278 | Kieran Healy on the Technology of Ranking People
What this covers
Sean Carroll speaks with sociologist Kieran Healy about how modern technology has made the measurement, classification, and ranking of people pervasive and granular. The conversation traces how data-driven rankings now shape life chances across healthcare, law, education, and hiring—yet are experienced by those ranked as deserved personal outcomes rather than imposed structure. Healy argues that behavioral data, from fitness trackers to financial records, gains legitimacy by appearing objective, which shifts blame for poor classifications onto individuals themselves while allowing the advantaged to interpret their inherited advantages as personal virtues. The discussion ranges across how financialization pairs with data collection, how systems of ranking persist because they are powerful and convenient even as people resent being subject to them, and why opting out carries real costs.
The episode explores how ranking and classification are ancient human practices, but what has changed is their scale, scope, and reach into forms of social life previously impossible to measure in real time. Healy examines the credit score as an institutionalized ordinal measure, online gaming's ELO ranking systems, and how companies like John Deere and General Motors increasingly profit from the data their products generate rather than the products themselves. The conversation turns to how price discrimination enabled by granular data has eroded the satisfaction of getting a good deal, and how the internet's transformation from a libertarian vision of digital homesteading into a centralized, hierarchical system was largely demand-driven rather than simply imposed. Healy argues that framing data concerns as individual privacy choices misses the institutional phenomenon, and that while these systems are expansive, they are not totalizing—people and social life tend to overflow and resist their boundaries.
Healy argues that modern technology has made the ranking and classification of people pervasive and fine-grained, producing an 'ordinal society' in which one's data-driven classifications shape life chances while being experienced as deserved personal outcomes rather than imposed social structure.
- Classification and ranking are endemic to human societies but technology has transformed their scale, scope, and granularity
- Behavioral data lends classifications legitimacy and reframes structural constraints as personal choices for which individuals are blamed or credited
- Incorporation into these systems is largely demand-driven (convenience) and opting out carries real costs
Classification systems govern life chances across healthcare, education, hiring, and law by treating structural constraints as individual behavioral choices.
- Before things can be ranked they must first be named/classified, but there is a very strong tendency for nominal (unordered) classifications to turn into rankings and to position people within them.
“before you can rank things, you must name them first... there's always a very strong tendency then for nominal classifications, which should be unordered or which are often unordered to turn into rankings”
- Your data-driven classification affects your life chances—the opportunities offered and their cost across healthcare, law, education, and hiring—and because the cultural logic treats these as reflections of your own behavior, structural constraints (background, opportunities, others' opinions) get 'stuffed through the behavioral channel' and recorded as choices you made, so you are judged as deserving your outcomes even when you had no real option, while the advantaged experience inherited advantages as personal virtues (born on third base thinking you hit a triple).
“It affects your life chances, as sociologists would say, it affects the opportunities that are offered to you and what those things will cost”
- Behavioral data (e.g. what your watch records about your steps) is presented as more truthful than self-report, which gives classifications derived from it tremendous legitimacy and introduces the idea that your category is your own fault or virtue—so the poorly classified are blamed for their social situation.
“If I ask you how many steps you walked today, you might tell me one number... but your watch on your wrist knows. And so that behavioral data is both very valuable and it does kind of capture something that's much more accurate”
- Quantitative classification always involves imposing a framework, illustrated by the fact that American and French butchers cut beef into very different named cuts—there is heterogeneity in how the same underlying reality is carved up, even though there is still ultimately a cow under there.
“even butchers carve nature at the joints quite differently. So there's heterogeneity, even if ultimately there's still a cow under there. There are different ways to cut it up.”
Ranking has always existed, but technology makes it fine-grained, pervasive, and capable of measuring social domains previously beyond real-time capture.
- Ranking and ordering people is not new—it is an endemic feature of human social organization wherever differentiation exists—but what is new is that technology has made this both much more fine-grained and much more widespread in scope, reaching forms of social life previously beyond any real-time measurement.
“it's an endemic feature of just how human societies are organized... But what's new now is... the ability to do this has both become much more fine-grained and much more widespread”
- An 'ordinal society' is one based around and justified by the pairing of massive datasets with algorithmic/statistical/mathematical processes that take in data and output scores and rankings, making inroads into every social institution to streamline and organize them.
“fundamentally, the idea of ordinality or an ordinal society is one that's based around and justified by the idea of kind of measurement and ranking”
- Data collection is not mainly external surveillance spying on a separate real world; rather social life itself now takes place mediated through these technologies, a qualitatively different feature of the modern world where interactions, jokes, and conversations that were once impossible and pointless to capture now occur as flows of data.
“It's more that the social life itself is now kind of taking place mediated through these technologies. And that's tremendously kind of powerful and a kind of qualitatively different feature of how the world is now.”
Ranking systems are paradoxically demanded because they simplify decisions and enable organization, yet individuals resist being subject to them at critical moments.
- Ranking technologies persist and are demanded because they are extremely powerful heuristics that radically simplify decision-making and make organizing possible, even though we would prefer not to be subject to them ourselves—so much of the social struggle is over who gets to use these measures and who gets to avoid being subject to them at key life moments.
