Philosophy of Power

An algorithm inside American hospitals decides who gets flagged for extra care, and it never sees race. Correct one bias in it, and the share of Black patients flagged rises from eighteen percent to forty-six. No meeting decided that. Somebody chose what the machine would predict, and chose cost. That choice has a date and a desk.

The desk is real. A person, a date, a room where somebody settled on cost as the stand-in for sickness. Then that person moves on, and the rule they left keeps sorting patients with nobody deciding again.

It is eleven forty at night, and you answer the message nobody required you to answer. Tomorrow you will take the number you did not want, in a voice that comes out agreeable. You are also the person at that table who can name them. The three funds that cast the votes at the annual meetings. The donors on the committee chair's disclosure, and the bill that quietly died in mark-up in the spring. Now hold that list against the eleven forty message and find one line of it that explains the message. Nobody on the list called you. For years you have drawn one conclusion from that gap, which is that you have not yet found the real name.

Before anything else, the suspicion deserves its strongest form, because the numbers are on its side. When the polled preferences of ordinary citizens are set against the preferences of the affluent, it is the affluent that the adopted policy tracks.¹ When access to a congressional office is tested directly, the word donor opens the senior door. We will take both of those apart properly later.

So the asymmetry is measured, and the room is real, and access to it is being sold. Which makes the next question the obvious one. Who decided. That question has a serious answer, and the serious answer came from a man who was sick of lists exactly like yours. The way American towns were studied in the 1950s was to ask well-placed informants who really ran the place, and write the names down. Robert Dahl's objection was not that the names were wrong. It was that a reputation is not a finding, and a claim nobody could disprove is not knowledge.

His fix was a measurement rule, and it is still the cleanest one anybody has. Pick a town. Inside it, mark the areas where something is genuinely at stake, such as redevelopment, the schools, the party nominations. In each area, find concrete decisions where two sides wanted different things. Then see whose want became the outcome. If one set of people prevails across all the areas, you have found a ruling elite, and if nobody does, you have not. He ran it on New Haven, Connecticut, and published the result in 1961. Different winners in different areas. No single group across the board. The big urban redevelopment programme belonged to the new mayor, Richard Lee, who initiated it and dragged local businessmen and federal officials along behind him.

Then the sociologist G. William Domhoff went into the archives, and this is his case. Redevelopment of the Oak Street area had been proposed in 1943, by a Yale planner. He wrote that the area should be allocated to future development of an institutional nature. Another architect repeated the idea in 1944. In October 1951, Redevelopment Agency staff sat down with Yale officials about it. A Chamber of Commerce minute of the ninth of November 1953 records that Mr Giese promised Mr Lee the full cooperation of all those present. And on the nineteenth of November 1953, five chamber officers took the mayor-elect to lunch at the New Haven Lawn Club.² He did not yet have the job. They briefed him, over lunch, on the programme.

Nothing in that paragraph contradicts anything in the book. The same mayor made the same moves, and won the same fights, in the same order. One variable changed, and it is not a fact about New Haven at all. It is where the story was started. Begin in 1953 and a politician initiates a programme and business follows him. Begin in 1943 and a plan nine years old is carried across a lunch table to a man who does not yet have the job. Every answer to who decided assumes a window, and the window is not found in the evidence. Somebody picks it. Ask who decided, and you must first say when the story starts, and there is no innocent place to start it.

Now the reply, at the strength it actually has, because it is not a dodge. If evidence from before the decision counts, then say where that stops. The 1943 plan had an author, who had teachers, working inside a profession that had spent decades deciding what a slum was. Push the clock back far enough and every outcome has an ancestor, and the ancestor is always upstream of whoever you were about to blame. At that point there is no evidence that could show the claim false. That is the exact thing Dahl objected to in the old reputation lists. His rule was built to cure it. The people correcting him have brought it back. This is not settled scholarship being reported to you. The disagreement with the sociologist who dug out those minutes was never settled.

Which brings the problem back to your list, and to a fact about that list you will not like. For most of the twentieth century, American elite research had a physical object to point at. Directors of large companies sat on each other's boards, so the boards overlapped into one dense web, and the web could be drawn. That is a statistic about social ties between named people. It is not a statistic about wealth, and not about outcomes, so nothing about winning or losing follows from it directly. In 2016 two management researchers documented that the web came apart in the 2000s. Well-connected directors stopped being preferred. They re-ran classic studies built on the old network, and the established findings no longer held.³ The tissue dissolved. The concentration it was invoked to explain went on concentrating. So whatever is producing the skew, it is not the network of acquaintances you can name.

The two results point the same way. In New Haven, a plan with authors and dates existed, and the method that looked for a decider found a broker. In the boardrooms, the acquaintance network that was supposed to explain the pattern died, and the pattern kept running without it. Every beneficiary of an arrangement can be named. You can list them by fund, by district, by disclosure, and the list will be accurate. And you can finish that list without having named a single person who chose the thing. So what were you actually looking for, when you were looking for a name.

In 1502 the Romagna was a patchwork of small lordships, and one man was welding it into a state. Cesare Borgia had taken the towns one at a time. He needed it quiet, and he needed it quiet fast. So he sent in a subordinate, Remirro de Orco, with plenary power to do whatever pacifying required.⁴ A Florentine envoy was attached to that court, watching him work, and writing home about it.

Remirro pacified the Romagna by terror, and the method worked. The province was pacified. It also became a province that hated the government which had done this to it. The hatred did not stay with the officer who ordered the killings. It travelled up, to the man who had appointed him.

That envoy spent the rest of his life turning what he saw into a manual. The manual contains an argument that almost nobody bothers to state as an argument. Start where he starts. A ruler's aim is to keep his state, and keeping it requires that people reliably do what he says. Compliance runs on one of two things: love, or fear. Love is a bond of obligation, and men break that bond whenever breaking it suits them. Its continuance sits in their hands. Fear runs on your own capacity to punish, which never passes, so its continuance sits in yours.⁵ Any agent thinking clearly prefers the instrument whose supply he controls. That is the whole of it. Power is what does not depend on somebody else's continuing goodwill.

The cliché gets the next step backwards. This is not a recommendation of cruelty. Hatred is the failure condition of the system rather than its goal. The same manual is emphatic that a ruler who strips his subjects of property and honour has wrecked his own instrument. Fear is a supply you meter. Hatred is what happens when you spill it.⁶

Which is what Borgia had on his hands at the end of that year, and it is why he did what he did next. On the morning of the twenty-sixth of December 1502, the people of Cesena came into the public square. The governor of the Romagna was lying there in two pieces. The executioner's block and a bloody knife had been left on the stones beside him.⁷ Cruelty is an instrument you can lend out and take back. Nobody in that square has to wonder who decided.

That is the model you have been working with, and it is a serious model, not a naive one. A century and a half later it got its cleanest definition, from a man watching England tear itself apart over who had the right to command. The power of a man, he wrote, is his present means to obtain some future apparent good.⁸ Take that slowly. Power is a stock of means held by a person: strength, prudence, eloquence, money, reputation, friends, luck. You hold it before you spend it, and you spend it to get outcomes. It is the same grammar as a bank balance, and it is why counting who holds what feels like the obvious way to study politics.

And that same author built the strongest available answer to the question you were actually asking. His problem was not why to obey a good king. It was that when every man judges for himself who has authority, no shared standard is left, and the result is war. So the many hand one man, or one assembly, the right to act in their name. Each does it only on condition that everybody else does the same. The consequence is exact. There is now a single nameable decider, and his acts are your acts, because you authored them. And that is the trap. Because his acts are your acts, nothing he does to you counts as an injury done to you. You have no standing to accuse him of injustice.⁹ The one person he answers to is you, and you have already answered.

Here is where it comes apart, and the first blow lands from inside the tradition. Two hundred and fifty years after the envoy wrote it, the man who wrote the founding text of popular sovereignty read The Prince. He said it was the book of republicans. His reasoning is in a footnote. The teaching of that manual contradicts the envoy's own larger works, so the manual has been studied only by superficial or corrupt readers. Then the detail that does the work. The Court of Rome sternly prohibited the book, and he says he can well believe it, for it is that Court the book most clearly portrays.¹⁰ Describing how power operates was treated by the people operating it as an attack.

And the larger work he points at is the one nobody quotes. In that larger work the envoy argues that the conflict between the nobles and the commons was the primary cause of Rome keeping its liberty. Every law favouring liberty came out of that dissension. He argues that the guardianship of liberty is safer in the hands of the many. The great want to dominate, and the commons only want not to be dominated. A guardian with no appetite for the thing he guards will not steal it. And a republic must give its citizens a lawful channel to accuse powerful men in public.¹¹ Otherwise the grievance goes out into the street as private violence. You have the names. What you do not have is the channel. The architect of the possession model thinks the interesting question is institutional, not personal.

The same reversal has happened to realism's other founding text. The most quoted sentence in all of power-talk is that the strong do what they can and the weak suffer what they must. It is a speech in a war history. Envoys from a great naval power stand in front of a small neutral island and tell it to surrender or be destroyed. The strong do what they can and the weak suffer what they must is their argument for why the island should not bother resisting.¹² They destroy it. And the historian who wrote the speech spends the rest of the book taking it apart. The island had been no threat to them. The same self-confidence sent the fleet to Sicily, and Athens lost the war. Everyone who quotes that line as hard-headed truth is quoting the losing side's reasoning and skipping the chapters where it fails.

That fight is still live, and it is not a costume drama. The plebeian reading points at institutions Rome and Florence actually had. Offices the great were ineligible for. Assemblies the notables could not enter. Trials of powerful men judged by everyone. Selection by lottery, used because elections return the wealthy and the well-connected. The other school says liberty just means not living at anybody's mercy, and that law and a well-mixed constitution deliver it without any of that. The answer back is that a Florentine constitution could write down who the plebs were. A modern electorate has no such line in it.

Both sides of that argument still assume the thing works. The possession model holds up wherever somebody can produce the stock on request. A governor with plenary powers, a sovereign with a sword, a name you can put in the square. The trouble starts when nobody can produce it, and you are left needing a test for who has it. The answer the twentieth century gave arrives in 1957.

A document on the table, seventeen pages, dated 1995. It was written by the scientific advisory committee of an industry coalition on climate, by the coalition's own experts, for the coalition's own members. It says that the scientific basis for the greenhouse effect, and the potential impact of human emissions, is well established and cannot be denied. In early 1996 the operating committee had one section struck out, the section rebutting the contrarian arguments.¹³ The redacted version was the one members were shown.¹⁴ The draft surfaced in a lawsuit years later. Now say who exercised power in that room. Nothing was decided. Nobody was defeated, and no vote was taken.

The difficulty there is not that the facts are thin. They are unusually thick, with a date, a committee and a deleted section. The difficulty is that the instrument you are holding was built to score something else. The first real test arrived in 1957. It says power is what shows up when two sides want different things and one side's want becomes the outcome. So the procedure is: find a decision, find a conflict, count who won. That rule has one enormous virtue, which is that it can come out wrong. A claim that some group runs the town becomes a claim you could check and lose. Before it, the going method was to ask around town who had the power and write the names down. That finds an elite in every town that has gossip.

Earlier you were handed a clock, 1953 or 1943, the mayor or the planning memo. Here the clock stops being a footnote about method. Setting the clock at the moment of decision is what makes the decision-maker the winner, every time, by construction. Each of the three tests in this part is a choice about where the story starts, and the flattering answers come from the late clocks.

