Every Relevant Detail

Salomé Viljoen* April, 2025

EVERY RELEVANT DETAIL¹

The Ordinal Society. By Marion Fourcade and Kieran Healy. Cambridge: Harvard University Press. 2024. Pp. vi, 373. $45.

Twenty years from now, the idea that someone looking for love won’t look for it online will be silly, akin to skipping the card catalog to instead wander the stacks because ‘the right books are found only by accident.’

—Wired Magazine, 2002 ²

I’m so sick of online love.

—Chappell Roan, 2024 ³

Introduction

Since Match.com launched in 1995, finding love online has moved from fringe practice (seen as something for the freaks and geeks, the sad and desperate) to the mainstream. Online love is big business—a near-billion-dollar industry in the United States alone.⁴ As of 2024, nearly half of Americans had met potential romantic partners on dating apps.⁵ Most Americans under thirty date via apps, as do a majority of lesbian, gay, and bisexual Americans of all ages.⁶ As of 2023, over 380 million people globally used online apps to find love.⁷

Love now happens online. Archetypical pre-online social forms of romance developed around informal social networks: college life, dinner parties, blind date setups between friends. A sense of serendipity and accident prevailed—love finds you when you least expect it. As evidenced by the rom-coms and TV shows of the 1990s and early aughts, love needed some dose (or, more cynically, a veneer) of happenstance to be done right. Hence an understanding of love wherein the meet-cute is venerated and the blind-date is considered a bit desperate and cringe. Ideally, one must let go and let love.

These old social forms of love have been replaced by digital swipes, likes, and matches. Success on the apps operates on a very different kind of social logic: Love is not a happy accident, but a game of moneyball. On dating apps, love is statistical. With a sufficiently large “n” and the science of optimal presentation, a person can achieve their “correct” rank via the win-loss distribution of dating. And there, balanced inside their dynamic pocket of accurate valuation on the ever-updating digital love market, a person might just find their match.

The transformation of love into online love is one instance of what Marion Fourcade⁸ and Kieran Healy⁹ call the “ordinalizing tendency” (p. 267). In The Ordinal Society, Fourcade and Healy trace this tendency across the arc of “the changing relationship between information technology and power” to argue that ordinality has become contemporary life’s organizing principle (p. 3). An ordinal society is a society of measurement and meritocratic ranking. It describes a social order “oriented toward, justified by, and governed through” classification and ranking (p. 1). We understand and express ourselves and experience the “ebb and flow of social and economic life,” via rankings and classification (p. 2). The result is a society in which “rationalized stratification” orders how people live (p. 2).

Technology can affirm, thrill, frustrate, shape, discipline, and delight us. Fourcade and Healy want us to step back from this first-person experience to take a systemic view of digital technology, too. Technology is also a means of social organization—a machinery that continuously categorizes us in the recursive loops formed between our actual behavior and interactions on the one hand, and our derived future of predicted behaviors and curated exchanges on the other. This recursive looping determines our future possible horizons. Fourcade and Healy diagnose ordinality as the central logic and ideology of digital social life. They aim to show how ordinality determines both how we make sense of and move through the world, and how the world makes sense of and moves through us.¹⁰

Coordinated ranking and matching across organizations “sort and slot people into situations and positions” (p. 261), where rank is derived from “objective measurement of actual behavior” (p. 260) and matching is based on criteria derived for a specific purpose such as finding a partner, hiring an employee, obtaining a loan, or getting a ride. Ranked matching is a market logic and a money-making scheme.¹¹ It is also a cultural feature of how we approach forms of self-identity. Personal activities—from step counts to Spotify wrapped playlists—are subject to pervasive quantification, and self-esteem is increasingly mediated through digital recognition by others via likes, retweets, and follower counts.

The Ordinal Society is an excellent addition to the collection of works seeking to provide a systematic theory of digital life. It is more accessible than Julie E. Cohen’s seminal work, Between Truth and Power: The Legal Constructions of Informational Capitalism¹² (though for legal readers interested in the topic, I will second Amy Kapczynski in saying Cohen’s book is well worth the effort¹³) and contains less polemics and more insight than Shoshanna Zuboff’s popular doorstopper, Surveillance Capitalism.¹⁴

I found the argumentative style of The Ordinal Society to be inscrutable at times, in the mode of other works of social theory. The book can jump between discrete anecdotes and sweeping statements about complex topics in ways that leave its argument unclear. There are also instances where the book extends its analysis beyond what it persuasively describes. For example, the financialization dynamics described in chapter 5 did not strike me as forms of ordinalization. Rather, the chapter details examples of how digital technologies like NFTs or meme stocks can serve as instruments of speculation and hype (pp. 170–71). These digital-financial fads are perhaps solutions in search of a problem, or perhaps just expressions of late-capitalist Id. Clearly some social logic is at work here. But I remain unconvinced that it is one rooted in ordinalization.

However, The Ordinal Society has many strengths. It balances a concise theoretical account of digital life with the authors’ specific theoretical contributions, both of which help the reader understand the genesis and logic of contemporary digital life. The authors acknowledge the individual thrills and pleasures of technology early in the book (p. 60). This sounds a welcome note for the genre of tech criticism, which tends to favor both systemic and skeptical perspectives.¹⁵

Relatedly, the authors do not shy away from the internet’s lessons on the (un)popularity of going it alone. If digital life is a cage, it is one we had a hand in building, and it is one whose door stands at least partially open. “Even today,” they write, “there is nothing intrinsic to the architecture of the web, nothing in its underlying protocols, that prevents the kind of widely distributed, robustly local, essentially decentralized network of free communicators” that early proponents of cyberspace imagined and that calls to reform digital life continue to feature.¹⁶ But it turns out that people don’t really want to “build a little online homestead” (pp. 18–19). Instead, they favor services that make it easy for them to interact, shop, and perform tasks online, even if that means giving up data and control. The internet’s history is a cycle of repeated consolidation and control at least partially because living in civilization is nice. It is nice not to be individually responsible for all the know-how and struggle involved in reproducing one’s life. The same goes for one’s digital life.

This is not to make light of the gravitational forces that constrain the scope of individual agency and shape the horizons of digital life (p. 27). But the perennial calls to change that life or expand and alter those horizons through the (re)decentralization of digital platforms may be less feasible, or face steeper and broader barriers, than proponents care to admit.

