numbers, ‘‘fear of falling’’ other numbers that affect our lives. The comes to dominate their experience; few ‘‘weapons’’ highlighted in Weapons of Math subject to this regime think the numbers Destruction are one new and pernicious have improved it. form these can take. In this widely reviewed Although the general view is bleak, Espe- and very readable book, mathematician land and Sauder allow a little room for agen- Cathy O’Neil walks us through numerous cy: individuals can construct narratives examples of models for decision-making around the numbers that tell a different story that are opaque, damaging, and scalable— (‘‘we chose to embrace our mission of pro- the characteristics she says define Weapons viding opportunities to underserved of Math Destruction (WMDs). students’’) than the purely quantitative one The proprietary LSI-R (Level of Service (‘‘we fell in the rankings’’). And their com- Inventory–Revised) model, for example, parison of law schools with business schools predicts a prisoner’s chances of recidivism suggests that when competing numbers from a questionnaire. It uses secret methods exist—business schools refer to several sets and is grounded in factors strongly associat- of rankings, not just one—individuals may ed with race and other forms of disadvan- choose to focus on the numbers that fit their tage. When the numbers produced by self-identity or that play to their strengths. WMDs lead to decisions that affect our lives, The effects of numbers, of course, are not we are forced either to respond to the model just felt at the individual level. The numbers (as when we try to raise our credit score) or also affect communities, organizations, and are trapped without recourse (if the model fields. Even self-quantification is experi- is fully opaque). WMDs often make deci- enced by many as a social phenomenon, sions based on statistical associations that whether in the original Quantified Self are nevertheless not causal—poor credit movement (complete with meet-ups and does not itself cause people to perform worse conferences), in the ‘‘SNPNet’’ forum that as employees. Or they reflect structural collectively puzzles over genotyping results inequalities—prior arrests are associated to improve treatment protocols for chronic with neighborhood, which in turn is associat- illness (as described by sociologist Mette ed with race; their accuracy as a predictor of Kragh-Furbo and colleagues in another of recidivism may have nothing to do with indi- Quantified’s chapters), or in the online com- vidual culpability. Pervasive WMDs may munities of Uber drivers who puzzle out require constant low-level attention: are our how to beat the algorithms and maximize web-browsing habits telling Amazon it their pay. should offer us high prices, or lower ones? Espeland and Sauder are particularly Engines of Anxiety: Academic Rankings, Rep- strong on showing the organizational and utation, and Accountability, in which Wendy field-level effects of law school rankings Nelson Espeland and Michael Sauder tackle and on identifying the mechanisms through the U.S. News & World Report law school which such changes take place. Rankings, rankings as a much simpler yet deeply trans- they argue, have several effects on law formative form of quantification, offers use- schools: ‘‘they transform power relations ful concepts for thinking about individual within schools, day-to-day organizational effects as well: numbers provoke reactivity practices, and the ways professional