The second test was built to catch what the first one drops. One 1960 formulation runs like this. All forms of political organisation favour some kinds of conflict and suppress others, because organisation is the mobilisation of bias. Some issues are organised into politics, and others are organised out. Take that plainly. Any committee has rules about what belongs on its agenda, and those rules are not neutral between the people who want different things from it. Winning by keeping a question off the table produces the same result as winning the vote, at lower cost and with no record. So the second measurement rule is this: find the issue that never became an issue, and find the arrangement that kept it out.¹⁵ The redacted report is that arrangement with minutes attached.

Now two towns in Indiana, next to each other, in a 1971 study of what cities do not decide. Gary was built around a single steel company. East Chicago had a plural industrial base, several firms, none of them the town. Similar populations, similar filthy air. East Chicago treated dirty air as a public question years before anyone in Gary did. When an ordinance finally came, in 1962, local business helped secure it. In the years before that, while the question simply never arrived, here is what the record shows the steel company doing. Nothing. No lobbying, no threat, no meeting. Everyone in Gary already knew what the company could do to the town's payroll, and that knowledge did the work an intervention would have done.

Follow the Gary case forward and it stops being about an agenda at all. After enough years, the smell in the air registers as employment, and people will defend it. That is the third test, and it is the hardest one to hold. The third view asks whether the most insidious exercise of power is to prevent grievances from forming at all. It works by shaping what people want, until they see the existing order as natural.¹⁶ The measurement rule is: find where people do not want the thing in the first place. Note where that leaves the first test. The most effective form of power available scores zero on it, because a contented population produces no conflict to count. An internal memo from a tobacco company in 1969 puts the ambition in one line. Doubt is our product, since it is the best means of competing with the body of fact in the mind of the general public. Read that again for what is being sold.¹⁷ Not a claim about smoking. A state of mind in the public, bought as a product, so that the question never firms up enough to be fought over.

Three instruments are now on the table. One counts open fights, one counts missing questions, and one counts wants that never formed. The numbers you have probably already met come in here. One study takes nearly eighteen hundred polled policy proposals from 1981 to 2002. It records what the median earner wanted and what the moderately affluent wanted, and asks whether each was adopted within four years. The estimated independent impact of the affluent on the outcome is large. The estimated independent impact of average citizens is close to zero. That figure is not a count of how often the middle loses. It is what is left of the average earner's preference once the affluent preference is held fixed. When the two come apart, knowing what the median earner wanted tells you almost nothing about whether the policy passed. And it comes with a mandatory companion, from a critic who went back to the survey data. Across a comparable set of questions, the correlation between what the affluent want and what the median earner wants is about zero point nine four. Those two groups mostly want the same things, and the cases where they clearly diverge are under a fifth of the sample.¹⁸ His conclusion is that the middle gets roughly the policy it would have got with real representation, by coincidence. The reply from the other side is the sentence worth keeping: a coincidence is not owed to you, and it can end.

One study does remove the confound. A political organisation asked congressional offices for meetings. It randomised one detail, whether the office was told that the attendees were campaign donors. Told, senior policymakers made themselves available between three and four times more often.¹⁹ Nothing varied but the sentence in the email.

Every test so far reads power off something that happened. There is a way of counting that reads it off nothing at all. Since 1954 there has been a formal index that computes voting power from the rules alone. It asks how often a member is the one whose vote turns a losing coalition into a winning one. Run it on the United Nations Security Council, where a resolution needs all five permanent members plus four of the ten elected ones. On that measure the ten elected members share about one point nine percent of the voting power between them. Fifteen seats. Fifteen votes. Seat for seat, a hundred to one. And a member who is never the one who tips it has a power, on this measure, of exactly zero.

Now the objection, and it has never been answered cleanly. To claim a suppressed grievance, you have to claim what people would have wanted under conditions that did not obtain. Those conditions did not obtain, so there is no direct evidence about them. The standard objection from the critics is blunt. The data you would need are hard to gather or simply unavailable, and what you can gather settles nothing about how power is distributed. The modern version is sharper, and it concedes every fact. Getting what you want without having to act is luck, not power. If the steel company never acted, never threatened, and simply benefited from everyone else correctly reading their own interests, then it exercised nothing. Calling that power credits an agent with a structural fact. A theory that cannot tell the two apart will keep handing winners a power they never had.

The numbers do not rescue you from this, and it is worth being exact about why. A correlation between preferences and outcomes identifies no mechanism, no actor and no intention. It cannot tell you whether anybody did anything. The data go quiet in precisely the place where you wanted them to speak. And the quietest issues, the technical ones nobody polls, never enter the sample at all.

So you are left with three measuring instruments and no clean reading. The first misses every quiet victory and calls the perfect win no win at all. The second and third catch the quiet victories and license claims that nothing could refute. Underneath all three sits a question you have not asked yet. Not which test to trust. Whether the winners in this story are the people you have been counting.

Who has to show up in order to win. There is a room like this in every rich country, and it is sitting today. A committee is revising the rules on who may buy a company. Also what a buyer may do to it afterwards, and what happens to the people inside it. It meets for months. No election turns on it. No newspaper covers it, and the transcript is public and unread. Ask that room the question this video has been asking. Who decided. The honest answer may be that nobody in there had to win anything. The rules were written by the only people in the building who knew how the rules worked.

Your first instinct is that somebody in the room was bought, and that instinct is not stupid. It is the instinct behind the email experiment. Saying the attendees were donors got senior officials to make themselves available three to four times as often. But go and look at a committee like this one and the money is usually not there. No donation, no threat, no dinner. What is there is a technical question, a deadline, and four people who have spent their careers doing this for one side.

So here is the argument that explains the outcome without anyone in the room doing anything. Charles Lindblom put it together in 1977, and it comes in four steps. Step one. In a market economy, the decisions that set employment, prices and investment sit legally in private hands. A government may tax those decisions and regulate them, but it may not issue them. Step two. A government's survival depends on how the economy performs, and how the economy performs depends on those private decisions. Step three. So a policy that business finds unattractive produces less investment and less hiring, which lands on the electorate and then on the government.²⁰

Step four is the one everything rests on, so take it slowly. Officials who can forecast step three do not wait for it. They adjust in advance. Which means the adjustment requires no lobbying, no threat, no meeting, and no coordination among firms at all. Nobody has to want it. Nobody has to know it happened. A town could know what its steel company would do to the payroll. A finance ministry knows the same thing about every company at once, and it knows it on a Tuesday morning with nobody on the phone. The four steps together give you a systematic bias with no biasing agent. It is invisible in any record of votes, donations or bribes, because it leaves no record. Who benefited and who acted have now come apart completely. There is a colder version with a paper trail, and it is coming later.

That is why the three tests for counting who won were reading zero. In 1988 an international political economist drew the distinction those instruments were missing. There is the power to win a bargain, and there is the power to set the terms on which everyone else bargains. The second belongs to whoever supplies security, decides what gets produced, or controls credit. It belongs to whoever governs what counts as knowledge and who is believed. None of them has to enter a particular bargain to determine how it comes out. The menu was written before anyone sat down. An actor like that can lose bargain after bargain in public while the frame stays exactly where it was. So a method that scores power by counting bargaining victories will file the frame-setters under losers.

Now the case that shows what all of this looks like on the ground. In 2011 Pepper Culpepper published a comparison of how four countries wrote their rules on corporate control and takeovers, in France, Germany, Japan and the Netherlands. Different parties, different coalitions, different decades. He asked what predicted the outcomes. Party control of government did not predict them, and neither did the alignment of interest groups. What the outcomes tracked was the stated preference of the managers' associations, in country after country. The managers' associations brought the expertise, and on a technical question the legislators deferred to it. They had the room to themselves, because the issues were technical enough that nobody outside the industry held a view. What breaks the pattern is the issue becoming visible, a scandal or a plant closure, rather than an election.

Business does not have to win the argument in a room nobody is watching. It has to keep the room from being watched.

Which brings us to your own headline number, the one you have been carrying since the start. Three index managers cast, on average, about 25 percent of the votes at the largest American companies. The two law professors who counted that project it could reach as much as 40 percent within two decades.²¹ The 25 is measured. The 40 is their projection, and they say so. What those authors are actually worried about is the opposite of what you assumed. Their complaint is that index managers have structurally weak incentives to spend money on stewardship, and structural incentives to defer to the management already in place. So the annual meeting that number describes goes like this. A quarter of the votes in the hall are cast by people who never spoke to the company and hold no view about the resolution. The concentration is real, it is measured, and it is largely unused. You have been carrying that figure as proof of a hand on the tiller. The people who produced it are complaining that nobody is steering.

You came in wanting a theory that could name somebody. What you are holding now is a theory that no fact could ever embarrass. Run it against a government that goes after a particular industry. The account absorbs that as the state's independence, which it needs in order to serve business in general. Run it against a government that hands the industry everything it asks for. The account absorbs that too, as the state doing its job. That charge was made in 1970, inside Marxism, by one Marxist against another.²² It had been aimed at the ruling-elite theorists a decade before that. Remove the agents and you have removed the evidence, because no observation is left that could count against you. A theory nothing can embarrass has told you nothing about your Tuesday. And if nobody has to act, you have already seen how hard it becomes to tell any of this apart from luck.

There is an answer to that, and it is only half an answer. Business power varies, and what it varies with is attention. Attention can be measured in advance of the outcome instead of read off it afterwards. That gives the claim a shape the older account never had. On quiet, technical questions, organised business gets what it asks for. On loud ones, parties and voters push the outcome away from it. So a quiet corporate rule that went against the managers' associations would sink the whole claim. A claim that can be sunk is finally saying something.

Then pay the cost in the same breath, because it is a real one. Go back to the committee from the start of this part, months into a technical rewrite, its transcript public and unread. And by the salience argument, that is precisely where the advantage is largest. The best structural theory in the field says business power lives where nobody is looking. The best data in the field cannot look there. They go silent in exactly the same place.

Still, everything in this part has one assumption underneath it that has gone untouched all the way through. Every mechanism here needed beneficiaries with assets. Investment that could be withheld, credit that could be priced, votes that came attached to shares. The winners were positions rather than people, but they were still positions with money in them. That is the last comfortable assumption you have left. Every winner so far held something. The next case is one where the people who owned nothing set the terms everyone else was judged by afterwards.

There is a book with a title on the spine that its author never chose. It is called The Will to Power. It is where most readers meet the line that made his name a byword for power. This world is the will to power, and nothing besides. Nietzsche collapsed in 1889. His sister took control of the papers, and with an assistant she selected notebook fragments, ordered them, and in places altered them. The book came out in 1901, and an enlarged four-book edition followed in 1906. In the 1960s two Italian editors went to the archive to work from the original documents.²³ Out of that came the first complete chronological edition of what was actually in there. The philological case, as Nietzsche scholars report it, turns on the closing section. The famous sentence comes from material Nietzsche discarded in 1887, notes he passed over while assembling his own books. A century of readers met his most famous doctrine in a volume that climaxed on a fragment he had thrown away.

That is a fact about publishing, and it decides what the doctrine even is. Take the notebooks away and go to the published passage the whole cosmological reading rests on, section 36 of Beyond Good and Evil. The grammar of it runs like this. If nothing is given as real except our world of desires and passions.²⁴ If we grant that the will operates. If, finally, our whole instinctive life could be explained as one fundamental form of will. Then, he says, one would have acquired the right to define all active force as will to power. Four ifs, stacked. He never gets out of them. Nothing in that passage is asserted. And the whole thing hangs on one of those ifs: that when you will something, your willing is what makes it happen. Nietzsche denies that flatly, elsewhere, in his own name.²⁵ He is showing you what would follow if you believed something he thinks is false.