Most notably, The Ordinal Society places central emphasis not on control, but on hierarchy. Control was and remains the dominant moral and analytic register for talking about digital life.¹⁷ Technology’s capacity for controlling people is part of the account here, but Fourcade and Healy want to shift our focus to how computerization unfolds as a social process (perhaps unsurprising, since they are, after all, sociologists). This not only frees their account from the classic preoccupations with (loss of) individual freedom but also allows for the dynamism of digital systems to come into view. The digital plane that permeates contemporary life is not static, nor is it a totalizing system of control. Understanding it requires toggling perspective from the first person to the third. Only then can we grasp the process that unfolds when the institutional (i.e., market and state) imperatives to impose order and control meet the inevitable messy “overflow” of individuals experiencing, enjoying, and using things in unintended, even subversive ways.

In its strongest sections, The Ordinal Society traces how the “exfiltration of sociality” in the form of data gets incorporated into industry’s value-making strategies, linking the capital accumulation machinery to the classification project (pp. 58–66). This focus on how datafication has produced a political economy of rationalized hierarchy, rather than the traditional emphasis on the resultant capacity for increased control, is, to me, exactly right.¹⁸ As adoption of these methods becomes widespread, classification situations become more inextricably tied to how one fights it out for one’s life chances in the market.¹⁹ It is a mark of a good theory that once you encounter it, you begin to see it everywhere. Take, for example, the recent debate within the Democratic party over data-driven messaging.²⁰ As someone familiar with the history and dynamics Fourcade and Healy describe, I still find several of their concepts genuinely analytically insightful.

Of my quibbles and qualms with the book, the biggest is that I am left unsure about the authors’ stance on the trouble of ordinality. Fourcade and Healy could fairly insist that their account is substantively descriptive. However, The Ordinal Society’s ending sounds in a distinctly negative register: Life in the ordinal society, the authors note, “may well be unbearable” (p. 285). I agree. All this social transformation matters, the authors suggest, because classification invites trouble (p. 109). But what kind of trouble?

Parts of the argument suggest that the move from life to online life is regrettable because it replaced a (relatively) (comparatively) open social field with a more determined and constrained one (p. 192). We may still be captains of our destiny, but we are sailing in a bathtub. Life’s experimental mess and play is constrained to the narrowed and impoverished terrain of Silicon Valley’s design (p. 275). Perhaps then, the trouble is not one of inequality per se, but one of ossification. The result of such a stultifying existence is individual alienation and disempowerment.

However, the same ordinality that rewrites the first-person experience of life also remakes the systems that dole out life’s opportunities. Rationalized stratification, the authors also seem to argue, makes our opportunity structure more sclerotic, (arguably) more unfair, and (certainly) less up for collective determination (pp. 2–3). So perhaps the problem is structural and distributional or a matter of limited democratic accountability.

But, as the authors note, not all inequality—and not all ranked classification—is an instance of injustice. For example, in both online and offline love, some people are more romantically desired than others. We should be hesitant to claim people have a right to be equally desired.²¹ Does the trouble lie in the rationalization of hierarchy then? Not always. Some hierarchies are wrong independent of, or prior to, their rationalization. For example, digital systems that recreate durable social inequalities along racialized or gendered dimensions may be unjust. But they are unjust for the same reason analog instances of racialized or gendered inequality are unjust.

So is the problem the inequality that ordinalization re-encodes, or is it the loss of life’s charming, spontaneous mess? In other words, does the problem with rationalized stratification lie chiefly in the rationalization, or in the stratification? While I share the authors’ concern about datafying society, diagnosing the source of that concern matters a great deal. How the law might attempt to eke out some measure of justice in an ordinal society depends on what ails such a society.

Part I provides an overview of Fourcade and Healy’s argument and introduces readers to ordinalizing as the process of reorganizing social worlds along the lines of determinate systems with rules-based competition. Part II dives deeper into the account of ordinalization, considering its key inputs—data and the intellectual technologies of social science that build digital ranking and matching. It then shows what ordinalization produces for its subjects. Part III situates The Ordinal Society in the universe digital technology and society scholarship. It considers the strengths and limitations, particularly for legal readers, of work that describes and critiques the current state of affairs but remains vague about what precisely has gone wrong in our embrace of rationalizing social life—and thus what, if anything, ought to be done to set things right.

I. From Open to Closed Worlds

Using computers to mediate social life commits us to a certain view of how social life does, and ought to, reproduce itself.²² Scholars of technology broadly agree that datafying social life is neither neutral nor incidental.²³ To paraphrase Milton Friedman, digital technologies are engines, not cameras.²⁴ They not only allow for more extensive and accurate recorded information about the world but also enable new changes to and activities in the world.²⁵

The reproduction of social life is always mediated and never neutral. But understanding social life today requires interrogating the specific politics and total mediation afforded by online life. If nothing else, digitalization is a remarkable achievement in scale and concentration. Never before have so many billions of hours across billions of lives been spent in common, in such immersive and responsive environs. And never before have such a small handful of actors been in charge of such extensive ecosystems of information and knowledge exchange.

Fourcade and Healy’s specific thesis emphasizes that computational methods make sense of—and intervene in—society via automated ranking and matching (pp. 1–3). In their account, digital life is a classification engine (p. 2). By datafying life, we render it not only more measurable but also more ordinal, or ranked. The activities, practices, and scripts of social existence increasingly operate by placing people into their “correct” or “accurate” ranked position.²⁶ Ordinalization, as a practice and an ideology, requires datafication to work. Ranking occurs through interaction and exchange, but it is only due to the carefully designed and comprehensive data flows about people in these interactions that rankings can reflect and produce, in real time, the personalized slice of the opportunity structure that one’s ranking affords.

As an argument, The Ordinal Society needs to prove two interlocking hypotheses. First, that a central feature of life in a computational society is that we are ranked. Second, that something about how or why computational systems rank us departs from how prior societies produced social order. The Ordinal Society does not argue either hypothesis directly. But it does lay out that argument’s basic elements over the book’s several chapters. And the result is largely persuasive.

A. Overview

The book begins with a history of computing as a ranking and sorting technology. It then reviews the kinds of technical and intellectual developments needed to create a ranked society, introduces the kinds of individual and societal goodies ordinalization endows to reproduce itself, and shows how ordinalization works to move social life’s “open worlds” closer to the “closed worlds” of online life.