Contemporary Sociology 47, 3 Review Essays 263 opportunities are distributed’’ (p. 7). By giv- returns to his neighborhood, he is more likely ing individuals (prospective students, most- to reoffend, having been in the prison envi- ly) more information about schools—if a ronment longer. When he does, the model’s flattened, one-dimensional kind of prediction—offenders from this neighbor- information—they redistribute power from hood will reoffend—is proven correct, and the schools to the applicants. This altering the model is reinforced. of power dynamics is one reason law schools Yet, as significant as these effects can be, may object so strongly to their use. Rankings a great deal of work has to be done in order also give more power to the parts of the orga- to make them happen in the first place. Bar- nization that produce the inputs into the man shows us how hard value entrepreneurs numbers: the job placement office, or the work to gather support for their numbers admissions office. Employees of those offices and convince others to use them. Merry are more at risk of consequences, though, demonstrates how human trafficking indica- should they fail to meet these new tors largely sit on the shelf because no audi- expectations. ence for them has been created. Espeland Within the organization, practices emerge and Sauder choose to study the U.S. News and relationships change in response to rankings because they are so effective, rankings. Marketing becomes more impor- but why are they so effective? Espeland tant (both to generate applications and to and Sauder’s business school comparison improve reputational scores), deans spend suggests the dominance of a single ranking a substantial amount of energy on managing over law schools is one factor. We might rankings, and colleagues at other schools hypothesize other reasons law schools become competitors. And across the field, seem particularly prone to fetishizing rank- the distribution of status and opportunities ings. In PhD programs, in contrast, specific changes. Reducing diverse schools to a single faculty members may play a significant measure means homogenizing them and part in applicant decisions, and thus it may allowing fewer niches for specific types of be clear that the school ranked three places excellence or the pursuit of goals that, while higher is not, in fact, the better choice if in conflict with the rankings, might them- that school has no faculty you want to selves be worthwhile. Small differences are work with. Plenty of questions remain about also amplified; applicants take seriously the not just whether numbers matter or how idea that their life trajectory might be mean- they matter, but when. ingfully worse if they attend #24 instead of #21, and so the schools must as well. Finally, beyond organizations and fields, How Should We Govern Numbers? numbers also have effects on other numbers, This leads us, then, to the ethical questions. and even on themselves. Numbers currently Numbers play a large and increasing role established affect what we will measure in in governing social life. But how should the future, as Merry’s concept of ‘‘data iner- we, in turn, govern the numbers? Must we tia’’ suggests. For example, since the decision simply accept their proliferation in whatever was made to exclude the value of domestic form as inevitable, or are there better and labor from Gross National Product some worse ways they can be used? At some level, eighty years ago, housework has never really all the works we discuss are interested in been accounted for, despite its clear econom- this question, though some address it direct- ic contribution. And numbers affect their ly and others allow us to draw our own own future meaning, as they produce ‘‘reac- conclusions. tivity’’ (as Espeland and Sauder describe it) O’Neil’s algorithm for identifying among those subject to them and come to Weapons of Math Destruction is among the measure something different. O’Neil’s ‘‘per- most explicit. She provides a rubric for decid- nicious feedback loops’’ are the most extreme ing whether any given model for decision- version of this: a model predicts a prisoner making deserves our condemnation. O’Neil from a poor, highly policed neighborhood tells us to ask three questions: Is it opaque? will be more likely to be rearrested and keeps Does it scale? And can it do damage? him in prison longer. When he is out and We might quibble with the ‘‘moneyball’’

Contemporary Sociology 47, 3 264 Review Essays approach to baseball for its lack of romance eligible for benefits for five years. This meant or inability to measure a player’s potential. that over the 10-year period that went into But the numbers that go into such models the CBO score, CLASS would look extremely (runs batted in, strikeouts) are publicly avail- lucrative, offsetting some of the Affordable able, there are clear causal relationships Care Act’s other costs. Yet over a 40-year peri- between those numbers and the desired out- od, it would cost $2 trillion. The program was come (wins), and a steady stream of new data included in order to make the numbers work allows us to test and improve the models. despite the fact that it was completely unsus- O’Neil gives them the thumbs up. tainable. Despite becoming law, CLASS was The LSI-R recidivism model, by contrast, is never implemented, because the Obama opaque. What goes into it? How are scores administration couldn’t find a way to make generated? It scales: twenty-four states use it remotely feasible. it at present, according to O’Neil. And it The transparency of the CBO process is has the potential to do serious damage, keep- precisely what made this gaming possible. ing someone locked up for longer based on Opacity, despite its other drawbacks, may factors they have no control over. O’Neil be the only way to quantify without chang- recommends some WMDs simply be aban- ing behavior. But opacity is no solution, doned (like value-added models of teaching, either, when the actual purpose of quantifica- which may be useful in understanding tion is to change behavior—in this case to aggregate performance but are terrible for encourage more financially responsible leg- evaluating individual teachers) and others islation. The trick is to create quantification be subjected to algorithmic audits, which techniques that encourage the desired would prevent models from making predic- behavior changes (attention to long-term tions based on illegal or unethical factors or costs) but minimize the opportunities for their proxies. These recommendations are gaming (with legislation that produces the appealing, although we suspect defining desired numbers but undermines their ‘‘opaque’’ or ‘‘damage’’ is harder than it intent). This is a needle that is truly difficult might initially seem. Moneyball models are to thread. still proprietary, after all, even if the inputs A few normative lessons do emerge from are clear, and a bad model can clearly dam- these books. Merry, for example, argues con- age a player’s career. vincingly that counts and ratios, despite their Moreover, transparency, which O’Neil relative simplicity, generally stay closer to presents as an unqualified good, can actually the intent of their creators than indicators be a double-edged sword. U.S. News rank- do, because counts and ratios ‘‘require less ings are totally transparent, but it is that interpretive work than composites and are very transparency that encourages reactivity, less vulnerable to the twists produced by as Espeland and Sauder show. Sometimes long interpretive chains’’ (p. 217). Rankings, this is just what is intended. Many numbers Espeland and Sauder remind us, are a zero- are meant to incentivize certain behavior, sum game: ‘‘one person’s climb requires and if their calculation is mysterious, they someone else’s descent’’ (p. 130). And the can’t produce the desired effects. As long self-quantification books highlight the posi- as numbers have consequences, though, tive effects of numbers, as well as their people will try to game them; and transpar- constraints. ency facilitates this. More generally, we would also emphasize Robert Saldin’s short but excellent case that we turn to numbers to make decisions study, When Bad Policy Makes Good Politics: that are already being made through other Running the Number on Health Care Reform, means. Even forms of quantification that provides one such example. The CLASS seem ethically questionable may be better (Community Living Assistance Service and than the alternative that actually exists in Supports) Act, designed to improve long- the world. New Jersey recently ‘‘replaced term care insurance, was included in Obama- bail hearings with algorithmically informed care purely to improve its Congressional Bud- risk assessments’’ (Livni 2017). While get Office (CBO) score. Users would pay O’Neil’s critiques could readily be applied premiums from day one but would not be to the ‘‘Public Safety Assessment’’ formula,