So set the cosmology aside and keep only what the books he actually published will carry. In the first essay of the Genealogy he makes a claim about grammar. We say the wind blows, as though there were a wind that could stay still and sit the blowing out. There is no being behind the doing, he writes. The deed is everything, and the doer is a fiction we add. Follow that where it goes. If there is no neutral agent standing behind its own force, then there is no substance for power to be a property of.²⁶ The question you keep asking, who has it, is not being hidden from you. It has no referent to attach to.

Then the political half, which is where this stops being a point about verbs. The lambs resent the birds of prey, so the lambs judge that birds of prey are evil. And whoever is least like a bird of prey is therefore good.²⁷ A lamb, for instance. The birds are unbothered. They have nothing against lambs, and they rather like them. But look at what the lambs have done, because it is not nothing. They have manufactured a scale of value out of a grievance, and the word for what they did is creative. Values made that way do not stay with the people who made them. They spread, and eventually the birds accept them too, and begin to feel guilty about being birds. Nothing was seized and nothing changed hands. The whole mechanism was the propagation of an interpretation.

The purest case in the book is a man with no army, no treasury and no office. The ascetic priest arrives among people who are suffering. Ascetic means what he sells: self-denial, less appetite, less comfort, less of you. What a sufferer wants above everything is a cause, somebody to blame. The discharging is what deadens the pain, on Nietzsche's account, and that is what the blame is for. The priest supplies one, and he turns the direction of it inward.²⁸ You yourself are the only one to blame, he says, you alone are to blame for yourself. That is not an order. It is an offer, and it is accepted, because it hands agency back to people who had none. You cannot be to blame for something unless you were able to act. The weakest people in his story produced the most successful exercise of power in it, and they did it by giving pain a meaning.

Nietzsche wants to judge slave morality, and that is where the position cracks open. He calls the ascetic ideal, the ideal of self-denial, a sickness, and he calls the whole apparatus life-denying.²⁹ Those are verdicts, not descriptions. But he has just shown that slave morality is itself an enormous exertion of will to power. So on what standard does it fail. If everything whatsoever is will to power, then nothing can fall short of it, and a test that everything passes is not a test. And that leaves you somewhere you did not ask to be. You did not choose the interpretation of your own suffering. It was handed to you before you were old enough to weigh it. On this doctrine that handover was just another exertion of the same force, and so is your complaint about it. The doctrine that dissolves the owner also dissolves the critic, and you were counting on being the critic.

The strongest live repair narrows what the word power means. Bernard Reginster reads it as the desire for the overcoming of resistance, and that has three conditions that must all hold. You need a first-order desire for something definite. You need resistance to getting it. And you need to actually succeed against that resistance. A climber wants the summit. Send a helicopter and set her down on the top. The desire is satisfied and the willing is empty, because what she wanted was the mountain refusing her. That is why the thing cannot be banked. An overcoming is not a substance you can be holding on an afternoon when nothing is happening. It exists while it is being done, and doing it consumes the obstacle it needed. The route has to get harder, or the power is spent.

Reginster owns the paradox that comes with this, and does not treat it as an embarrassment. A will to power wants resistance, which means it wants what frustrates its own first-order desires. A fully satisfied will to power would destroy the conditions under which it could be satisfied at all. Readers still argue about how wide the doctrine runs, from a claim about all of reality down to a claim about human motives. The version that would reach you is the narrowest one. The aim is built into acting as such, so that merely by acting you commit yourself to it. The objection to that move is small and fatal. Approving of an action once you know what caused it does not commit you to approving of the causes. People endorse what they did and repudiate the resentment that drove them, every day.

What survives all of it is the negative result, and it is enough to be going on with. Power is not a quantity anyone is sitting on. It is something being done, in a relation, and it can be done to devastating effect by people who own nothing. Which leaves one thing standing that everybody still believes, the master and the priest and you and the man at the next desk alike. That power gives orders, and people obey them. That had better be true.

On the fourteenth of August, 1971, a basement corridor in the psychology building at Stanford was dressed as a prison. Some students were made prisoners, others were made guards. It was halted on the sixth day. Almost everyone who has heard of it remembers the same lesson. Put an ordinary young man in a uniform, and he will turn cruel on his own. Hold that basement in view, because we are coming back to it, and the corrections to the story make the point sharper rather than gentler.

Start instead with the picture the whole search depends on. Somewhere above you is a person whose will is executed by other people, and the chain of execution eventually reaches your Tuesday. That is the prince model from earlier, extended downward: power as a thing somebody holds and discharges through commands. It is not a silly picture, and it has the most famous experiment in psychology behind it. At Yale, years before that basement, Milgram sat volunteers at a shock generator and told them to punish a learner's mistakes with rising voltages. In the condition everyone quotes, every single participant went to 300 volts, and 65 percent went to the maximum.³⁰ No weapon, no threat, no penalty for walking out.³¹ A man in a grey coat, and people did it anyway.

Then somebody went through the paperwork. The psychologist Perry worked from Milgram's own archive at Yale. Across Milgram's studies as a whole, she found, only about half the participants fully believed the setup was real. And of those who did believe it, the majority refused to go on. She reported a troubling mismatch between the published descriptions of the experiment and the evidence of what actually happened in the room.³²

And the famous number is one setting of the machine among dozens. Milgram ran variant after variant, changing who was in the room and how close the authority stood. Across those variants, the proportion who went all the way to the maximum ran from zero to nearly one hundred percent. Across the studies as a whole, a majority of participants chose not to go that far. A range that wide is not a fact about human nature. It is a fact about the arrangement.

Which leaves the question of what, inside the arrangement, was doing the work. Milgram's script gave the experimenter four things to say when a participant hesitated, in a fixed order. The first simply asked him to keep going. The second: the experiment requires that you continue. The third pressed that it was essential he continue. And then the fourth: you have no other choice, you must go on. One of those is a command. The first three are a man in a coat saying that a piece of science needs help. The fourth says that a person's will has been cancelled.

When researchers went back and tracked which prompt preceded which behaviour, the pattern was consistent. The further down the sequence the experimenter had to go, the less likely the participant was to continue. And when Haslam and his colleagues re-ran the paradigm as an experimental analogue, they stated it as a double prediction. Continuation was positively predicted by how far a prompt appealed to the goals of the science. It was negatively predicted by how far that prompt constituted an order. The fourth prod, the only sentence in the entire script phrased as a command, is where people stopped.³³ So the evidence built on the founding paradigm does not show people obeying. It shows people being recruited, and quitting at the moment recruitment turns into instruction. The name for this is engaged followership: the participant is not surrendering to a voice, he is helping with a project he has been persuaded matters.

In 2019 the historian Le Texier went into the Stanford archive. The reflex he was testing is a good one, and it is the story most people reach for the moment the command model dies. If nobody effective was giving orders, then the cruelty came from the situation itself, from the role and the uniform and the corridor. That is exactly what that basement is famous for proving. What Le Texier published in American Psychologist gives it the same problem. The guards, he reported, received precise instructions regarding the treatment of the prisoners. The induction is on record. You can create in the prisoners feelings of boredom, a sense of fear to some degree.³⁴ You can create a notion of arbitrariness, that their life is totally controlled by us, by the system, you, me. We are going to take away their individuality in various ways.³⁵

That is the experimenter, before the study began, describing the finding the study is credited with discovering. There is tape of the warden pressing a guard to be tougher for the benefit of the experiment. And the breakdown that has been shown in a thousand lecture halls as proof that the situation was overwhelming has a second version. The prisoner in question was released after about thirty-six hours.³⁶ He said decades later that he faked it to get out and study for an exam. The flagship exhibit for structure without an author turns out to have an author, with a clipboard, giving instructions. Le Texier's conclusion is blunt. The conclusions were largely written in advance of the experiment.³⁷

That should not comfort anyone. In 2006 Haslam ran a comparable study with the guards left to define their own role, and the results went somewhere unfamiliar. The guards would not identify with being guards. They were reluctant to impose authority and were eventually overcome by the prisoners. The participants then built an egalitarian arrangement between them, which is the outcome everybody hopes for. It proved unsustainable. And as it came apart, moves to impose a tyrannical regime met with weakening resistance.³⁸ Nobody had to be briefed for that part.

There is a phrase most people reach for at this point. Arendt wrote it watching Eichmann in a Jerusalem courtroom, and her own claim was narrow. His only personal distinction, she wrote, was a perhaps extraordinary shallowness. Evil is never radical, only extreme, with neither depth nor any demonic dimension.³⁹

The popular version has drifted a long way from that. In it the perpetrator is a sleepwalker, a functionary who was not really present for what he did. The banality of evil comes to mean that nobody was doing anything. The two laboratories contain the opposite, and the researchers who went back to them name that thesis and reject it by name.

So both halves of the laboratory story fail for the same reason. Each one claims the cruelty arrived with nobody putting it there. In the shock room somebody wrote the four prods. In the basement somebody briefed the guards on what to create. And the 2006 study adds the part nobody briefed: people abandoned by a group that is coming apart go looking for someone to run it. What survives is a claim about the mechanism, and it is not a soothing one. Haslam's reading of this evidence, which currently holds the field, says something harder. The tyranny in that study was made possible by the active engagement of enthusiasts, and perpetrators are engaged followers rather than blind conformists. The alternative to a puppet master is a room full of volunteers who believe they are contributing to something worthwhile.⁴⁰

Which is the finding that should trouble you at your own desk. The strongest lever in that laboratory was a sentence about why the work mattered. The weakest was the one that told a person he had no choice. Nobody at eleven at night tells you that you have no other alternative. They tell you the launch depends on it, that the team is counting on you, that this quarter is the one that decides things. Compliance runs on a shared project rather than a voice giving orders. Which means the thing doing the telling is not giving you an order. There is nothing there to disobey. So where do you send the complaint.

A clerk closes a file and puts it back on the desk. Not an image for something else, an actual folder of written documents, preserved in their original or draft form.⁴¹ It sits on that desk, and when the person at that desk retires, it does not leave with them. The next occupant opens it at the page where the last one stopped, and carries on. Max Weber built his entire account of modern rule out of that folder. It is the least interesting object in this video, and it is the reason nothing changes when you replace the person.

Start where he started, which is a problem about scale. In a household, one man can be obeyed by being present, watching, and remembering who did what. Past that scale, being obeyed requires a staff, and a staff has to be recruited, paid, supervised and replaced. That last word is the hinge. A post has to survive the departure of whoever is in it. So its duties must be specified in a way that never mentions him. So you get what Weber calls fixed and official jurisdictional areas, generally ordered by rules, that is, by laws or administrative regulations.⁴² Each desk gets a boundary, and the boundary is written before anyone sits down behind it. Which is what the files are for. Which is why the office is separated from the official's home. And why, in his phrase, public monies and equipment are divorced from the private property of the official.⁴³

That last item carries more weight than it looks. Under the old regime a man could buy an office, and having bought it, he owned it the way he owned a field. He could sell it on, leave it to a son, or bargain with it when the crown wanted something. The modern official owns nothing. He cannot take the post with him, cannot bequeath it, and has nothing to trade when he threatens to go. When he goes, the budget stays, the seal stays, the archive stays, and the half-finished case stays with its annotations already narrowing which rule applies. Continuity here is a fact about paper, not a fact about loyalty.