The first section of the book canvasses a familiar story. It provides a historical recap of Silicon Valley, the rational sciences and computing technologies of the twentieth century that preceded it, and the intellectual technologies²⁷ and datafied inputs²⁸ necessary to the creation of Webs 1.0 and 2.0. The authors present this account through their unique lens. Chapters one and two frame data as an exchange between users and platforms akin to the pre-modern practice of ritualized gift exchange.²⁹ The authors trace how the gift exchange of data for the delights of online life produced the “exfiltration of sociality,” or the near-depthless reserve of analytic possibility offered by the sea of data about us (p. 58). This analytic reserve is a necessary precondition for matching and ranking to occur at the granularity, scale, and speed of online life.³⁰ This early portion of the book recounts how the technological capacity to collect data empowered the intellectual and moral virtues of numerical and statistical value, most notably in its capacity to depoliticize the political. What began as a project of statecraft³¹ evolved into a lucrative activity as data became an asset for companies with the business savviness and technological capacity to make sense of it.

Chapter three, the heart of the authors’ argument, lays out how algorithms engage in social classification through looped “cybernetic triage”: “[T]he analysis of tracked and classified behaviors forms the basis of differential treatment, thus affecting social stratification through the allocation of similar sets of opportunities to similarly situated people—what sociologists since Max Weber have called ‘life chances’ ” (p. 104). The rationalized stratification of life chances creates what the authors call classification situations: the positions people find themselves in (for better or worse) within matching algorithms, feeds, curated exchanges, platforms, and digital products (pp. 104–05). These “classificatory infrastructure[s]” of online life engage in ceaseless ranking and slicing (p. 104). Organizations rely on these ranked classifications to produce value (p. 104). This, in turn, produces the settings where people socialize with one another.

The resulting taxonomies overlap with, but are distinct from, demographic categories like race, sex, gender, and class—the categories that normally feature in descriptive and normative accounts of social life and its durable inequalities.³² Online classification occurs via “post-demographic” features that are behavioral and probabilistic instead (p. 104). These categorize people based on the things they are likely to like or do.³³ However, like demographic categories, a person’s position in these relationally generated categories will have consequential effects on her life outcomes (pp. 104–05). Fourcade and Healy name the value of one’s social position “eigencapital” (pp. 117–18). Eigencapital draws on and extends Bourdieu’s theory of cultural capital,³⁴ to index the resources one enjoys due to one’s social location.

The second half of the book then turns to consider how ordinalization produces and remakes specific institutions and forms. Chapters 4 and 5 review how classification situations produce value in the market; while chapters 6 and 7 focus on how ordinalization reconfigures political life. These chapters are engaged with at greater length in Part II below.

B. From Open to Closed Worlds

In the early 2000s, Microsoft patented TrueSkill, a ratings method that built upon longstanding methods to rank players based on head-to-head competitions by extending the approach for games with more than two players (p. 267). TrueSkill is the backbone of Xbox’s online gaming platform, powering its leaderboards and tiered-ranking systems. This system matches millions of users across thousands of games, constantly updating the rankings with the results of each new game. TrueSkill is, the authors note, a “near perfect exemplar of the twofold character of an ordinalizing system” (p. 267). For one, it identifies the best players and elevates them to a leaderboard for others to see. This gives players a goal—if they put in the time and develop the skills, they can climb the ladder to higher and higher public standing. Second, categories and tiers of players “fall out” of the ranking. This allows players to be matched against opponents whose past success rates and skill levels roughly match their own, and who therefore are neither far too difficult nor far too easy for them to face (p. 267). This maximizes all players’ satisfaction.

“The ordinalizing tendency,” the authors write, “is to bring real-time ranking methods from the closed to the open world” (p. 267). In closed, fully constructed environments like online gaming, the terms of success are set by designers and are “fully tunable” (pp. 267–68). By contrast, in the open, social world, success and failure are “still definable,” but winning and losing play out in more diffuse, high stakes ways (p. 268).

What makes algorithmic ranking and matching especially powerful, and perhaps especially troubling, is the way the technology acts to make the social “open world” more like the rule-based “internal” worlds of games. Google and Facebook’s ad auction technology, the two-sided ranking mechanisms of Uber and Lyft, and dating apps all deploy two-sided matching mechanisms to iteratively assess and determine individual standing within the mechanism.³⁵ Applying the powerful insight of Google’s PageRank system—that “status is best conceived as the high regard of highly regarded others” (p. 268)—these systems weigh individuals’ assessments by the scores of those assessing them.

Moving from the diffuse social processes of the open world towards a system of rationalized ranking empowers certain institutions and actors. The companies and designers behind a ranking mechanism can claim an outsized role in determining social order and designating the fruits of social esteem. By moving to algorithmically designed environments, ordinal systems don’t just produce groups and social strata, they also “set[] the metrics through which competition is assessed” and reward the “players” who successfully compete on the terms set out by the system designers.³⁶

The authors correctly note that algorithmic ranking’s novelty is not the mere fact of quantification, nor the mere fact of ranking. Categorization—applying rule-based “algorithms” and taxonomizing attributes—has always worked to draw boundaries between groups of people. Algorithmic methods are unsettling. But not simply because they are opaque and unaccountable. Nor because they are built from shoddy and imperfect data and violate our privacy. While these critiques apply, they miss the generalized social upshot of ordinalization: that it transforms and reorganizes the process (and purposes) of categorical meaning-making itself.

Others, with distinct emphases, have made similar claims. For instance, Shoshanna Zuboff has raised the concern of epistemic capture—that platforms have enclosed and appropriated the process of knowledge and meaning-making.³⁷ Zuboff, however, is chiefly concerned about where this leaves the individual in relation to the public sphere and the impact of epistemic capture on the conditions requisite to cultivate democratic citizens.³⁸ Cecilia Rikap is similarly concerned with firms that have captured the knowledge formation process. However, as a political economist, she primarily focuses on how the structural advantages of creating knowledge for large incumbents (better data means better ranking and matching systems) produces what she calls “intellectual monopolies.”³⁹ In contrast, Fourcade and Healy neither focus on the information ecosystems individuals need to be good citizens, nor (at least centrally) on how the mass enclosure of data creates long-term structural advantages for the largest firms. Instead, they are interested in how digital meaning-making operates to re-encode and rematerialize the social structure.