Contemporary Sociology 47, 3 Review Essays 265 in its first six months of use New Jersey’s jail Explicitly comparative projects, like population fell by roughly 20 percent (Kelly Barman’s work on social value, Merry’s 2017). Given how broken the bail system is research on human rights indicators, and and how much it already disadvantages O’Neil’s analysis of destructive algorithms, poor defendants, even a biased algorithm offer important insights about the conditions seems like a step in the right direction. Sub- under which quantification matters. While jective or expertise-based decision-making the U.S. News rankings clearly altered legal is no panacea; and while quantification often education, Merry’s work shows how little crystallizes existing problems, fixating on it oomph some rankings possess. Absent can be a distraction, rather than a solution. a comparative lens, we collectively tend to These examples all highlight the need to fixate on the most prominent forms of quan- focus on the how of quantification as much tification (like GDP or presidential approval as the mere fact of it—as with any other polls) and miss the forest of much less potent powerful tool, the effects depend on inten- indicators that surround them. Comparing tions and implementation. across cases within a domain reminds us of the mundane organizational and political factors that shape the reception of quantifi- How Should Quantification Be Studied? cation by highlighting the presence and Studying quantification is a growth industry. absence of powerful, motivated actors push- Alongside the rise of data science in indus- ing or resisting forms of quantification. try, academia, and government we see a par- Comparing across domains, as O’Neil does, allel emergence of new dedicated venues for productively highlights the distinctive examining these transformations, including aspects of contexts that may inhibit or privi- the Data and Society Research Institute and lege quantification, showcasing why (for journals like Big Data & Society. What can example) algorithmic decision-making is rel- we learn from this recent generation of atively uncontroversial in credit but fiercely quantification studies to inform the path fought in education. ahead? Finally, studies of new quantified phenom- We see three rough approaches represented ena help to sharpen our conceptual catego- here: charting a single (often new) quantified ries for thinking about what quantification phenomenon, comparing quantification proj- really is. Quantification is not a single, uni- ects (often bounded to a single field or fied process. Work that explores the bound- domain), and tracing a particular historical aries of numbers, data, and measurement case where quantification took center stage. helps to flesh out our vocabulary—from indi- Each of these approaches offers particular cators and rankings, to models and algo- advantages and characteristic weaknesses. rithms, and now big data and biosensing, Tracing a single case, as in Saldin’s history and so on. Indeed, one of the clearest takeaways of the CBO and health reform, or Espeland from these books, read as a group, is the blurri- and Sauder’s study of the effects of law ness of ‘‘quantification’’ and the need for con- school rankings on legal education, may be ceptual categories that will help us unpack it. most useful for thinking about the mecha- What qualities are specific to rankings, or indi- nisms through which relatively influential cators, or models, or algorithms? What does forms of quantification exercise that influ- quantification share with related concepts like ence. The nitty-gritty of the policy-making commensuration or categorization? process and the ability to compare organiza- Some phenomena that, when one is tional practices before and after the emer- primed to think of them in terms of quantifi- gence of law school rankings provide rich cation, seem to fit right in aren’t quite exam- within-case details that illuminate how par- ples of quantification at all on closer inspec- ticular numbers do their work. These cases tion. Nafus discusses her decision to talk are less useful, however, when it comes to about ‘‘biosensors’’ rather than the ‘‘quanti- specifying scope conditions. This problem fied self’’; when one is interested in under- is not unique to single case studies of quanti- standing, for example, how heart rate moni- fication, but it does impel us to reach for tors change human experience, which of more. these terms directs our attention to the most

