The last lesson (12.3.1) showed that digital life turns behaviour into data — datafication . This lesson follows that data to its sharpest edge and asks the questions that decide whether the digital age is liberating or something darker: who owns your recorded life? who profits from it? and what happens when automated systems begin sorting human beings at scale? It is the cui bono? — who benefits? — question (the reflex this whole course has been building) turned on the datafied society.
And it is where digital sociology stops being merely interesting and becomes urgent, because the stakes are your autonomy, your opportunities, and the shape of power in the century you are living in. But urgency is exactly where analysis is most likely to slide into either panic or complacency — so this lesson holds the same discipline as the rest of the course: take the real dangers seriously, measure them rather than merely feeling them, mark honestly where the evidence is strong and where the confident stories run ahead of it, and refuse both the techno-utopian shrug and the dystopian scream. The truth about the datafied society is alarming enough without exaggeration, and clear seeing serves you better than either fear or denial.
The exhaust that became the product.
In the early days of the commercial internet, companies collected data about their users mostly to improve the service — to fix what was broken, to make the product better for you. The data was a byproduct, a kind of exhaust.
Then someone realised the exhaust was worth more than the product. The data trail you left — what you clicked, lingered on, searched, ignored, where you went, who you knew — turned out to be enormously valuable, not for improving your experience, but for predicting and influencing your behaviour , and selling those predictions to whoever wanted to shape what you do: advertisers, above all, but not only them. The business model flipped. The service became the bait; your behaviour became the raw material; and predictions about your future actions became the product being sold — to others, about you, usually without your real understanding.
The scholar Shoshana Zuboff named this surveillance capitalism — an economic logic that claims human experience as free raw material, translates it into behavioural data, and trades on predictions of what people will do. Her sharpest point: you are not the customer, and you are not even really the product. You are the source — the free supplier of the raw material — while the actual product (predictions about you) is sold to others, in markets you never see, about a future you have not yet chosen. The "free" service was never free; you paid with the most intimate currency there is — the record of your life — and you paid it to firms that turned it into a means of anticipating and nudging your behaviour at a scale no institution in history has ever commanded.
The plain-words core: three hard questions
Zuboff's framework, and the wider field, can be organised around three plain questions. Take them in order, because each sharpens the last.
Who owns the data? Not you. The record of your behaviour — generated by your life — is largely owned and controlled by the platforms that capture it. You supply the raw material continuously and for free; they own it, combine it, analyse it, and monetise it. This is a genuinely new kind of property relation: a vast asset, made of human lives, owned by a small number of firms. (Note the echo of Marx, 4-territory: a resource produced by the many, owned by the few — here the resource is behaviour itself .)
What is done with it? It is turned into predictions — and increasingly into influence . The mildest use is targeting (showing you an ad for a thing you might buy). The sharper concern is behaviour modification at scale : designing the platform to nudge, hook, and shape what you do — the infinite scroll, the notification, the variable reward engineered to hold your attention (drawing, deliberately, on the psychology of addiction). Whether this rises to the near-total behavioural control Zuboff warns of is contested (see below), but that platforms are designed to shape behaviour, not merely record it , is not in doubt.
Who gets sorted by it? This is the edge that matters most for inequality. The predictions are used to sort people — to decide who sees which price, which job ad, which loan offer, which risk score, which stream of content. Increasingly this sorting is automated , done by algorithms at a scale and speed no human bureaucracy could match. And automated sorting of human beings is where datafication meets the whole of Part 8 — because who gets sorted up and who gets sorted down turns out to track exactly the old lines of inequality, now automated and hidden behind a veneer of mathematical objectivity.
Going deeper: algorithmic bias and the automation of inequality
The most consequential thing about the datafied society may be this: decisions that used to be made by humans — who to hire, who to lend to, who to police, who to release on bail — are increasingly made or guided by algorithms, and those algorithms can automate and launder the very inequalities of Part 8.
How the math launders the bias — the mechanism, step by step.
The training data carries the past's injustice. An algorithm learns by finding patterns in historical data. But history is not neutral — it is the record of a society structured by inequality (Part 8). A hiring algorithm trained on who was hired and promoted before learns the past's biases about who "succeeds"; a policing algorithm trained on where arrests happened before learns where police went before, not where crime was (10.2.1 on the dark figure). The model faithfully reproduces the pattern in its data — and the pattern is the injustice.
The feedback loop amplifies it. Predictive policing sends more police to areas the model flags; more police find more (recorded) crime there; that feeds back as data confirming the area is high-crime, sending yet more police (10.2.1's rate-producing agencies, now automated). The algorithm does not just reflect the bias; it acts on it , generating the data that appears to justify it — a self-fulfilling prophecy (7-territory) running at machine speed.