II. Inputs and Outputs

There is a push and pull to ordinalization. Individuals can feel a magical, seamless rightness when ordinality “just works.” This can be as simple as an algorithmic social media feed serving up the perfect off-beat joke or a car pulling up right as you need it on a city corner. This is not only an affirming experience, but over time, an affirmation of ordinality as an ideology.

On the other side, there is frustration when our matching algorithms fail. When we are not ranked highly (or are shunted out of the game entirely), we feel the arbitrariness, the artifice, and the “unjustness” of the system. Users thus afflicted may accuse a system of “shadowbanning,” a claim that a deceptive and unfair algorithmic interference has been applied to content that “artificially” reduces its reach. More troubling are the hiring algorithms, lending algorithms, and tenant-screening algorithms, all of which may flag a subject as high-risk in ways the subject may never see and for reasons neither subjects nor the employer, lender, or landlord even understand.⁴⁰

Sufficient friction can even lead to exploitation or attempts to “game” the rules of the game itself. Taken to its extremes, exploitation can also lead to the ridiculous. Consider one recent example about Elon Musk. Frustrated that his posts were not receiving the attention he believed they deserved, he worked with X (formerly Twitter) engineers to alter the platform’s algorithm so that it more heavily promoted his content.⁴¹ If this desire for exploitation reflects certain background social facts, it can create a market opportunity. Institutions that trade on the legitimacy and merit of ordinality can begin to offer, as a service, “cheats” to undermine its own methods.

Paid subscriptions for online dating sites are a prime example.⁴² Paying users can, in essence, supplement the “eigencapital” of one’s desirability (as determined via the ordinal system of a matching algorithm), with the monetary capital of a paid subscription. This may offer dating apps a surer pathway to financial viability, but it also undermines the basic logic and legitimacy of the ranking and sorting mechanism. By allowing a paid tier of users to avoid being ordinalized, dating apps reveal how ordinalization both supports and undermines the dynamics of innovation. The tendency for online systems to engage in a self-undermining, downward-spiraling ouroboros of classifying, matching, and then interfering is what Cory Doctorow refers to as the “enshittification” of digital systems.⁴³ This process is set off by individuals who are both seduced by and exasperated with ordinalization and by firms that attempt to have their ordinal cake and eat their paid subscriptions (or big-money advertisements) too.⁴⁴

Fourcade and Healy are at their best when they keep this duality firmly in view—which, to their credit, is often the case. Compared with the totalizing, and thus overdetermined, account of Zuboff, their “soft structuralism” is more accurate while still providing a generally coherent frame of analysis. In particular, their approach allows explicit room for the messiness of social transformation. It enables the reader to trace how ordinalization as a theory of power works itself out through freedom—via the “mechanisms of the market and the imperatives of self-realization.”⁴⁵

Ordinality as an organizing social logic requires a few things. Most notably, it needs large reserves of granular data and analytic technologies to value and make sense of them. Closely related, yet distinct, it needs digital matching mechanisms to encode and enact data analysis as categorical ranking. Ordinalization also results in new capacities and worldviews. These inputs and outputs are discussed in greater detail below.

A. Data Value and the “Exfiltration of Sociality”

Fourcade and Healy correctly diagnose the scale, and the point, of social life’s datafication: “[W]hat the largest contemporary organizations have at hand is the immense yield of the social substrate in the broadest sense, a source of data about what people are doing, writing, or saying . . . that exists in a form that can be parsed, saved, and mined for information” (p. 86). The Ordinal Society is right to emphasize the systematic nature of data assets. Taken as a whole, mass datafication of human behavior produces a resource that approximates the generalized social structure.⁴⁶ This does not merely pose a privacy risk for the individuals (although a mass looting of privacy is required to produce this resource). It provides a source of potential value production that can be pursued using a broad swath of strategies.⁴⁷ Such projects often entail concerted and local efforts to tweak social replication’s imperfect processes to serve a designer’s own ends (pp. 102–04).

Accounts of data value are by now widespread. The Economist famously dubbed data “the new oil” in 2017.⁴⁸ Shoshanna Zuboff, Erik Brynjolfsson, and Andrew McAfee also view data as a commodity, though they disagree profoundly about whether data commodification is good.⁴⁹ Others, myself included, argue data is more appropriately understood as a product of social relations that can be encoded as capital—but that data’s status as capital is downstream of both political economic pressures and of law. ⁵⁰ Whereas the authors draw on Bourdieu and Mauss in their account of data value (pp. 64, 67, 270), Julie E. Cohen extends Polanyi to provide an account of social data as a vital factor in contemporary production. She also provides an important account of the legal choices involved in constructing data value. In my view, Cohen’s is the more persuasive account of how data produces value in the informational economy.⁵¹

As Jathan Sadowski notes, “[d]ata is not out there waiting to be discovered as if it already exists like crude oil and raw ore.” ⁵² Fourcade and Healy similarly emphasize that data isn’t a natural information repository; it instead requires constant and active construction. It needs to be organized, managed, cleaned and stored in a very specific way to be useful for extracting insight. Moreover, distilling useful insights from data is far from straightforward and often quite hard (p. 88). It is, to borrow from Fourcade and Healy, “easy to do badly” (p. 88).

Data value, in their account, is not just the creation of a capital asset, but a “cultural and political accomplishment” in achieving certain institutional norms and beliefs about what good business operations look like (pp. 76–77). The belief that “everything is data and is there for the taking” requires “a collective, sustained overhauling of the sociomaterial environment” (p. 93). Adjusting human exchange to allow for extensive data collection and to index behavior via logins and reliable sustained tracking is a huge material undertaking.

Central to the authors’ account of data value and of ordinality is the concept of eigencapital.⁵³ The authors use eigencapital to refer to the value of one’s social location—the value that “incorporates whatever value is in your social network, along with synthetic measures of your trustworthiness or accountability in the world” (p. 117). The authors admit that one’s eigencapital is more of a theoretical concept than reality. The sum of behavioral and relational data about us is “not all gathered into a single place or condensed down to a single quantity” (p. 117). But in principle, some “vector of information” that summarizes our ordinal position across many interfaces would represent our “position in the multidimensional space of classification situations” (p. 117). This quantum would “characterize [our] social location” (p. 117). And in the reputational and relational world of the ordinal society, this location itself is a form of capital—value that an individual can trade and exercise.⁵⁴

Eigencapital is an extension of Bourdieu’s concept of cultural capital—the idea that people may possess some resource that, though not distilled into a direct measure of income or wealth or class position, benefits them in life.⁵⁵ Cultural capital is famously hard to measure, and its “flexibility of application” can be fairly held, at times, to “make[] a virtue of endogeneity” (p. 120). However, I agree with the authors that there remains something “deeply true” about the insight that some people have internalized a set of resources that makes life go more smoothly—that equips them to navigate the inequities and obscure rules of life with greater ease and success. Forms of social or cultural capital index the value of an elite milieu. In this sense, eigencapital names the endowments that flow from our place among the ranked systems of digital life.