“these technologies are just tremendously powerful because they just are heuristics, they simplify decision-making immensely... people demand in that sense, the ability to make these kinds of decisions”
- Framing data concerns as 'what can I do to protect my own data / opt out' is relentlessly individualizing and loses sight of the broader institutional phenomenon; policy like GDPR cookie consent and California's 'Do Not Sell My Information' buttons push choice onto individuals, who in practice just click 'accept all', so this is the wrong terrain for debate.
“the tendency to think about how people think about these problems at all is also relentlessly individualizing... to the extent to which that becomes the terrain on which most debate about this is happening, well, then you're kind of losing sight of the broader kind of institutional phenomena”
- Refusing to feed the system has a concrete cost: because you don't give information, the system gives you back only the lowest common denominator—e.g. turning off search history means YouTube can no longer target ads and serves up tawdry, irrelevant ones—illustrating the 'mostly bargain' where you get out only what you give in.
“the price you pay for not wanting to be incorporated is that you get the worst, they're like, Okay, we'll have to serve you up the lowest common denominator”
Organizations operate under a data imperative to collect granular information on everyone, paired with business models that financialize ongoing transactions into recurring revenue streams.
- Embedding software and data collection in everything pairs with financialization: instead of a one-time purchase, companies turn each transaction into an ongoing stream of income (rent) for shareholders, and the data enables new mechanisms—e.g. patented kill switches tied to financial records to remotely disable a car when payments stop.
“what we're interested in doing, what businesses are interested in doing is turning every potential transaction into a stream of income. And that a rent as economists would call it”
- New technologies are initially given away free as a delightful 'gift', and like all gifts this sets up an expectation of return—what we return is information about ourselves (location, behavior, interactions, questions asked)—which organizations then make profitable; the same pattern is now playing out with large language models.
“as any sociologist or anthropologist will tell you, gifts set up these expectations of a return. What we give in return is information about ourselves”
- Organizations are mostly not interested in any particular individual, but they feel a 'data imperative'—a strong internal conviction that they should be collecting granular data about everyone—supported by a whole infrastructure of occupations tending 'data lakes', even though practically extracting useful per-person granularity is difficult.
“We call this kind of the data imperative. They know that this is something they should be doing. And there's a whole infrastructure, there's a whole set of occupations of people who tend data lakes”
- Companies we think of as making hardware or consumer goods often make most of their profit from their financing divisions—General Motors makes more from auto loans than autos—and even Apple, the last major Silicon Valley hardware maker, increasingly derives growth from services, subscriptions, and its credit card.
“The General Motors is mostly a bank now. They make more money out of their auto loans than they do off their autos.”
- Granular data collection lets every manufacturer become a software company and every software company a provider of software-as-a-service, so that data and the activity it generates can be sliced into tranches and sold—e.g. John Deere knowing when farmers plow and plant could become a provider of market intelligence and futures-market data, not just a tractor seller.
“this tendency to make data collection more and more granular fits really nicely with this tendency for finance to wanna make sort of products that are more and more abstracted... every manufacturer becomes a software company and every software company becomes a provider of software-as-a-service”
Digital platforms and services were not imposed but demand-driven; people chose convenience and sociability over decentralized alternatives that no longer seem plausible.
- Early computing fused Cold War defense/command-and-control origins with a flat, libertarian hacker culture, and by the 1990s web the high watermark was a romantic vision of 'digital homesteading'—a free frontier where individuals could escape local status, ranking, and the 'weary giants of flesh and steel' to simply be themselves.
“this kind of, that's the sort of high watermark in a lot of ways of excitement about the kind of pure freedom associated with just setting out on your own, setting up your website, these homestead dreams”
- The centralized, hierarchical internet was not simply imposed on unwilling users by corporations; it was largely demand-driven, because people wanted the sociability and convenience of having someone else administer servers and help them find others quickly, so it is a mistake to think 'we could have had nice things but the corporations ruined it.'
“it's tempting to think that, oh, we could have had nice things, but then the corporations came along and sort of made this world terrible for us. Now, it's not that there's nothing to that critique, but it is the case that people do want different things”
- When predicting a new technology's impact, people project their ideal-world desires onto the change; when things don't work out that way, the utopian vanguard tends toward immense disappointment verging on contempt for everyone else—'revolutionaries end up disappointed with the peasants they have liberated.'
“people project their own desires about what an ideal world would be onto whatever the sort of social change... there can be a kind of immense disappointment amongst the utopian vanguard with everyone else and their failure”
Individualized data enables near-perfect price discrimination and transforms insurance, credit, and other markets from risk-pooling into prepayment for predicted behavior.
- The credit score is the best-institutionalized example of an ordinal score: a single number characterizing your debt-paying behavior that you are held morally responsible for, whose use has blossomed far beyond its initial restricted purpose—much like the driver's license became a de facto national ID—into a gateway affecting areas it was never designed for.