The veneer of objectivity hides it. Here is the sharpest turn. A biased human decision can be challenged — you can point to the prejudice. But when the same biased outcome comes from "the algorithm," it wears the authority of mathematics and data, and becomes far harder to contest: the computer decided, the data is objective, there's no prejudice here. The bias is not removed; it is laundered — given a false neutrality that makes it more powerful and less accountable than the human bias it replaced. This is the "God trick" of false neutrality (13-territory) automated and sold as fairness.
So the danger is not that algorithms are more biased than humans — often they are not. The danger is that they scale the bias (one biased model decides millions of cases), hide it (behind mathematical authority), and entrench it (through feedback loops that manufacture their own justification). Automated decision-making can take the inequalities of Part 8 and make them faster, larger, and much harder to see or challenge.
The digital divide: inequality of the digital itself
Alongside the inequality that algorithms produce , there is inequality in access to the digital — and the sociological understanding of it has deepened well past the early, naïve version.
The digital divide has three levels, not one.
First level — access. The earliest concern: who has a device and a connection and who does not. Real, and still real globally — but in wealthier societies, access has spread enough that people wrongly concluded the divide was closing. It was not; it was moving.
Second level — skills and use. Among those with access, huge differences remain in the skills to use it well and the ways it is used. The same connection can be a tool for education, opportunity, and advancement — or mainly for entertainment and consumption — and which it becomes tracks class and education closely. The advantaged use the digital to get further ahead; the disadvantaged often use it in ways that do not build advantage. Same access, divergent outcomes.
Third level — outcomes and exposure. The deepest level: who benefits from digital participation and who is harmed by it. The advantaged extract opportunity, information, and connection; the disadvantaged are disproportionately exposed to the harms — surveillance, harassment, exploitation, algorithmic sorting downward , the predatory targeting of the vulnerable. The digital does not level the playing field; it re-inscribes the old field with new tools — and at the third level it can actively widen the gap, giving the advantaged another engine of advantage while exposing the disadvantaged to another source of harm.
The early dream — that the internet would be the great equaliser, dissolving hierarchy in a free and open commons — was, on the evidence, wrong (12.3.1). Inequality did not dissolve in digital space; it adapted to it, and in some ways sharpened.
Hold the alarm to the evidence — where the strong claims are strong, and where they run ahead.
This is a domain where the stakes invite overstatement, and honesty (the discipline of Part 7, and of 11.3.2 on measured fears) requires marking the line carefully.
What the evidence strongly supports : that behavioural data is captured at vast scale and largely owned by a few firms; that platforms are designed to shape behaviour, not merely serve; that automated decision systems can reproduce, scale, and launder existing biases; and that digital inequality is real, multi-level, and often widening. These are well-established, and none should be waved away.
Where the confident stories run ahead of the evidence : the strongest versions of the "total control" thesis — that surveillance capitalism has achieved, or is close to, near-complete power to predict and dictate human behaviour — go beyond what is demonstrated. People are not infinitely manipulable; the prediction products are often far cruder than either their sellers claim or their critics fear; targeting frequently fails; and human beings resist, subvert, and ignore the nudges (the tactics of de Certeau, 12.1.3, operate online too). Zuboff's framework is illuminating and influential, and it is also contested — critics argue it overstates the coherence and success of the manipulation, understates human agency, and treats a still-uneven set of practices as a finished, totalising system. Take the mechanism seriously; hold the most sweeping conclusions provisionally.
The honest posture is the one this course has held throughout: the datafied society concentrates a genuinely new and dangerous form of power in few hands, and automates inequality in ways that are hard to see and challenge — and the case for alarm does not require believing the maximal claims about total control. Name the real, evidenced dangers precisely; resist the temptation to inflate them into an all-powerful system that erases the very human agency that could resist it. Overstatement is not just inaccurate; it is demobilising — a machine described as omnipotent is one no one bothers to fight.
One loan application, the whole lesson.
You apply for a loan online. Watch every thread of this Part run through that single act.