Eigencapital is somewhat distinct from other forms of cultural and symbolic capital. It has a “generalized, relational character” (p. 121). Its meaning and value are context-specific, but both are, in principle, calculable. I find eigenvalue quite useful to theorize the sociological endowments of “data doubles.” I remain less sure it is the best concept for analyzing the practices of monopolization and market power surveyed in chapters 4 and 5—where indeed, the concept seems to recede from the authors’ analysis. In such contexts, social data can be deployed in value accumulation strategies that apparently have little to do with their ability to act as a locational referent for a given data subject in the social structure. More accurately, the same stream of social data can be simultaneously deployed in multiple value accumulation strategies—only some of which are about its value as a marker of the data subject’s reputational or ordinal location.

Take for instance the prodigious volumes of driver data Tesla cars generate. Tesla uses this trove of driver data to engage in a variety of business practices the authors characterize as the unbundling and bundling of digital goods: bundling physical goods with datafied connectivity and commodifying data to produce new informational goods that disaggregate economic rights (pp. 158–59). Driver data produces value in several ways.⁵⁶ First, engineers can use driver data to make cars safer and more efficient. These improvements derive value from data by improving Tesla’s product—the kind of true or ideal innovation that both benefits drivers and rewards the company for making more efficient, safer cars.⁵⁷ Another strategy is to use behavioral nudges to improve driver behavior.⁵⁸ Driving data from many cars can be aggregated and used as training data to further automate driving. Companies like Tesla that have access to superior aggregate data possess an edge (in theory) in developing automated vehicles. Tesla may also sell generalizable insights about driver behavior to other companies, perhaps who are developing automated vehicles of their own, or that want insight into Tesla drivers’ movement and behavioral patterns for other purposes. Tesla can use aggregate driving data to establish a baseline against which it then compares the driving behavior of individual drivers—either using this to develop its own insurance products or selling this to insurance companies to help them with both general and individual rate setting.⁵⁹ Finally, Tesla can use driver data as a contractual enforcement and accountability mechanism whereby it catalogues vehicle use as a dossier of evidence against any future liability claims.⁶⁰

Importantly, only some of these strategies appear to trade on individuals’ eigencapital or derive from ordinality. These might include setting a driver’s insurance rates, using driver behavioral data as a ledger of contractual compliance between the driver and Tesla, and nudging drivers into safer driving. Drivers and Tesla may be aligned in some strategies, whereas in others, they may have a more adversarial relationship. Whether the production of informational value harms or benefits a subject is highly contextual. How information is acted on to produce value may itself loop back into further action that raises or lowers subjects’ standing in the multidimensional social structure—sometimes in competing ways, along the same informational vector. The “value” or “power” of information to raise or lower individual standing is not endogenous to the value of the information itself; rather, it is determined by institutions’ choices about how to guide action based on that information and the laws that constrain and guide such action. What is powering firms’ new profit strategies is not limited to their ability to estimate, trade against, and reward or exploit individuals’ eigencapital. Instead, there is distinctive value in the aggregative data for predictive and adaptive systems modulating across actors and at scale.⁶¹

To be clear, I do not think this point contradicts anything the authors describe. As they note, what matters in understanding the role of social data in value creation “is that the technology allows for more fine-grained disaggregation of services, producing flows of data that can in turn be abstracted, rebundled, and sold” (p. 161). But I find ordinality, and more specifically, eigenvalue, to be marginal to these processes and strategies of disaggregation, even though eigenvalue is quite helpful for expressing the novel endowments of ordinalization for individuals in a datafied social structure. Given the aggregative, iterative, and multi-market settings firms wrest information value from,⁶² I will admit I lost sight of the concept when the book turned from individual endowments to firm-level price and wage discrimination, and to the disaggregation of rights and income streams.

B. The (Imperfect) Science of Social Life

Alongside their account of data value, Fourcade and Healy provide a compelling case for another key ingredient to the realization of an ordinal society: the development of rationalist social science. In my view, the drift from social science (a set of methods to understand social life) to a scientific social life (the background institutional cultures and ideologies that shape how life is lived) is significant. The first offers a way of understanding the world. The latter produces technologies of prediction and economies of prediction value.⁶³ As the authors note, “[t]he kind of learning that takes place when the aim is to accurately predict is different from the kind of learning focused on the real structure of social processes” (p. 97). The slippage between understanding a causal account and enacting a predictive outcome can offend moral sensibility—for instance, the latter can result in the allocation of credit in ways that follow historic and highly racialized patterns of dispossession (pp. 95–96).

The toolkits of social science were picked up by institutional actors who had a whole lot of data sitting around and needed a way to make money out of it. As Fourcade and Healy suggest, there is a long history of rationalizing human activity in service of industry. For example, Frederick Taylor’s famous work studies used a stopwatch to time workers’ tasks. Frank and Lillian Galbraith used film cameras “to observe and record the work process in order to understand and reorganize its component parts” (p. 73). Tellingly, at the Ford Motor Company, “the name for the division responsible for internal security and the surveillance of workers was the Sociological Department” (p. 73).