Then comes the move that makes this a theory of power rather than a description of stationery. The sociologically interesting fact about a command is not that it gets carried out. Fear does that, and so does habit, and so does calculation. What matters is that the command is accepted as a valid norm. The clerk does not obey because the rule is backed by police, though it is. He obeys because the instruction was issued within jurisdiction, by the correct procedure, under a rule that is on the books. The ground of the duty is the legality of the rule. Not one property of the man who issued it, who is bound by that same rule. Which does something to the list you can recite. The funds, the donors, the committee chairs are, on this account, carriers. The tax assessment posted the morning after a coup is not a different assessment.

Peter Evans and James Rauch published a much-cited study in a sociology journal in 1999. State agencies scoring higher on meritocratic recruitment and long careers, they reported, were associated with higher economic growth. Later work has gone back to re-test it. What that score measured was how completely the posts were insulated from the people filling them. Here is that structure producing an outcome. A licence application is refused. The caseworker signed the refusal, but she did not write criterion four, and the criterion is unambiguous on the facts in front of her. The official who drafted the criterion never saw this applicant and could not have. The minister who signed the regulation has never heard of her. The annotation that decided which criterion applied was written by someone who retired in 2019. Ask every link in that chain whether they decided this, and every one of them can answer honestly. No, I applied it. The refusal is still signed, still real, still enforceable, and the applicant still cannot work.

Hannah Arendt gave that arrangement its name in 1969. Rule by Nobody. If we identify tyranny as the government that is not held to give account of itself, rule by Nobody is clearly the most tyrannical of all. Since there is no one left who could even be asked to answer for what is being done.⁴⁴

The attack comes from two directions. The first is internal, and it comes from his best modern critic. Weber grounds the whole structure in legitimacy, and defines legitimacy as belief in legitimacy. Ask why people obey and the answer is that they believe they ought to. Then ask what makes the order legitimate, and the answer is that they obey believing they ought to. The circle closes and nothing independent has been said. The critic's replacement is one sentence. A power relationship is legitimate because it can be justified in terms of the beliefs people already hold. That moves the test outside their heads. You are no longer surveying whether the governed accept the arrangement. You are asking whether it could be defended to them, which is something an outsider can check. And it explains what Weber's version cannot see, which is a regime steadily losing legitimacy while every single person still complies.

The second attack is empirical, and it is worse for the theory, because the theory half concedes it. The impersonal office badly understates how much of the outcome real officials generate. The rules underdetermine cases, and somebody has to close the gap. Weber himself says where the leverage comes from: technical expertise, plus official secrecy about what the file actually contains. Neither is conferred by any rule. Both belong to a person. The machine built to keep persons out has a door in the back, and the trained official is standing in it.

One piece of Weber's own scheme has been missing from this, and without it the end of his life makes no sense. He held that anyone claiming a right to be obeyed rests that claim on one of three grounds. The first is legality. You obey because a rule is on the books, and it was applied by the correct procedure. That is the office, and it is what this part has been describing. The second is tradition. You obey because it has always been so, and the man giving the order holds a status the past has made sacred. The third is the strange one. You obey because of something about the person, a quality you take to be exceptional. His word for it is charisma.

That word has drifted since, and his sense of it is narrower. It is not charm. It is a recognition, granted from below, by the people who follow. Nobody is appointed to it and nobody is issued with it. Which means it sits with the followers. You have handed it out yourself, to a manager or a candidate or a writer, without ever being asked to. It can be taken back. The prophet who stops delivering stops being a prophet, and no rule protects him, because rules are the thing he arrived without.

So that kind of authority carries a problem the other two do not. It cannot survive an ordinary week. A movement resting on one man's recognised exceptionality has no salary structure, and no answer to who comes next when he dies. So it builds them. The staff has to eat, so there are stipends, and stipends need accounts. The succession has to be settled, so the grace gets attached to a son, or to a position, or to an election. Weber's name for that drift is routinisation. What it does is manufacture an office out of a person. You have watched a small version of it. A founder's habits become the way things are done here, and then a written policy, and then a form somebody fills in. The exception becomes the rule, and the file appears.

The direction of that matters. It runs from the person to the office, and that is the only direction his own history ever shows it running.

Consider a man in Germany in 1919. The empire has collapsed, and there is a constitution to write. He is invited onto the draft board, and onto the German delegation to Versailles. He is famous among scholars, and almost entirely unsuccessful as a politician. His political efforts in those months bear next to nothing. One thing gets in. His advocacy of a strong plebiscitary presidency, a head of state elected directly by the whole nation, standing outside the parliament and above the ministries. He had a word for the arrangement, leader democracy. An election, on this picture, is a contest in which one man acquires a personal following. The whole nation picks him, not a party committee, and it picks him for the pull he has over people. And then the part the arrangement stands on. He stays answerable to the people who elected him. Which is Weber putting back the one thing the office had removed, somebody who can be asked.

One historian spent a career arguing that this followed from Weber's sociology rather than lapsing from it. And that it was wide open to an authoritarian reading. His critics answer that the elected leader was designed precisely to control the bureaucracy he had argued was far easier to capture than to destroy. That dispute is still running. But both sides agree on what happened, and what happened is the strange part. The theorist who proved that the office beats the person spent his last political energy trying to install a person strong enough to master the office. His own argument is the reason. If the apparatus outlives every change of master, nothing inside it can check it. The only counterweight left to a position is a human being.

So the machinery is real, and the file does not care whose hand is on it. There is nobody to argue with and nobody to petition. A machine that nobody can be asked about is still rule, running exactly as it was written to run.

Two documents, side by side. The first is a criminal sentence handed down in Paris in 1757, against a man who had attacked the king with a knife. It orders that his body be tortured in public, torn apart, and burned, in front of whoever comes to watch. It was signed. The second document is about eighty years later, and it is a set of house rules for a house of young prisoners in the same city. Rising at six in winter and five in summer. Nine hours of work a day, two hours of instruction.⁴⁵ The day ends at nine in winter and at eight in summer, and every transition from one activity to the next is marked by a drum. The first document has an author who put his name to it. The second document's author is a schedule.

The obvious reading of that pair is the humane one, and it is mostly right. Somebody stopped being burned alive. Read as a story about power, the obvious reading says that the amount of power in the room went down, because the violence went down. Foucault's objection is not that the drum is worse than the fire. Nothing in it asks you to prefer the scaffold. The objection is about what quietly went missing between the two pages. On the first page there is a man who ordered the punishment and can be named. On the second page the punishment runs every waking minute, over far more people, for years at a stretch. It also produces a whole literature about the character of the offender. Nobody in that building is doing anything except following the clock.

Which raises the design question the reformers were actually working on. How do you get continuous obedience out of people without paying for continuous supervision, since guards cost money and sleep. The answer Jeremy Bentham published in 1791 was architectural, and it runs as a short chain of premises. First, separate seeing from being seen. Cells arranged in a ring, each one open to a central tower, and no way for the inmate to tell whether the tower is occupied.⁴⁶ Second, the inmate is therefore permanently visible, and the inspector's presence is permanently unverifiable. Third, a man who cannot tell when he is being watched, and who is punished when caught, has one rational policy available. Behave as though the tower is always occupied.

Now follow what that does to the operator. If the inmate has already adopted the policy, his conduct no longer depends on anyone actually looking. The watching has been transferred into the watched, who now carries it around at his own expense, for free, in an empty corridor. So the identity of the person in the tower stops mattering to the mechanism. A cruel inspector and a kind inspector produce the same corridor. An empty tower produces the same corridor, provided nobody can confirm that it is empty. That is an arrangement in which power is exercised continuously and held by nobody in particular, and the building was meant to prove it in brick.

It never got built. The scheme lost in procurement. Bentham handled the purchase of the ground at Millbank on behalf of the Crown, for twelve thousand pounds. Authorisation to build never came, and the plan was abandoned in 1812. The ordinary prison put up on that ground instead cost about five hundred thousand pounds, more than twice its own estimate.⁴⁷ That is roughly forty times what the land had cost. The cheap self-running inspection house was beaten in a tender by an expensive conventional one. And then it was built anyway, a century later and an ocean away. Five circular blocks, cells in tiers around central observation posts, designed for about five thousand men. Raised between 1926 and 1931 on a Cuban island, by the Machado dictatorship. After the 1953 attack on an army barracks, that state took in its most dangerous prisoner, a young lawyer who had led it. It held him there until 1955. It did not put Fidel Castro in the circular blocks. It put him in the hospital wing.

The self-watching premise is the one everybody repeats and almost nobody measures, so here it is measured. In June 2013 the scale of state internet surveillance became public, worldwide, inside a single news cycle. The legal scholar Penney tracked what happened next to traffic on privacy-sensitive Wikipedia articles. Views dropped immediately, and the longer-run trend in views shifted too, which is not what a passing news story does.⁴⁸ Nobody in that population was arrested. Nobody was addressed, warned, identified, or contacted at all. They were only made aware that they might be visible, and traffic to those articles bent.

Take the version of this you have already sat through. A child sits an examination. Nobody in that room intends to dominate anybody, and the teacher may be kind, underpaid, and measured by an inspection regime of her own. The exam still distributes thirty children along a single scale, and writes each one into a file that will follow them. It makes deviation visible, which is the condition of deviation being corrected. And it manufactures the vocabulary the child will use about himself for the next forty years: able, average, careless, gifted, behind. You still use one of those words about yourself, and you did not choose it. The exam forbids him nothing. It makes him someone.

The data went missing. Two economists, Levitt and List, recovered them and published in 2011. The data were from a factory outside Chicago in the 1920s, and they are the source of the most repeated anecdote about being watched. Workers there were said to improve under observation, whatever was done to the lighting.⁴⁹ Their verdict on the famous patterns was that existing descriptions of them are entirely fictional. They did find some subtler effects. The story everyone tells was not in the numbers.

The historians hit harder, because they hit the archive. The design was never built as designed, as we have just seen. Real nineteenth-century prisons were chaotic, underfunded, disease-ridden, and run by staff who neither knew the disciplinary programme nor implemented it. One criminologist's charge is that this whole way of punishing, correcting people rather than hurting them, barely shaped Victorian criminal justice in practice. Then, in 1981, a historian of the prison went at the foundations. The historians who follow Foucault write as though the state does the punishing. His charge is that punishment was never the state's monopoly.

The reply is available and it is decent. Foucault is analysing a diagram, a way of arranging visibility, and not conducting an inventory of the institutions that happened to house it. And failure is written into his own account, which says the prison has always been known not to work. Its failure is productive, because it reliably manufactures a class of delinquents somebody finds useful. Say the cost of that reply out loud. An argument confirmed when the mechanism works, and confirmed again when the mechanism collapses, has stopped taking instructions from evidence. That is exactly the complaint his philosophical critics make from the other side, that this is normative work wearing the clothes of empirical history. The historians ask for an inventory and get a diagram; the philosophers ask for a standard and get a description.

The tower stands in the plan and not on the ground. It lost, in 1812, on cost, to a building with corridors and locks and men who were paid to walk them. And you have been behaving all week as though it were finished, staffed, and looking at you, in a country where it was never built. The argument did not need it.

What would it mean to own a thing that only exists while it is happening? Ownership has a test, and the test is absence. Your bicycle is yours while it is chained to a rail and you are three floors up in a meeting. A field belongs to someone in February, when nothing at all is growing on it. Property survives the stretch of time in which nobody is using it. That is most of what having something means.

We talk about power in exactly that grammar. He has a lot of it. She lost some. They are holding on to it. Nobody flinches at those sentences, and they commit you to something you would never sign if it were written out plainly. They commit you to a quantity, sitting somewhere, which stays what it is when nothing is being done with it. In 1982 Foucault wrote that power exists only when it is put into action.⁵⁰ There is no stock of it sitting anywhere in between.