Datafication : the application, and thousands of data points about you (bought, inferred, scraped), are fed to a model.Surveillance capitalism : much of that data about you was generated by your life and is owned by others, assembled and sold in markets you never saw.Algorithmic sorting : an algorithm, not a loan officer, scores you and sets your rate — or rejects you.Laundered bias : the model was trained on historical lending, which was shaped by the discrimination of Part 8; it may deny you, or charge you more, for reasons that trace to your postcode, your name, or your network — proxies for race and class — while presenting the decision as neutral mathematics you cannot argue with.The digital divide (third level) : if you are already disadvantaged, you are more likely to be sorted downward , exposed to worse terms or predatory targeting; if advantaged, sorted upward , offered better. The system re-inscribes the old hierarchy.The reflexive check : and the maximal-alarm reading (the algorithm is an all-seeing oracle of your creditworthiness) is itself to be resisted — these models are often crude and error-prone, which is its own injustice (being wrongly sorted by a bad model), different from the injustice of being accurately sorted into a disadvantaged group. One ordinary loan application contains surveillance capitalism, algorithmic bias, digital inequality, and the need for measured judgement all at once — which is exactly why the datafied society is not a niche topic but the water more and more of your life now swims in. The cui bono? question — who owns this data, who profits, who gets sorted which way — is no longer abstract. It is your loan, your rate, your rejection.
Why this matters
Because the datafied society is being built around you now , mostly without your consent and largely without your awareness, and the defaults are being set by those who benefit from them. The single most important thing this lesson can give you is the reflex to ask, of any digital system you touch, the cui bono? question: who owns the data this generates, who profits from it, and who gets sorted — up or down — by it? That question cuts through the friendly interface and the language of "free" services and "personalisation" to the power relation underneath: a new form of property made of human lives, a new form of influence engineered into design, and a new form of sorting that can automate and hide the oldest inequalities we have.
And the second most important thing is the discipline with which to hold that alarm — because this is a domain where fear and denial are both being sold to you, and both disarm you. The techno-utopian says relax, it is all convenience and connection, the market will sort it out — and thereby talks you out of noticing the power being assembled. The techno-dystopian says it is already too late, the system is total, resistance is futile — and thereby talks you out of the agency you still have. The measured truth is more useful than either: real and serious dangers, unevenly realised, still contested in their reach, and still — crucially — open to being seen, regulated, resisted, and built differently. The datafied society is not a finished machine; it is an arrangement, made by choices, and arrangements can be changed (the lesson of the whole course). Seeing it clearly — neither shrugging nor screaming — is the first condition of changing it.
This closes Topic 12.3, and with the next lesson (12.4.1) we take stock of the whole Part — what a specialised and applied sociology is finally for, and how the many rooms of the house and the exits into the world fit into one understanding of the discipline as both a way of seeing and a way of acting.
Surveillance capitalism — Zuboff's term for an economic logic that claims human experience as free raw material for behavioural data, from which prediction products are made and sold; a contested but influential framework.
Behavioural data / behavioural surplus — the record of human behaviour captured by platforms, beyond what is needed to provide the service, used to predict and influence future behaviour.
Algorithmic bias — the reproduction of historical injustice by automated systems trained on biased data, often amplified by feedback loops and hidden behind a false appearance of objectivity.
Automated decision-making — the use of algorithms to make or guide consequential decisions about people (hiring, lending, policing, bail), sorting human beings at scale and speed.
Digital divide (three levels) — inequality in access to digital technology, in the skills and uses among those with access, and in the outcomes and exposure (benefit vs harm) that result.
Laundering of bias — the process by which prejudice, when enacted through an algorithm, acquires the authority of mathematics and data, becoming harder to detect and contest than overt human bias.
Try it yourself:
One — Run cui bono? on a "free" service. Pick a free app or platform you use. Ask: how does it actually make money? What data does your use generate, who owns it, and who is the real customer being sold something (a prediction about you)? "If you're not paying, you're the product" is a start — but push further: you are usually the source , and the product sold to others is a forecast of your behaviour.
Two — Find the laundered decision. Think of a consequential decision now made or guided by an algorithm (a credit score, a job-application filter, a content feed, an insurance rate). Ask: what historical data was it likely trained on, and what inequalities might be baked into that history? Then notice how much harder it is to challenge "the algorithm decided" than "a person decided" — that added difficulty is the laundering at work.
Three — Practise measured alarm. Take one worrying claim about the digital future (algorithms control us, privacy is dead, AI will decide everything). Sort it: which part is strongly evidenced, and which part runs ahead into total-control territory? Hold the evidenced danger firmly and the sweeping claim loosely. Notice that this makes you more able to act, not less — because a danger you can see clearly is one you can actually do something about.
What's next: Topic 12.4 — the single lesson 12.4.1, The Uses of a Specialised Sociology — takes stock of the whole Part: how the many specialised rooms of the house (12.1), the exits into applied and public practice (12.2), and the digital frontier remaking all of it (12.3) fit together into one picture of sociology as both a way of seeing anything and a way of acting in the world.