In contemporary parallels, economists and other social scientists play a vital role in constructing the ordinal systems of Silicon Valley. These social scientists, like the Galbraiths before them, help companies adapt the academic methodologies designed for apprehending the world into the commercial enterprise of acting on it.⁶⁴ While the Galbraiths were engaged in detailed sociological study of the factory line, companies rely on economists and computer scientists to help them apply a different social scientific discipline to the digital platform: mechanism design. ⁶⁵

Though the authors do not pick it out, the economic discipline and method of mechanism design has played a particularly significant role in helping firms develop their ordinal systems. It is an important intellectual technology behind ordinalization, providing the science of optimization—how to engineer and coordinate individual choices in light of a set of determinate goals or rules.⁶⁶ Algorithmic or automated mechanism design forms the backbone of ordinal systems. Indeed, Google’s chief economist, Hal Varian, has said that “[o]nline advertising serves as a poster child for algorithmic mechanism design.”⁶⁷

Distinct normative concerns arise when methods of study become methods of intervention. Using social scientific evidence to reconstruct an accurate picture of injustice in society can act as a camera on the world—providing evidence of injustice. For instance, in a racial society, race is inherently predictive of outcomes. If race was not predictive of outcomes, then race would not be a social category that carries the moral and political content that it does. Thus, determining how the causes and effects of race’s predictive effect impact outcomes is important for detecting injustice inscribed along racial lines. However, transferring those same methods to now engage in predictive allocation swaps evidence of injustice for instances of injustice. By using these techniques to act in the world, ordinal systems can recreate and strengthen the ligaments of classified hierarchy.

Social scientific methods play a key role in managing the scale and complexity of online life for both individuals and companies. Identity and status are now forged via engagement, views, follows, and likes. Platforms develop strategies, premised on iterative data mining and matching via intricate mechanisms, for relevance and domination. These are intricate bureaucratic practices. It can become tempting to misunderstand how these sophisticated, responsive, complex systems exert control.

To their credit, Fourcade and Healy do not. The authors helpfully extend influential critical theorist Gilles Deleuze’s work on social control to describe the relationship between the self, control, and ordinality. The infrastructures of ordinal life express a “distinctive conception of power” (p. 26). Deleuze distinguishes enclosure spaces (like prisons or factories) from “spaces of control” (p. 26). The latter are not concentrated physical spaces but are distributed and connected via technical gateways and standards. Movement through gateways requires authentication, but it happens smoothly (pp. 26–27). This kind of power, Deleuze suggests (and the authors agree), is not about brute control. Instead, power resides in the capacity to modulate action “at a finely detailed level through continuous adaptation and feedback loops. That is to say, control is accomplished cybernetically rather than mechanically.”⁶⁸

Digital scholars broadly share this view of digital systems as cybernetic, exercising power through classification and modulation over time.⁶⁹ But other work often overstates the force this kind of power has on individual action.⁷⁰ Or rather, the understandable impulse to account for the magnitude of cumulative power achieved via feedback loops and scale can sometimes mistakenly come across as an argument for the coercive strength of the form itself. For example, a factory employee clocking into work experiences a stronger direct force of control than a user signing onto a platform. But one factory—even one company that owns many factories—cannot house 3.07 billion workers.⁷¹

The authors’ emphasis on the distinctive conception of cybernetic power is particularly effective because it highlights the enduring social truth of messiness and surprise. As they note, “social life tends to overflow the organizational and institutional matrix imposed on it, even when those institutions provide a powerful basis for coordination and control” (p. 260). In practice, ordinalization is “patchy” and often “broken” (p. 261). It also develops in response to user demand.

Consider two examples. First, in 2009, Facebook introduced the “Like” button.⁷² The company aimed to “simplify the comments feedback process from a verbal form to something that could be counted” (p. 61). So now, instead of many users writing “congratulations!” under a user’s engagement post, they could simply press the “Like” button. This allowed the company to have greater insight into the popularity of posts while producing a new form of interpersonal interaction that introduced new social norms, expectations, and obligations around how we express care (p. 63). The succinct visibility of likes also produces a new kind of social currency, which fuels “psychological obsessions and competitive frenzy,” and an ecosystem of influencers and attention markets (p. 63). Over time, users asked for more buttons to express a greater range of emotional reactions, with each additional emoji reaction “provid[ing] an additional bit of information about the user and the post” (p. 62).

Second, in 2015, “in response to audience demands for representation and inclusiveness, the Unicode consortium, which manages emojis, introduced skin-tone modifiers” (p. 62). These were soon incorporated into the operating systems of iPhone and Android, along with apps like Slack and Discord. As the authors note, this kind of digital representation was a “step forward for diversity, to be sure, but also a marvelously compact means of unobtrusively generating data on the race or ethnicity of users who choose to react to things” (p. 62).

These examples illustrate a larger point. By changing sociality’s social form or language to be more measurable, the underlying behaviors by which people engage with and make sense of one another also change. These examples also provide a more precise account of how subtly corporations exercise ordinality’s power. Datafication does not impose itself onto human behavior bluntly or violently. Ordinalization’s coercive effect is not like a virus invading the cell of self-hood and scrambling the DNA of identity. Facebook did not reprogram users in some dystopian fashion, “force” them to use emojis, or “colonize” their social activities.⁷³

Instead, as these examples show, ordinality becomes encompassing over time. It becomes more total and more compelling in dynamic response to users. As sociality spills over the bounds of evaluative ranking and the evaluations adjust to account, ordinality comes to alter the grammar for how social interactions occur. By making social interactions more conducive to measurement, ordinal systems reconfigure the etiquette of socializing. To be sure, changes suit the institution’s purposes and aims, but they also occur in response to users.

III. Ordinalization, Legibility, and the State

In Seeing Like a State, anthropologist James Scott details a now-famous account of the German state’s attempt to “remake” its forests to increase timber production.⁷⁴ Germany began by precisely detailing and mapping its forests, making this resource legible to the state. Thus equipped, state actors embarked on an ambitious plan to rationalize German forest management and maximize timber production.⁷⁵ Such rationalized forestry involved expunging all the supposedly unnecessary mess of forest flora and fauna. At first, the forests flourished, producing incredible amounts of timber. But then, the forests became susceptible to disease and fire.⁷⁶ The complexity and messiness of the forest ecosystem, far from being extraneous or unnecessary, was in fact vital to its long-term health. This episode offers a cautionary tale against the hubris of rationalization. It reminds us that we should think twice about projects of social engineering that aim to separate the measurable and useful wheat from the allegedly extraneous chaff.

Scott’s work is often referenced by digital technology scholars, including Fourcade and Healy.⁷⁷ Too often, digital technologies—and those who make, fund, and legitimate them—exhibit the same tendency for hubristic, oversimplified remaking. The power endowed by legibility unleashes competitive imperatives to remake the world along lines that may be more optimized (over the short term at least) but that excise important features of life in the process.⁷⁸ As German forests were once optimized for timber production, social life is now reconstituted into a digital form to yield more rationalized ranking assessments. Scott’s example of the German forests is not merely descriptive. The example invoked not just rationalization, but also its devastating consequences.