What exists is a relation between actions. To exercise power is to act on the actions of other people, to structure the field of what they can do next.⁵¹ It does not move a body. It arranges what there is to choose from, and then leaves the choosing to the person. This is also why the who question keeps coming back empty when you press it on a real institution. Ask who runs a prison, and every candidate is himself placed. The warder is watched by the inspector, and the inspector answers to the ministry. The ministry answers to the statistical returns and the newspapers. There is a strategy and there is nobody occupying the position of strategist.

Set the premises out so you can check them yourself. One, power exists in relations between actions, and only while those actions are being performed. Two, whatever exists only in the relation between exercises is not a substance stored inside a person or a building. Three, so talk of holding power is shorthand for standing well inside relations that continue only while they are run. Otherwise it is a mistake about what kind of word we are using. The move underneath all three is the one you met earlier, that there is no doer sitting behind the deed, no quiet substance behind the force.

There is a fourth premise, and it is the one that earns everything else. Positions can be radically unequal without any of them being ownership. An advantaged position is a standing likelihood of winning exchanges, renewed every single time one happens. The firm that can wait six months and the applicant who cannot wait six weeks are not equals, and nothing here says otherwise. Their inequality lives in the exchange, and it has to be run again next time to still be there. Without that premise this sounds like a denial that anyone is stronger than anyone else. With it, the claim is only about grammar, and every inequality you can name stays precisely where it was.

Two limits stop this becoming mysticism. Chains are not power on this account. A man in irons is in a physical relation of constraint, and the word does no work there. Power is exercised only over free subjects, and only insofar as they are free, which means it needs someone who could have done otherwise.⁵² That is why a timetable counts and a locked box does not. Late in his life Foucault drew a second line, between power relations that are still mobile and states of domination. In a state of domination the margin of manoeuvre has collapsed, and the relation is fixed and runs one way. So the account has an outside, and chains and a collapsed margin of manoeuvre both sit on it.

Which is why the search that brought you here keeps coming back empty. A deed still exists in a drawer at three in the morning. Power has no three in the morning. It is not hidden from you. It is not the kind of thing that can be had.

And every timetable was written by a hand. In February 1971, four years before the book that says power belongs to nobody, Foucault founded a group with a historian and a journalist. The purpose was small and exact: to give prisoners and ex prisoners a voice in French society. The premise was that the penal system converted petty criminals into hardened delinquents. The method was to collect what prisoners themselves said, and then publish it. They held press conferences. When Toul rioted in December 1971 and the riots spread to other prisons, the group was there with the words of the men inside. There was a police crackdown, and arrests, repeatedly. It ran for about two years and drew a few thousand people before it wound itself up.⁵³ The man who would write that power has no owner spent those two years behaving as though the sentence had a second half.

Now turn on all of it, because it does not survive its first two objections intact. You answered the message at eleven forty. Nobody made you. The caseworker who refused the licence could say the same thing, and say it honestly. No, I applied it. So who imposed what on whom. The objection standardly credited to Charles Taylor, in 1984, starts right there. To say power was exercised over you is to say something ran against what you wanted. Strip that out and the word covers every exchange between any two people, which is the same as covering nothing.⁵⁴ But saying the message was IMPOSED on you means knowing what you would have wanted if nobody had shaped you. That is a truer version of you, sitting underneath the shaping. And the histories you met earlier deny that anyone has one. They trace a value back to the need that produced it. The priest's self-blame back to a sufferer's want of somebody to blame. The same trouble hits the unmasking. The last hour has been telling you that the thing running your Tuesday is not what you took it for. That claim needs the new description to be truer than the one you were living by. Take the going rules of a time and a place, the ones that decide what can count as true at all. They are what made your list of funds and donors look like the answer. If truth only exists inside them, nothing was unmasked, and one description was simply swapped for another.

The second objection is sharper, because it comes from inside the sympathy. It also starts from the question you are probably holding right now. If nobody is holding the thing that ran your evening, what are you supposed to mind, and who do you take it up with. In 1981 one political philosopher granted that the account of modern power was genuinely new, and called what he built on it normatively confused. He suspends the question of what is wrong with an arrangement, on principle, and then never comes back for it. So take an arrangement that hurts nobody. Nothing forbidden, no violence, no bones broken, every hour of your life shaped and every shaping accepted by you. On his own terms that is another regime of power, no better and no worse than the one you are in. Say you prefer yours. The reason you give will already assume the standard that is in dispute. Where there is power there is resistance describes what happens. It gives you no reason to be on either side of it.

The best current answer, in a 2019 reappraisal, supplies the missing standard, and the standard is freedom understood as continuous self transformation. The arrangement that hurts nobody is objectionable because it forecloses your capacity to become someone other than what it has made you. And critique is problematisation rather than judgment handed down from a standpoint. It shows that what looks necessary was in fact made, and can therefore be made differently. The decision is left to whoever the argument reaches, which tonight is you. Then the honest part. Half of that answer rests on Foucault's own conduct as a critic rather than on a principle you could state and test against a case.

No owner does not mean no author. Somebody wrote the timetable, and the man who enforces it obeys it too. The rule that decides whether you did well this quarter was chosen by a person, and that person can be found.

In February 2022 a mathematician published a document about his own employer. Michael Thaddeus teaches in the mathematics department at Columbia University. What he had done was audit the figures Columbia reported to the magazine that ranks American universities. Class sizes. The share of faculty holding the top degree in their field. The share working full time. The student to faculty ratio. And about three point one billion dollars of reported instructional spending, which is more than Harvard, Yale and Princeton combined. Roughly one point two billion of that was patient-care expense at the medical centre, counted as instruction.⁵⁵

The gaps he found were not small. Columbia reported that 82 percent of its classes had fewer than twenty students. His reconstruction from the public course directory put that figure between 63 and 67 percent.⁵⁶ Columbia told the magazine that 96 percent of its faculty worked full time. Its own filing to the federal government implied 74 percent.⁵⁷ These are discrepancies between numbers sent to different addresses, not a court finding, and the university contested parts of the account. But hold on to what a ranking is supposed to be. It is a mirror held up to an institution after the fact. The institution does the teaching, and the number arrives later and describes it.

Something has changed about what is being governed, and it changed long before any magazine ranked anything. The prison and the timetable in the last two parts work on bodies in a room. Somebody is present, somebody is watched, somebody is corrected. The things a modern administration actually worries about will not fit in a room. Birth rates. Epidemics. The price of grain, and how many people leave the countryside in a decade. None of those is a deed that anyone commits, so none of them can be forbidden. There is no order you can issue to a birth rate.

Foucault's name for that object is population, and he means something narrow by it. A population is not a pile of individuals. It is a thing with regularities of its own, which appear only at scale and vanish when you look at any one person. No single family has a birth rate. The country does. And an object like that has to be counted before it can be governed. Which is why the modern state and the statistical office grow up together. The counting brings the object into view rather than reporting on one already there. Which is the hinge for everything that follows. A centre that governs by numbers is not being lazy. Numbers are the only form in which its object exists for it at all. Which is what is wrong with the mirror.

That picture of measurement is the thing this part has to break, and the reason it breaks is boring rather than sinister. A ministry cannot stand in ten thousand classrooms, and a magazine cannot sit in on a seminar. So a centre that wants to govern places it cannot visit has to convert them into something portable. Forms, standards, returns, published tables. Two sociologists writing in 1992 described power that works this way as acting at a distance. The chain runs like this. The centre defines a trace. The trace makes distant places comparable, and comparable places can be ranked. People working in those places know they are being rendered comparable, and they reorganise what they do toward the trace. No order is issued at any point in that chain.⁵⁸ The registrar filling in the form experiences it as doing the job properly.

Take the strongest case for the defence first, because the honest version of this argument starts with the indicator working. From 2000, the English health service ran national targets with published hospital data and real consequences for managers who missed them. The share of accident and emergency patients waiting more than four hours fell from 23 percent in 2002 to about five percent by 2004.⁵⁹ Patients waiting more than twelve months for planned surgery fell from sixty-seven thousand in 1998 to twenty-four by 2005. Twenty-four people, in the whole of England.⁶⁰ That is not a rounding error in a graph, and nobody who worked in that service before 2000 remembers it as a golden age.

Now the same literature, on the same regime. Patients were held in ambulances outside the doors of the department, the vehicle parked in the bay with its doors shut. They waited there until the staff inside were confident the clock could be beaten. The four-hour clock starts when the patient crosses the threshold, so the patient does not cross the threshold.⁶¹ Where you start the clock has been a question about historians so far in this video, a matter of which memo you begin the story with. Here it is a physical door, and a sick person on the wrong side of it. Extra staff were drafted in on measurement days. The health department reported 96 percent compliance with the four-hour target in one year, against 77 percent in an independent survey of patients.⁶² The national auditors found nine hospital trusts that had inappropriately adjusted their waiting lists.⁶³

The tempting reading is that the shut ambulance door cancels the fall in waiting times. It does not. The maximum wait genuinely came down. Ambulance response times in England pulled away from Wales, which had no such regime, over the same years.⁶⁴ The indicator did what it was built to do, and it also produced a shut ambulance door, and both of those are the same event. An indicator that only failed would be incompetence, and somebody would eventually switch it off. It is because it succeeds that it can reshape a hospital around itself.

The ranking story ends like this. Columbia sat tied for second in the country, with Harvard and MIT.⁶⁵ A mathematician in its own mathematics department rebuilt the numbers from the course directory. The university declined to submit data. In September 2022 it appeared eighteenth.⁶⁶

An indicator only does half the work. The other half needs somebody at the far end who responds to it, reliably, without being told to. Foucault spent his 1979 lectures describing that person, and the description is the part of this that reaches your own week.

The older economic picture had a man of interest in it. He trades, he has preferences, and the job of government is to leave those preferences alone. The version Foucault found in the economics of his own day had rewritten him. He is an entrepreneur of himself. His own capital, his own producer, the source of his own earnings.⁶⁷ Which changes what a wage is. It stops being the price of your hours. It becomes the income on a stock you have been building since childhood. Schooling, health, and the time somebody spent with you before you can remember it.

Then the sentence that matters for a video about power. A subject defined that way is, in his phrase, eminently governable. The reason sits inside the definition. He is someone who responds in a systematic way when the variables around him move. So a centre holding a subject like that has no need to command anybody. It changes a price, a risk, a default, or what you are told before you choose. Rational response does the rest, and it is done by you, and it feels like sense. The order never comes, because on this account there was never going to be one.

Here is the objection that should be worrying you by now. If a target that works and a target that is gamed both count as governing through freedom, what would ever count against the theory. A regime tightens and the account says the centre is steering conduct. A regime loosens and the account says the centre is governing through autonomy. Any outcome confirms it, which means it forecasts nothing, and a claim compatible with everything has told you nothing about anything. That objection lands against the slogan, and the slogan deserves it.

What it does not land against is the specific claim, because the specific claim is about a particular instrument and could have come out otherwise. The reported class sizes could have matched the course directory, and they did not. The four-hour target could have left the queue exactly where it was, and it did not. It moved the queue outside the building. That is the live disagreement in the field, stated plainly. One side holds that audit inspects performance that was already there. An accounting scholar writing in the nineties argued the opposite. Audit makes organisations auditable, rebuilding how they work until the work throws off traces a checker can verify.⁶⁸ The hospital did not get better at emergency care and then get measured for it. It moved where it kept the patient so the measurement would come out. Those two predict different things, and you can go and look.