Contemporary Sociology 47, 3 266 Review Essays relevant aspects of the new technology? Or of talking about these diverse phenomena when, as in Barman’s work, valuation of an and to build insights across a broad range organization’s social impact produces not of quantification processes. The qualitative a number but a binary yes/no, are we still study of quantification may, at the moment, talking about quantification? Maybe not, be producing a significant body of new and yet the process of producing that binary work. But for producing a coherent sociology seems quite similar to that of calculating the of quantification, we have a long way to go. social return on investment. Following Barman, then, we see promise in distinguishing more explicitly among the References intended uses of numbers—their ‘‘communi- Espeland, Wendy Nelson, and Mitchell cative goals’’ (or, we would add, ‘‘performa- L. Stevens. 2008. ‘‘A Sociology of Quantifica- tive goals’’). Law school rankings have both tion.’’ EuropeanJournalofSociology49(3):401–436. an explicit purpose and an intended audience; Hacking, Ian. 1982. ‘‘Biopower and the Avalanche these shape the process of quantification and of Printed Numbers.’’ Humanities in Society 5:279–295. itseffects.Sodohumanrightsindicators,and Hecht, Gabrielle. 1998. The Radiance of France: CBO scores, and activity trackers. Though Nuclear Power and National Identity after World numbers in the wild may certainly be used in War II. Cambridge, MA: MIT Press. ways their creators never anticipated, clearer Kelly, Albert B. 2017. ‘‘Six Months In, N.J. Bail and more explicit attention to the purposes of Reform Is Working.’’ NJ.com. Retrieved Sep- both producers and consumers would help tember 4, 2017 (http://www.nj.com/opin ion/index.ssf/2017/07/six_months_in_nj_bai us think more clearly about variation in the l_reform_is_working_ opinion.html). quantification process across contexts. Livni, Ephrat. 2017. ‘‘In the U.S., Some Criminal There is also plenty of room for more com- Court Judges Now Use Algorithms to Guide parative work that explores how numbers Decisions on Bail.’’ Quartz Media. Retrieved are produced (the same number in different September 4, 2017 (https://qz.com/920196/ organizational or social contexts, or different criminal-court-judges-in-new-jersey-now-use- algorithms-to-guide-decisions-on-bail/). numbers in similar contexts) and on when Porter, Theodore. 1995. Trust in Numbers: The Pur- and how they have effects (including suit of Objectivity in Science and Public Life. comparisons of failure and success, as well Princeton, NJ: Princeton University Press. as variation across successful cases). And Star, Susan Leigh. 1999. ‘‘The Ethnography of there is certainly more space for synthetic Infrastructure.’’ American Behavioral Scientist work and theory to give us a common way 43(3):377–391.

Who Owns Lefebvre? The Forgotten Sociological Contribution to the New Urban Sociology

MARK GOTTDIENER University at Buffalo-SUNY [email protected]

‘‘Edward W. Soja, Fredric Jameson and Mark Gottdiener played a key role in introducing Marxist Thought and the City,byHenri Lefebvre’s thinking about space into the Lefebvre, translated by Robert United States’’ (Trebitsch 2005:xxiv). Bononno. Minneapolis: University of Minnesota Press, 2016. 158 pp. $20.00 Who owns the intellectual commodity paper. ISBN: 9780816698752. ‘‘Henri Lefebvre’’? I ask this question because of the disconcerting way his work has been disseminated, absorbed, and established social science. He has been commented on in the Anglophone sphere called a critical geographer or a Marxist phi- of academia. He is hardly mentioned in losopher, but he was all his life a professional

Contemporary Sociology 47, 3