Ordinalization is not just a camera that captures existing hierarchies, such as who is the most popular, who needs a ride the most, who should get this job, or who qualifies for this benefit. Ordinalization is an engine that drives forward the ceaseless social ordering process—who becomes the most popular, who gets the ride, who lands the job, or who qualifies for the benefit. Ordinalization enacts a claim regarding who deserves the fame, the ride, the job, or the benefit. In other words, ordinalization works to reconfigure not just how we allocate wins and losses, but also what it means to be a winner or a loser along any number of social criteria.

The intellectual instinct here—not just of Fourcade and Healy, but the field writ large⁷⁹—is that what we need to pay attention to is not just how these systems rematerialize “real life” into “digital life.” We must attend to how this intensifies or alters the rules of the game. Perhaps even more importantly, we must chart how the digital transformation concentrates power in the small set of hands that rest on the levers of algorithmic management.

But I am left basically where I am after reading many other similarly insightful analyses of digital life. In Data and Good?, Daniel Susser notes the tendency of digital surveillance scholarship to detail and describe the effects of computerization, often in critical tones, but not to advance direct normative accounts of such critique—and far less, to ground the critiques in alternative positive accounts. Susser calls on surveillance scholars to engage more directly with the normativity often implied, but rarely explicitly laid out and defended, in their work.⁸⁰

The greatest difficulty I had with Fourcade and Healy’s account is that it shares in this tendency. It implies a normative critique but does not lay out or defend it, nor does it suggest a cure for what its diagnosis entails. While I can hazard a guess or two based on a close reading of the text, I am not sure why, or under what conditions the authors think ordinalization is bad and, by implication, how we are to intervene to make it all less bad. The implied normativity of The Ordinal Society is common in social scientific work.⁸¹ And, Susser calls out, the same tendency prevails among surveillance scholars, too.⁸² But for legal readers—who, for better or worse, are in a field concerned with what is being done and how to do it differently—this common omission is especially noteworthy.⁸³

A. Ordinalization and the Ground Truth of Social Structure

Complexity poses one barrier to direct normative engagement. The relationship between the social categories that commonly carry the most normative significance and the classification and ranking of digital social ordering requires some explanation. In fact, clarifying this relationship requires one to already take a stance on how the descriptive and normative work of social categories relate to one another. To their credit, Fourcade and Healy’s descriptive account clearly links the value-producing work of rationalized stratification for markets and states with the hierarchy-entrenching processes for groups.

As Fourcade and Healy note, classification relies on behavioral and relational characteristics that are post-demographic: People are grouped based on behaviors, propensities, and relationships—not explicit or direct measures of sex, gender, ethnicity, race, or socioeconomic status. For instance, an ad-targeting system does not directly “apprehend” that I am a woman. Nor is it built to do that. Rather, it is designed to “apprehend” my behavior by analyzing the kind of content I like and engage with and the types of users I follow. From this, it aims to “predict” what kind of customer I am to match me with sellers that want to reach that kind of customer.⁸⁴ The “kind of customer” that emerges will be defined by a cluster of behavioral attributes (e.g., “buys cat food,” “lives in Ann Arbor,” “reads e-books”) rather than demographic attributes.

So, on first blush, rationalized stratification, insofar as it produces hierarchies, does so based on observed behavioral attributes and not based on legally salient categories like race, sex, or class. However, “[b]y digging deeper into individual-level measures, by standardizing decision-making through increasingly detailed behavioral benchmarks, data collection ends up reconstituting, one tiny piece at a time, the ground truth of social structure” (p. 241). As I have noted elsewhere, the datafication of social life means that social relations have been rematerialized as data relations.⁸⁵

One result is that the post-demographic behavioral and relational characteristics ordinal systems are interested in end up reconstituting—one tiny piece at a time—the material and social content of demographic categories. In the latter half of The Ordinal Society, Fourcade and Healy detail the deep interlinkages between the relational and behavioral inferences that form the basis of “postdemographic” categories and the reasons we care about group-based inequality (p. 104). These post-demographic categories recreate the very kinds of inequality that animate social, moral, and legal projects that center demographic categories.

What are social categories like race and gender, this line of thinking might provocatively ask, if not a durable heuristic for a bundle of socially predictive behavioral and relational observations? As the authors put it, the persistent statistical “residual” gap between groups “is in fact a measure of advantages or disadvantages that have been produced and reproduced over the very longue durée” (p. 245).

Fourcade and Healy do not overtly defend this account of social categories. But in detailing ordinalization’s processes and effects, they lay out a powerful descriptive case for why we care about demographic categories to begin with. Namely, demographic categories are of moral and legal concern insofar as they durably sort certain populations toward the bottom across a variety of life outcomes. In an ordinal society, the social salience of these categories—that they describe the social logic of marginalization and impoverishment—remains, even as social ordering methods marshal other indicia through rationalized stratification (often, it should be said, unintentionally).

This has profound implications for why datafying social life may both reveal and perpetuate injustice. It also suggests that datafication’s accompanying allocation of rights and obligations has significant stakes. These legal rights determine who has claim to the engines that rematerialize and remake social relations; thinking carefully about their reform matters a great deal for achieving some measure of justice in a digital society. It’s possible that life in an ordinal society is unbearable because the methods of social sorting and social order remove the legal wrenches we have wedged between inequality’s grinding gears: protections against discrimination in housing, jobs, and educational opportunity allocation based on formal demographic categories.⁸⁶

Another related possibility—that Fourcade and Healy, to their credit, engage—is that ordinalization recodes structural barriers as individual behaviors (p. 244). This implies a whole host of distracting diagnoses and interventions that, in trying to fix structural issues with programs to educate better behavior, miss the mark. Relatedly, Fourcade and Healy, like others, emphasize that ordinalization turbocharges the problem of value.⁸⁷ Ordinalized settings are powerful mechanisms for rewarding good behavior. But “from the point of view of any particular market, what counts as good behavior is . . . defined by what makes for a profitable sale” (p. 248).