And there is a trap in the going and looking. The only available defence against a rigged number was another number. A federal filing, a course directory, a consolidated financial statement, patiently reconciled by a man who does this for a living. There was no other move on the board. Critique of quantification proceeds by quantification. Every check you can actually run on a metric is itself a metric. So the number that reports on your work is also the thing you now work toward. And your only appeal against it is another number. You would have to assemble that one yourself, out of whatever filings and directories your employer already publishes. The win consisted of moving a university from second place to eighteenth on the same list, by the same magazine.

Still, that win took a federal disclosure regime, a published table, a professional body of auditors and an institution with a name to lose. There was somewhere to send the document. Now put the same mechanism in your pocket, and take all of that away.

It is eleven forty-seven at night, and the phone is face down on the arm of the chair. Nobody ordered you to pick it up. There is no rule about it, no supervisor, no penalty for leaving it there until morning. You pick it up. The strange part is not that you did. It is that the doing of it was sold in advance, to a buyer who has never heard of your name.

Every older picture of power has somebody forbidding something. A sovereign, a statute, a foreman, a gate that locks at six. That picture cannot see the chair at eleven forty-seven, because nothing there was forbidden and nothing was commanded. So either power ended in that room, or it changed address. Byung-Chul Han took the second option in 2017, and the argument is worth laying out slowly.

Prohibition tells a person what they may not do, and in doing so it hands them a target. You know the rule, you know who wrote it, and refusal has somewhere to point. Permissive power does not forbid. It offers: optimise, realise yourself, become the best version of the thing you already are. A subject who takes achievement as its own project supplies the pressure itself. So the demand no longer arrives in the second person, as an order. It arrives in the first person, as I want this, as I still have an hour tonight.

Coercion supplied in the first person meets no resistance, because resisting it would mean resisting yourself. That is the whole engine. The permissive arrangement is therefore more efficient than the prohibitive one rather than gentler. Freedom stops being the limit of power and becomes the material power runs on.⁶⁹ The last part showed the governmental version of this, where the variables around a choice get adjusted and rational response does the rest. What is new here is that the adjusting and the responding happen inside one person.

That argument has a popular version, and the popular version is the one that travels. In it, the feed rewrites what people believe, and the platform installs the opinions. Now the part that cannot be defended, which is exactly that popular version. The claim that the feed changes what you think is not supported by the best test of it. In 2023, a team led by Guess published the experiment in Science. Consenting Facebook and Instagram users were moved onto a plain reverse-chronological feed for three months around the 2020 election. Time on the platforms fell substantially, and so did activity. Issue polarisation did not move.⁷⁰ Affective polarisation, meaning how much people loathe the other side, did not move. Political knowledge did not move. A companion study stripped reshares out altogether, which cut content from untrustworthy sources by about 30 percent, and shifted no opinions either.

So take the thesis off belief and set it on what was actually measured. Exposure, attention, time, defaults, the option set, market access, and rules. In 2000, Lawrence Lessig wrote that conduct is regulated by four things, law, norms, markets, and architecture, and that online the architecture is code.⁷¹ Code is written by employed people with deadlines. His sharpest consequence survives every deflation of every mind-control story. The absence of public regulation is not the absence of regulation, and it means the people writing the code set the rules instead.⁷²

On 17 December 2012, Instagram published revised terms. The advertising clause was widely read as permitting user photographs in paid advertising, with no payment and no notice. The outcry took a day. On 20 December, a co-founder announced that the advertising section would revert to the version in effect since the service launched in October 2010.⁷³ Roughly a hundred million people had a rule written over them and then unwritten, inside seventy-two hours. None of them voted on either version. The terms were legislation, adopted by continued use.

Exit worked that week because switching was still cheap, and it has not worked like that since. That is the deflationary critic's entire point. Cory Doctorow argues the problem is monopoly rather than mind control, and that the persuasion story is the advertising industry's own sales pitch. Against him, Shoshana Zuboff had already described, back in 2015, markets in the prediction and modification of behaviour. That is a trade in what you do rather than in what you believe, and it is indifferent to your beliefs while doing it.⁷⁴ Notice what the deflation does to that fight. It removes the least defensible rival, and leaves the architectural claim standing alone.

Change the ranking and you change what hundreds of millions of people see. You still do not change what they believe. Exposure moved and belief did not, and the power is in the half that moved.

The strongest evidence against a puppet-master comes from inside the industry. On 11 January 2018, Facebook re-weighted its feed toward interactions, comments and reshares and reactions, and away from public content. The company's own researchers later recorded that the approach had unhealthy side effects on important slices of public content, such as politics and news. They also recorded that misinformation, toxicity and violent content were inordinately prevalent among reshares.⁷⁵ Those documents left the building in 2021. Nobody at any desk had written the words make politics angrier. A metric was chosen, and a world followed, including for the firm that chose the metric.

Then the objection that actually hurts the volunteering thesis. We exploit ourselves describes a salaried professional with a laptop and some discretion over her own evening. It does not describe the warehouse worker whose scanner counts her idle seconds. It does not describe the courier deactivated by an automated system, with no hearing and no appeal. Those people have a very visible enforcer, and they can name him. Han takes one niche of the labour market and calls it an epoch, and it happens to be the niche his readers live in. There is a repair available, and it costs him a word. In both cases the rule arrives as a system parameter rather than as a person. Which makes self-exploitation one special case of power that issues no orders, and not the general form of it.

Which does something to the tower from earlier, the inspection house that works whether or not anyone is inside it. The version you live in has no isolated cell and frequently no watcher at all, because the occupant is connected, publishing, and doing the illuminating himself. The effect is measurable even when nobody is watching in particular. That is what the Wikipedia readers of June 2013 did, unaddressed and unpunished. They simply learned the room was lit, and they lit less of it themselves, at once and then durably.

Exposure is theirs. Belief is still yours. Which leaves the thing you actually feel at eleven forty-seven, sitting up with a lit screen and no order to disobey. That has no name at all.

A card in a shop window, propped up between the onions and the carrots. It carries an official slogan, in the official phrasing, the same one hanging in every other shop on the street.⁷⁶ The grocer set it there before opening, the way he sets out the crates. Nobody delivered it to him this morning. No inspector is coming. If he left it in the back room, no one would arrive at noon to ask why. He puts it up anyway, and tomorrow he will put it up again.

The scene comes from a Czechoslovak dissident's essay, and it earns its place in the philosophy of power because of what fails to happen in it. Ask what was done to the grocer today and the honest answer is nothing at all. No order, no threat, no penalty, no interference of any kind. And yet nobody wants to describe his morning as a free man's morning. The trouble is that the ordinary vocabulary of freedom has no line to write that on.

Take that vocabulary at its strongest first, because it is the one almost everyone reasons with. On this account, you are free to the extent that nobody interferes with you. Freedom is measured in acts done to you and doors closed in your face, and where there are no acts, there is no unfreedom to count. It is clean, it is checkable, and it refuses to moralise, which is a large part of why it won.

That vocabulary meets its problem with two slaves. One has a master who beats him weekly. The other has a master who is indulgent, who has never raised a hand, who mostly lets him be. Acts of interference against the second man are close to zero, so by this measure he is the freer of the two. Worse, his freedom now rises and falls with his master's temper, going up on the days the man is in good humour. Frederick Douglass had already refused the whole framing: it was slavery, he wrote, and not its mere incidents, that he hated.⁷⁷

If you reject that conclusion, you have to reject a premise, and only one is available. The premise was that freedom varies with what is actually done to you. What separates those two men is a standing capacity to interfere at will, held by someone, uncontrolled by the person it hangs over. That capacity is what the neo-republicans call domination, and everything turns on its being a matter of what could happen. Which means it is fully present on a Tuesday when nothing whatever happens.

Then comes the case that should sting. Suppose the slave gets good at his position. He studies the man, learns which mornings to stay out of the room, anticipates the order and has it done before it is given. Interference against him drops almost to nothing. On the account we started with, he has become freer, and he became freer by becoming more servile. Nobody believes that. The pre-emptive conforming is the shape the injury takes.

Philip Pettit gives the test that catches this. You are adequately free, he says, when you can look others in the eye without reason for fear or deference. Not fear that he will hurt you. Fear of what he is free to do and you cannot stop. The test is whether you can look them in the eye without reason for deference. Not whether they have ever done anything to you.

Two things turn a power over you into something other than domination. The first is whether the power over you is reliably constrained by rules that actually bind, and that are common knowledge to everyone in the relation. Rules the other party knows you can hold him to. The second comes from a small example of Pettit's. Andrea wants to drink less, so she hands Bob the key to her alcohol cupboard. The terms she writes bind him on when he may give it back. Bob now has real power to frustrate her, and Andrea is not dominated for a moment, because she wrote the terms of his power. The live question is whether you have any control over the power that has you.

That is what replaces the list you can recite. The funds, the donors, the committee chairs: the list can always be extended, and extending it has never once changed a Tuesday. A relation, unlike a name, can be pointed at. It has a clause number, a ranking rule, a visa condition, a notice period. And the inspection tower from earlier finally stops being about eyes. What made that building unfree was never that a man was up there watching. It was that one could be, at any moment, and the people below had no say in whether he was.

Elizabeth Anderson runs all of this at the place you spend your weekdays. Government, she says, exists wherever some have the authority to issue orders to others, backed by sanctions, in one or more domains of life. That describes your employer exactly.⁷⁸ The authority reaches your speech, your body, your searchable belongings, who you sleep with, what you post on a Sunday. There is no right of appeal, the governors are not elected by the governed, and nothing limits which domains of life the authority may enter. We would call that a dictatorship in any other setting, and here we call it nothing at all. Anderson's explanation for the silence is historical. The language of free markets was built for a world of small proprietors. A genuinely free market, on that expectation, would leave most men working for themselves. Industrialisation destroyed that world and the vocabulary outlived it. So the largest authority relation in your life has no name in the words you use for it.

A playground bully is physically able to beat every child in the yard. He only ever beats some of them. Are the untouched children dominated? Something in us balks, and the pure capacity account says that they all are. The repair on offer is that a capacity counts only where appropriate penalties are absent, and that what matters is social possibility rather than physical possibility. The yard sorts itself out under that rule. The child with an older brother in the year above has a penalty standing behind him. So does the child with a teacher who would act. Neither of them is dominated. The child with nobody has no penalty standing anywhere, and he is dominated on the days nothing happens to him. The balking is what the account corrects, because it was counting bruises.

The deeper objection has no repair yet. The orthodox position still needs a dominator: structures facilitate domination, and only agents dominate. The critics answer that patriarchy, slavery and apartheid were impersonal, and dominated through unequal standing before laws, norms and customs rather than through anyone's decision. The sharpest recent charge is that neo-republicanism lacks an adequate conception of structural domination. In the labour-republican tradition that charge comes from, what dominates a worker is not this boss but the requirement of having one.⁷⁹ One line of reply is that exit plus a basic income would settle it. But if quitting means joining another government of the same design, a right to leave is not a right to be free of the relation. That argument is open right now, and it is the same frontier this whole video has been standing on.

What you have gained is real, and it is smaller than what you wanted. You now have a relation you can point at instead of a name you cannot find. That tells you what is wrong with your Tuesday. The whole account now rests on a capacity being fully real in the hours when nobody uses it.

Two colleagues in the same building, on the same salary band. Same title, same access to the same files, same standing invitation to the same meeting. One of them can get a decision reversed by Thursday. The other writes the better memo and cannot get a call returned. Nobody in the building thinks these two are equally powerful, and nobody can say what the difference consists of.

A nine-page article in a British political science journal, 1972. Its author had read the New Haven study. He noted, in passing, that the celebrated study of who governs an American city almost totally ignores its Black residents. His name is Peter Morriss. Fifteen years later he published a book, and the book said what the 1972 article had only circled. Both sides in the fight over who has power had been arguing about the wrong thing. The whole quarrel was a dispute between two bad measurements.