“The credit score is the most obvious one where you have, in the United States... a single number that characterizes your behavior in a way that you are sort of morally responsible for”
- Insurance traditionally works by pooling people of varying risk and spreading risk across them; individualized behavioral data from connected cars enables near-perfect price discrimination, which transforms insurance into something like American dental insurance—not real insurance but prepaying for what the insurer already knows you will do.
“for insurance to work, you have a pool of people of varying degrees of risk, and then you spread that risk across individuals. But if you have individualized data on people, well, then you can kind of have a different version of an efficient market, where you can price discriminate perfectly”
- As these systems approach their efficiency limit, the result is a perfectly price-discriminated market where everyone pays exactly what they are willing to pay, which makes consumer surplus—the satisfying sense of 'I got a good deal'—evaporate, contributing to a pervasive modern sadness.
“it's a perfectly price discriminated market where everybody pays exactly what they're willing to pay, and so that means then... that sense of kind of, Hey, I got... What the economists would call consumer surplus... can evaporate”
AI systems marketed as autonomous often depend on hidden human labor, and algorithmic content may degrade public information ecosystems that trained them.
- The naive divide between 'old fuddy-duddies' and 'digital natives' is misleading: what makes the young 'native' is not deep understanding of how the technologies work but comfort swimming in the environment like fish in the sea without thinking about the water.
“what makes them kind of native is not their deep understanding of how these technologies work, but more that they're kind of comfortable swimming around in this environment like fish in the sea without thinking too much about water”
- Online game consoles already exemplify the ordinal society: subscription multiplayer games run ELO-like ranking systems that are genuinely useful to players (matching them against fair opponents) while simultaneously giving the company a global ranked view of who plays most and what kind of people they are.
“they all run some ELO-like ranking system to make sure that you're matched against people who are kind of competitive with you... but then also provide a global view of the whole system about who plays the most, what kind of people”
- These classification systems are 'democratic' in being expansive and inclusive—their logic is to incorporate everyone (you must measure and track to do so) and then re-rank and stratify them—so being outside the system is a double disadvantage because you become unclassifiable, much as relying on cash leaves you outside the banking system that incorporates even poor people while extracting fees from them.
“the idea is to incorporate and then stratify, and to incorporate you must measure and track... to be outside of that sort of system then is to be at a double disadvantage because you're not even kind of classifiable”
- AI is distinctive in that, having trained on the free gift of the entire web's content, it now emits AI-generated 'effluvia' back into that environment; meanwhile genuine social activity has retreated to private platforms (Substack, Discord, Slack) so that if only AI 'slop' remains publicly accessible, this may pose problems for the technology's future.
“Now they're in the process of kind of emitting the effluvia of AI-generated output back into that environment. And it's not clear to me what's going to happen.”
- Modernity holds the individual as an autonomous, authentic agent forging their own self, while organizations simultaneously maintain a 'data double'—a shadow of recorded digital traces that they treat as representing us truthfully—so the romantic drive to be 'really yourself' coincides with the institutional need to 'authenticate' that you are who you claim to be.
“this sort of set of digital traces that produce this kind of shadow of ourselves, a data double of ourselves, if you like. That's a phrase that Dan Book, as a historian coined”
- Systems marketed as AI are often actually low-paid human labor behind the scenes, as with Amazon's 'Just Walk Out' stores that were hyped as AI but turned out to rely on overseas workers tagging purchases—and similar arrangements may underlie 'smart' fridge monitoring.
“As we saw recently with Amazon, right? That with their Just Walk Out stores... initially it was all this hype about it being AI, but then it just turned out to be a bunch of people. Poorly paid overseas who were kind of tagging everybody”
- Carroll and Healy belong to an intermediate generation 'cursed to explain computers both to people older than us and people younger than us.'
“we're in that sort of, I think, just about in that intermediate generation that is cursed to explain computers both to people older than us and people younger than us”
- When IBM France needed a name for the computer in the 1950s, instead of the direct translation 'calculator' they consulted Sorbonne Latin professor Jacques Perret, who proposed 'ordinateur', rooted in 'ordination'—the theological sense of a God who brings order to the world—capturing computation as both ranking and bringing order.
“They consulted with this guy, Jacques Perret, who was a professor of Latin at the Sorbonne. He objected and he said, what about ordinateur?... its ultimate root is to do with ordination... like a religious sense of a God who brings order to the world”
Despite their pervasiveness, these systems are not totalizing; people overflow their boundaries, institutions remain poorly connected, and new arrangements will eventually emerge.
- Although these systems are pervasive and expanding in scope, they are not totalizing—people tend to overflow the boundaries of systems built to enclose them, social life is messy, things break and happen at random, and these systems do not exist forever, so we are never permanently stuck in a particular way of organizing things even though what comes next is not necessarily better.
“People tend to overflow the boundaries of the systems built to enclose them”
- A real form of freedom—'interstitial liberty'—arises from the institutions organizing our lives being poorly connected and unable to transfer information efficiently, which created the freedom to move somewhere and start with a clean slate without being easily found; as systems integrate and friction disappears, this freedom that bubbled up from the cracks between organizations is increasingly lost.
“freedom in the book we call this interstitial liberty, that there's a kind of freedom that comes from the institutions that organize our lives being so relatively poorly connected to one another”