B. The Politics of Legibility: From High Modernism to Informational Capitalism

Perhaps a greater barrier to direct normative evaluation than complexity is genuine, if sublimated, political disagreement about what ails an ordinal society. Let’s return to James Scott’s Seeing Like a State. There are important differences between the coherent and ambitious (if ultimately misguided) vision of social improvement Scott details and the engine of world-building that produces the ordinal society.

One is planning. The parable of German forestry features coordination and telos.⁸⁸ Whereas, no single company or institution planned the ordinal society—it is the result of generalized market imperatives and historical and technological confluences. The engine of ordinalization churns forward through practices that harden market advantage into market power and consolidation.

In the give-and-take between structural forces and individual agency, these two visions begin at different poles. The engineers of German forests began with a social objective—improve the total timber production—and then enacted individual transformations on trees and forests to achieve that goal.⁸⁹ The institutional ordinalization project, by contrast, proceeded from the specific and individual. Profit-seeking firms adopted the kind of strategies that they have been socialized to associate with success. Individuals, out of curiosity and vanity, quantified themselves to affirm their social status through likes and views. From the bubbling up of these piecemeal incentives came first market rewards, then pressures, and then imperatives. Individuals first played with, then engaged in, and finally became subjected to ordinalization.

Consider the example of law school ranking by U.S. News and World Report (p. 264). It was not the collective ambition of law schools to centralize and flatten the meaning of success into one set of generalized metrics. Law schools bemoan the loss of power that legibility implies: Ranking metrics not only eliminate the nuance of what a good education or good legal career may entail but also influence how schools decide to grow and change.⁹⁰ This situation emerged from an iterative process of ranking, incentive alignment, chafing, competition, and negotiation. U.S. News, like the far more sophisticated digital platform ordinalizers, made a market their product. Its value derives from the market’s capacity to activate competitive pressures between law schools and capture the results. I doubt U.S. News had grand social ambitions to remake the meaning of quality and prestige in legal education. But in carving value from the competition among schools and setting the terms of success, this is, in effect, what they have done.

Given this rather distinct starting position from that of ambitious, high-modern social engineers, I have always found the temptation to compare our present to the past described in Seeing Like a State a bit odd.⁹¹ High modernism and informational capitalism are rather different, though they admittedly share a common central belief in the power of rationalized quantification. Whether it is timber production, the quality of a legal education, or love, faith in quantification means that if we can render “every relevant detail” into analyzable form, we can fully grasp our object of study (p. 73). Once grasped, we can then unlock the pathway to answers or success.

But a common faith in the governing capacity of numbers masks the rather different orientations stakeholders have to the political terrain so governed. The hubris of the high modernist social planner is an overconfidence that we can improve our world in conscious and planned ways. The planner is seduced by quantification, but its coercive effects are outsourced—to German forests or minority communities. This is different from the hubris of the individual meritocrat or market actor, which smacks more of affirmation than action. As Fourcade and Healy note about law school rankings, “[w]hat initially felt right to law school administrators about the rankings was that they expressed their sense of status in a visible and apparently objective way” (p. 265). The comparative ranking lure normalized a sense of status derived from competition between schools—and pushed schools to “compete on the dimensions the ranking method values” (p. 265). In this account, the ordinal subject experiences both the seduction and the coercion of quantification.

These distinctions matter for the politics they imply—it is worth making our critiques of the datafication of social life careful and precise. The Ordinal Society joins a sizeable body of work cataloging the ways that digital society can be stupid, stultifying, and perhaps even evil. And to be upfront about my own views, I think that along many dimensions the society emerging from the algorithmic engines of digital life is stupid,⁹² stultifying,⁹³ and even evil.⁹⁴ However, in developing a case for what has become unbearable about digital society, my own political commitments make it imperative not to have such critique further undercut faith in our collective capacity to apprehend the world and to intervene for the better.

For one, it is only through our collective institutional capacities that we can hope to alleviate what ails digital life. Digital companies are the largest and most powerful companies in the world. It is only through some measure of state regulatory action that forms of value, distinguishable from individual market value for private firms, can be expressed and enacted at scale.⁹⁵ Imperfect (and imperiled) as it is, the administrative state is perhaps the only institution big enough to serve as a counterweight to digital capital. As Thea Riofrancos put it in the context of achieving decarbonization, if we hope to exert some measure of control and constraint on the power of large companies and develop alternatives to the world they offer, we need a state both for its disciplinary power and for its infrastructural capacities.⁹⁶

Critiques of digital life that gloss over the distinctions between civic action and private gain may muddy the waters in other ways, too. While state and market ordinalizers may share methods of legibility, reformation, and control, firms and agencies are distinct institutions with different mandates, functions, constraints, and cultures. In extending the critiques of legibility from the civic social planner to the ordinal mechanism designer, we must take care against slippage; lest the critique of the hierarchy digitalization get mistaken for—or worse, harden into—the belief that what can be gained through progress is never worth what is lost in the process. As Paul Seabright wrote of Seeing Like a State, “[t]here is no feebler pretext for conservatism than the anxiety that progress is somehow inimical to charm.”⁹⁷ There is no doubt that online love has destroyed much of pre-online dating’s charm. Serendipity and chance may be central to the process of spontaneous love, and itself a reason to prefer what came before. But too often, this impulse extends a similar diagnosis to prior forms of social order that only take on the glow of charm with the benefit of hindsight. In their own time, many were themselves rightly judged to be in dire need of reform. To borrow from Riofrancos again, digital life critiques must always proceed with two parallel steps. If with one foot we critique the offered future’s stupidities and evils, with the other we must insist that a better vision of progress is possible.⁹⁸

Making such critiques clear seems especially important now, against the backdrop of the significant threat facing the U.S. administrative state’s future.⁹⁹ The more the administrative state—which does the quiet, mundane work of saving millions of lives through air quality science, workplace safety, welfare administration, and food and drug safety—is under attack, the more we are left with a state that is only punitive and carceral and a market that provides us with a thin and fleeting trace of sociality for a price. If anything, our problem now might well be a lack of ambition and imagination on what we can do, positively and progressively in our collective capacity as citizenry, that cedes the terrain of grand ambition to the titans of tech.

* Assistant Professor of Law, University of Michigan Law School. My sincerest thanks to Don Herzog and Ben Green for their thoughtful comments. I am additionally grateful to Grace Vedock, Elena Schultz, and the rest of the editorial staff at the Michigan Law Review for their helpful edits, sage advice, and patience.