Before the charge, there is a crack in this video that a careful listener has already found. An hour ago you were told that power exists only when it is put into action. It has no three in the morning, and it is not the sort of thing anyone can have. In the last hour you were told that what is wrong with the grocer's morning is a standing capacity to interfere. That capacity is fully present on a Tuesday when nothing whatever happens. Those two sentences cannot both be true of the same word. Either the thing is there while unused or it is not, and this video has now insisted on both, an hour apart, at full volume.

Morriss goes back before the fight, to a question neither camp asked. Not when is power exercised. What is it to have power at all. The first premise is a fact about the word. To call someone powerful is to say they could bring outcomes about, which is a claim about what would happen, not a report of what did. The second premise follows from the first. A capacity can be held for thirty years and never once used, and it is held the whole time. So no enumeration of exercises settles who has power. The researchers counting decisions were measuring exercises and writing up capacities, and Morriss calls that the exercise fallacy.

The obvious repair is to count what people hold instead. The money, the votes, the office, the shares, the seat on the committee. That is the honest instinct behind every list of names, and it fails for a reason Morriss makes as plain as anything in this video. It is a mistake, he says, to identify whisky's power to intoxicate with its alcohol content. Whether you are drunk depends on what you have eaten, on your tolerance, and on your genes. The same forty percent goes into two people and comes out as two different evenings. The alcohol is the vehicle. It is not the power.

Put the two colleagues back beside each other and the point stops being about drink. No inventory of what either of them holds will ever show the difference, because the difference is not among the things they hold. Every method that scores power by counting resources is counting alcohol content. They have spent years counting alcohol content.

And now the crack closes. A capacity is not a stock in a vault, sitting there waiting to be spent. It is an arrangement in working order. The inspection house is a capacity that nobody is exercising, and the corridor below it runs the same on the day the tower is empty. Power exists only in the running, and it is fully real on the morning when nobody runs it. Those two claims stop fighting the moment you stop picturing a substance. The whisky is in working order in the bottle, and what it does depends on the evening it meets. One colleague is an arrangement that reverses decisions by Thursday, and the other is not, on a week when neither of them tries.

The repair is not agreed. In 2008 Keith Dowding published a paper against it, titled The Fallacy of the Vehicle Fallacy. The title is the objection, and the objection has not been settled. What has moved is the other flank. In the second edition of his book, in 2005, Steven Lukes adopted the dispositional analysis himself. He treats power there as a potentiality rather than an actuality, and narrows his subject to domination. He did not retract the third face, the power that shapes what you want before you want it. Do not hear this as him giving up his own position. But the man who argued for thirty years that the deepest power leaves no visible act now agrees. On his own account now, power is a capacity that a person holds whether or not the day contains an act.

You met Dowding's other objection in the third part, at the steel company that never lifted a finger. Getting what you want without acting is luck, not power. His name for it is systematic luck. Everybody else's interests stay arranged well enough to deliver your preferences year after year while you do nothing. Here is the part you were not told then.

Dowding concedes one thing, and the concession is where the argument turns back. Systematic luck can itself be the product of power, when the incentives that produce it were deliberately built. Push that one step past where he leaves it, and the same holds where somebody is actively keeping those incentives in place. Either way the question turns into an empirical one with dates in it. Who wrote the rule that makes everyone's interests run through that payroll, in what year, and who has kept it from being rewritten since. That is not a question about anybody's inner life, and it is answerable from the files. Where nobody built the arrangement and nobody is holding it in place, Dowding's charge stands and the beneficiary is lucky rather than powerful. This video does not get those cases back. What the last two hours recover is the narrower set, the arrangements somebody built and somebody is still maintaining.

Follow the repair to its end and something larger goes. If power is a capacity, and the resources are only vehicles, then the object of criticism may not be a holder at all. Clarissa Hayward's proposal is to stop asking who wields power over whom. Ask instead how the limits of the possible get drawn, and who is placed to affect them. A cleaner cannot walk into the boardroom, and a chief executive cannot tell a shareholder meeting that next quarter's numbers will be bad. Power, on her account, is the network that draws those limits: laws, rules, customs, and what a person in that role is allowed to be. That network sets the field of what can be done, and it does that for the powerful as well. Freedom then becomes the capacity to take an effective part in shaping those limits, which people have in wildly different amounts and nobody has all of.

That buys a great deal of this video. It explains why replacing the occupant changes nothing, because the boundaries were never his to take with him when he cleared his desk. It explains why the man who has never once refused you anything is still, in the exact sense that matters, your boss. And it explains the list. The funds, the donors, the committee chairs: an inventory of vehicles, assembled by someone looking for a holder. In this field the holder is not the unit.

Then the bill arrives, and it is not payable. Remove the face from power and you have removed the defendant with it. That is the standing objection from Lukes to any account built this way. The two of them published an exchange in 2008 on structure, agency and where responsibility lands. Its title is the whole problem in three words: nobody to shoot. A conception of power that indicts no one licenses nothing.⁸⁰ This video has now spent an hour proving that nobody chose your Tuesday. You can now describe the wrong done to you in full, and the description contains no one. So the wrong you can finally name is a capacity nobody exercised, inside boundaries nobody drew. Which means the person you would send the bill to has, by your own account, done nothing.

Sandy works at a shopping mall and raises two children on what it pays. A developer buys her apartment building and begins converting it into condominiums. She is evicted, legally, with notice. The apartments near her job cost more than she earns. The neighbourhoods she can afford she judges unsafe for her kids. So she looks at the suburbs, where the rents work, and finds that the suburbs require a car. She buys the car. Then the suburban landlord asks for three months of rent as a deposit, and the car has taken the money. Sandy is weeks from homelessness, and she has broken no rule and met nobody who broke one.

That case opens Iris Marion Young's Responsibility for Justice, and she built it to be hard. Find the person who did this to her. The developer bought a building and filed permits. The landlords charged what the market gave them. Follow the case one step past where Young leaves it and you reach a zoning board, voting decades ago on what shape the suburbs would be. Those people are retired or dead. Every actor did something ordinary, permitted, and in most cases boring.

Ordinary responsibility is a tracing procedure, and it asks three things. Did this agent cause the harm. Did they intend it, or at least foresee it. Did they act knowingly and freely. That procedure is not stupid, and it is what courts and consciences both run on. It fails here for a reason that has nothing to do with how hard you look. Structural injustice is cumulative, so no single action is a sufficient cause of it. It comes from people pursuing normal goals inside accepted rules. And the contributors cannot know in advance what their contribution adds up to.

Young is careful about what this does not say. Discriminatory landlords exist, corrupt politicians exist, dodgy developers exist, and all of them ought to be held accountable. Her point is that convicting every one of them would not move a single one of Sandy's constraints. And there is a cost to trying. Liability applied to a structure picks out a few salient wrongdoers. Naming them quietly clears everybody else, and leaves the background exactly as it was. There are villains, and they are not what is holding Sandy in place.

Which produces the objection this part exists to answer. If nobody is blameworthy, then surely nobody owes anything, and the whole analysis ends in a shrug. That inference only looks valid because one English word, responsibility, is doing two jobs. One job is asking whether an agent can be blamed for what happened. The other is asking whether an agent can be handed a burden of repair going forward. A defender of Young's separated the two cleanly in 2019. Blame needs a tracing story, and here there is none. The burden of repair does not need one. It needs only a position, and a position is something anyone can check.

And positions can be read off. Young's allocation runs on four questions, each with an address. What position are you in to affect this structure. What do you actually get out of it. What stake do you have in it changing. And who are you already able to act with. Answer those four and responsibility comes out graded. A shopper who buys a cheap shirt and the corporation that invented fast fashion are both connected to the same factory. On all four questions they score nothing like the same.

Now the harder case, where beneficiaries exist and no chooser does. A government drafts a tax that business dislikes. No executive telephones anyone, no donation changes hands, no meeting is held. The currency slips, two firms announce they are pausing investment decisions, the revenue forecast is revised down, and the proposal is quietly shelved. On any count of who lobbied and who voted, nothing happened. The version with a paper trail is duller and larger. A rule written for one economy is simply left alone while the economy moves under it. Distributions shift, no new law passes, and a blocking coalition holds. Say plainly what that is. That is a mechanism doing the work of an intention.

Here is what a mechanism looks like when somebody finally opens it. A commercial algorithm used across American hospitals, running on millions of patients, decides who gets flagged for extra care. It never takes race as an input. In 2019 a team led by Ziad Obermeyer published what it does. Correcting the bias would raise the share of Black patients flagged for extra help from seventeen point seven percent to forty-six point five percent.⁸¹ Somebody had to pick what the model would predict, and picked cost. Cost stands in for illness only if equally sick people get equally expensive care. Less money gets spent on Black patients who are just as sick. So the model reads them as healthier and passes over them. Nobody decided that outcome. Somebody chose what the machine would predict, and chose cost. That choice has a date and a desk.

Which is the shape this video keeps finding. No owner, many authors. The prison timetable had a hand that wrote it, and so does the target variable in a health model. The difference is that a target variable can be rewritten on a Tuesday by whoever owns the specification.

There is a second address, and it is not a person. Take three engineers voting on whether a bridge is safe. Safe here means two things, that the design holds and that the materials hold.⁸² Two of the three engineers say the design holds. A different two say the materials hold. Only one of them says both. Vote the questions one at a time and the committee affirms the design, affirms the materials, and denies the bridge is safe. Not one engineer holds that combination. You could read all three minds and still not know what the committee holds. Groups form positions of their own, and that is what makes the next argument bite. If the group can hold a position no member holds, the group is the thing that held it. And it is the thing that can be answerable for it. When a company ships contaminated food, its employees usually have real excuses. Badly informed, following established precedent, working under constraining expectations, afraid of losing the job. Often we should accept those excuses. And if the company itself is not on the hook, nobody is, and incorporating becomes a loophole for ends that harm others and reward you.

Nussbaum's charge is that this hands out a free pass. A responsibility that is permanently forward-looking is a responsibility that can never be failed. Every morning the ledger resets and the task lies ahead again. So everybody in the chain that put Sandy in that car has discharged their responsibility today. Tomorrow they will discharge it again, and she is still weeks from homelessness. Nothing goes onto the debit side, ever. A costless responsibility is an attitude, and a flattering one, since it lets comfortable people feel implicated without ever being indicted. That puts a fork in front of you, and it is a real one. Either somebody can fail at this, in which case you can name a person and say what the failure was. Or nobody can, and the four questions are a mood you carry around. Two repairs try to keep the first branch open. One puts guilt back for the failure to act, so something finally hits the ledger. The other says the ledger was always the wrong picture, because nobody gets a clean slate while standing inside many structures at once. The first repair abandons Young's insistence that this be a different conception of responsibility altogether. What the four questions buy you is an address to send the bill to. What you do not get, yet, is any way to say the bill went unpaid.

And the sharpest counter goes after the premise itself. Two philosophers of justice argue that structural injustice is not the sum of morally benign decisions by diverse agents. It is powerful actors engaging in subordination, exploitation and exclusion. If they are right, the whole apparatus is a sophisticated way of losing the defendant. The concession that has to be made here is real. Some injustice is avoidable, meaning powerful agents could change it and do not. Some is deliberate, meaning they work to keep it. There, moral responsibility comes back and lands on them. Structural explanation does not license benign intentions all the way down.

The guardrail was written in 1887, in a letter almost nobody quotes accurately. Acton told a bishop he could not accept a rule that judged popes and kings with a favourable presumption that they did no wrong. His words: if there is any presumption it is the other way against holders of power, increasing as the power increases.⁸³ Then came the famous sentence about power tending to corrupt. The line now used to excuse office-holders was written to abolish the excuse. The presumption runs against them, and it rises as the power rises.

Which leaves the clerk. His defence is that if he had not done it, somebody else would have. As a description of the office, that is usually true. Arendt does not argue with the description. She asks why, then, he became one, and why he remained one. That question moves the register from the system to the person sitting in it, and no structural fact answers it for him.

One thing here is genuinely unsettled, and it matters for Sandy. Young assigns political responsibility to victims as well, because they know most about the harm and are best placed to name it as injustice. Whether that blames them depends on whether political responsibility is a species of moral responsibility. Her own text supports both readings, and she never closed it. So responsibility can be allocated, by position rather than by tracing. That is a claim about who ought to act. Whether anything moves when they do is a different question, and that one has data.

August 1968. Soviet tanks entered Prague against a population with nothing to shoot back with. The occupier had total superiority in the means of violence, and could not rule the country. Street signs came down overnight so the columns could not navigate. Officials willing to sign an order in the occupier's name could not be found.⁸⁴ Hannah Arendt read this as a confrontation between violence and power in their pure states, and she meant those two words to name different things.

On the ordinary picture, an armoured division is the maximum quantity of power available to anyone. That picture makes Prague unintelligible. An army that owns every road and cannot get one form signed is not holding a small amount of something. It is holding a large amount of something else. So here is the distinction, and it is the counter-concept this whole video has been circling.

Power, Arendt says, corresponds to the human ability not just to act, but to act in concert. It is never the property of one person. It belongs to a group, and it lasts exactly as long as the group holds together.⁸⁵ When we say a man is in power, what we are describing is that a number of people have empowered him to act in their name. Power springs up whenever people get together and act in concert, and it disappears the moment they scatter. That is a claim about the room you sat in this morning as much as about Prague. Whatever authority the person running that meeting had, it lasted because the rest of you kept turning up.

Violence works the other way, because it is instrumental. Its implements are designed to multiply one person's strength until they can stand in for numbers. That is why a single man with a rifle can hold a street he could not hold with his hands. That produces the asymmetry the whole argument turns on. The extreme form of power is all against one. The extreme form of violence is one against all, and that is never possible without instruments.

Now the step that makes this more than a matter of vocabulary. Take any apparatus of violence: an army, a police force, a security service. Each is a chain of command, and its effectiveness consists entirely in orders being carried out by the men below. That carrying out is not itself produced by the order. It rests on the willingness of the ranks, which rests on what they think and how many of them think it. Where commands are no longer obeyed, the means of violence are of no use. So violence can destroy power and is incapable of creating it, and a regime escalating its repression is publishing a measurement of its own lost support.⁸⁶

This is where the phrase that has been running through this video changes its job. No owner, many authors began as a fence against nihilism, and at the last turn it became an address for the harm. Here it becomes the strategy. An arrangement nobody owns still runs on hands, and every one of those hands is attached to a person with a name and a shift pattern. The prison officer, the tax clerk, the sergeant at the barricade.

A French law student had already put it in one image, four hundred years earlier. The man who dominates you has two eyes, two hands, one body, and nothing more than any other man, except the power you confer on him. Where would he get the eyes to spy on you, if you did not lend him yours. So support him no longer. You will see him fall of his own weight and break in pieces, like a colossus whose pedestal has been pulled away.⁸⁷

That sounds sentimental until you read the rest of the essay, which is mostly an explanation of why it does not happen. Habit does part of it, and spectacle does part of it. But the load is carried by a tiered structure of beneficiaries. A handful around the tyrant, hundreds profiting under them, thousands under those, down through everyone who gets something out of the arrangement.⁸⁸ Withdrawal is a defection problem inside a patronage network.

Gene Sharp spent a career turning that pyramid into a list you could work from. He asked which institutions the ruling actually runs through, and the tradition that grew out of his work named them. The army and the police. The civil service, the courts and the newspapers. The business elites, the religious authorities and the schools. Every one of those is a payroll, and a payroll is a set of people who can stop.

Which gives the strategy its target, and the target is never the ruler. It is the personnel of those institutions, and the event you are waiting for is defection rather than persuasion. That is what makes Serbia legible instead of lucky. The strike at the Kolubara coal mine cut electricity generation, and the current ran through men who had stopped digging. The regime never ran out of people who supported it. It ran out of people willing to carry out an order, one layer down from the top.⁸⁹

Which leaves the question of what makes a man one layer down decide to stop. Sharp built half his theory around one answer, and he took the name for it off a wrestling mat. Political jiu-jitsu.

It is a claim about how force reads to the people watching it. Violence used on people who are shooting back looks like a fight, and a fight has two sides with a case each. The same violence used on people who are not shooting back reads differently, and it reads that way to the audience the regime can least afford. The conscript holding the rifle. The magistrate who has to sign the committal. The editor deciding what goes on the front page, and the government abroad deciding whether to keep selling something.

So the regime is in a bind with no clever way out of it. Leave the square alone, and the appearance of control goes. Clear the square, and the clearing becomes the movement's best argument, delivered free of charge to the exact people the whole apparatus is standing on.

The mechanism has conditions, and they are worth stating before anyone leans on it. Somebody has to see the repression. The people who see it have to have somewhere to move. And they have to be inside the arrangement that the regime actually depends on. Where any one of those fails, the same courage in the same street buys nothing at all.

Autumn 1956, Budapest. Within days of the rising, workers' councils appeared in the factories. Then revolutionary councils in the neighbourhoods, and students' councils, and writers' councils, and councils among the soldiers, federating upward into something that was, briefly, a government.⁹⁰ Nobody designed it and no party issued it. It came out of nothing but people getting together, which for Arendt is the demonstration that power is generated rather than seized. And it was gone the moment the tanks dispersed the assemblies.

Pair that honestly with the study from earlier in this video. In that prison experiment, once the participants' own egalitarian arrangement had come apart, the moves to impose a tyrannical regime met with weakening resistance. Acting in concert is real, it is not automatically stable, and people who have watched it fail will ask for a strongman.

Three checks, then, on everything just built. The first is evidential, and it helps. James Scott's point is that public deference is performed under surveillance and sanction, so it tells you nothing about belief. The private speech behind it is a running critique. Which means the defiance that looks sudden is the public declaration of something that already existed offstage.⁹¹

The second check does not help at all. On Bourdieu's account, compliance is not a decision in the first place, so it cannot be withdrawn by one. It sits in dispositions acquired before you had articulate beliefs, and it is misrecognised as simply the way things are.⁹² If that is right, resolve to serve no more is not an available act. The self doing the resolving is itself a product of the arrangement.

Then the psychological check, which cuts the other way and should be said out loud. The flagship finding here was that the worst-off defend the system hardest, which would be internalised domination with a number attached. It has the weakest replication record of anything in this video. Large cross-national tests found little or no trace of it. In some data the sense that the distribution is fair rises with advantage rather than falling. So the confident claim that the oppressed have internalised their oppression is currently underdetermined by the evidence.

The last check is empirical, and it is the hard one. Erica Chenoweth built the dataset that made all of this look like engineering. Nonviolent campaigns beat armed ones by roughly two to one. And no campaign in her record had ever reached three and a half percent of a population and then lost. Read the caveats she puts on that figure and it thins immediately. The three and a half percent is peak participation, not total sympathisers. It counts people actively and repeatedly in the streets. And it counts only maximalist campaigns, the ones trying to remove a head of state or expel an occupier, not the ones after a single concession. The record it describes runs to the mid-2000s.

And she is the one publishing the decline. From the 1960s, success rates for revolutionary nonviolent campaigns stayed above 40 percent, climbing as high as 65 percent in the 1990s. Since 2010, less than 34 percent of them have succeeded, and a mere eight percent of the violent ones.⁹³ Average peak participation has fallen from about two percent of a population in the 1980s to one point three percent now.⁹⁴ The colossus still falls when the people holding it up walk away. There are fewer of them walking, they are marching instead of striking, and the regimes have read the same literature you have.

She gives reasons for that, and two of them have teeth. The first is tactical. Movements over-rely on mass demonstrations and neglect the techniques that disrupt a regime's stability, general strikes and mass civil disobedience. A march shows the regime how many people are against it. A strike takes their work away from it. One is a message. The other is a cost. The second reason is that the regimes adapted, and in a specific way. They infiltrate a movement early to provoke militancy before it has built a broad base. And they work on their own security forces in advance, so that defection is harder to reach. The record moves with that. From the 1970s until 2010, between 30 and 35 percent of nonviolent movements carried a violent flank. In the decade after 2010 it was more than half.⁹⁵

The precision is contested, and it should be. One recoding splits the violent category apart, counting stones and defensive force separately from armed insurgency. Recoded that way, the celebrated advantage of pure nonviolence largely disappears. The reply is that almost every campaign mixes methods, so the coding is a judgement call either way. And violence in a crowd may be a symptom of momentum rather than a cause of victory. The direction survives that exchange. The decimal points do not.

There is also the question of when you stop counting. Mubarak resigned in February 2011, after eighteen days of protest on the standard account, once the army declined to fire. In a dataset of maximalist goals that is a success. By the summer of 2013 the outcome was a military coup and a harder autocracy. Same events, two answers, decided entirely by where the observation window closes. And whoever gets to close the window is doing something to the record, not just reading it.

Then a recent case the theory has to survive. Hong Kong, the summer of 2019, against a bill that would have allowed extradition to the mainland. On the sixteenth of June the organisers said two million people were in the streets. The police said a fraction of that. By any plausible count the movement went far past three and a half percent. The campaign lost outright. In June 2020 the National Security Law was imposed and the argument was over.⁹⁶

Set that against the list of pillars and it stops being a mystery. Hong Kong's own police could have walked off to the last officer and it would have changed nothing. The decision was not being taken in Hong Kong. The security apparatus that mattered sat outside the movement's reach entirely, which is the condition the consent theory has been resting on all along. Withdrawal works where the consent you are withdrawing is the consent that regime actually depends on. Where it is not, the count in the street stops meaning anything at all. That answer costs something, and it should be paid here. The theory only earns it by naming the apparatus that matters before a campaign starts, not after it has lost. In Hong Kong anybody could have named Beijing in advance, so this case is cheap. Can it do that generally. Not reliably, and an account that relocates the decisive pillar once the result is in is explaining failure after the fact.

One more limit, and it is the one nearest to your week. A peace researcher put it to Sharp in 1989, and he was arguing with the theory rather than against it. The model assumes a ruler on one side and subjects on the other, with a line between them. That is a fair map of a dictatorship. It is a poor map of an office, where almost everyone is somebody's superior and somebody's subordinate at the same hour. You sign off one person's leave and wait on another's signature for your own. There is no side of the line to defect from.

His second point is colder. The loci of cooperation are not automatically on freedom's side. A union that came out of a strike becomes an institution that disciplines its own members.⁹⁷ And the concession he made himself should be kept, because it is the honest shape of the thing. The theory is jargon-free, it is immediately actionable, and it works best exactly where the ruler and subject picture approximates reality. The sharpest instrument in this video was built for a shape your Tuesday does not have.

So here is what replaces the name you were looking for. Whose power over you is uncontrolled by you. Who benefits from this arrangement staying as it is. Who holds a position from which it could be changed. And what would change if enough people stopped cooperating at once. Those four are checkable, and none of them is a name.