The fundamental flaw of democracy is not that it gives power to the people. It is that it gives power to the people who want it.
This is not a new observation. Plato made it in the fourth century BCE, in Book VI of the Republic, through the mouth of Socrates, who compared the state to a ship and asked whether you would want the ship's course to be determined by a popular vote among the passengers or by a navigator who had spent years learning the stars, the tides, and the rocks. The passengers would choose the loudest voice, Socrates argued. The loudest voice would not be the most competent. It would be the most ambitious. Plato's solution — a philosopher-king, a trained guardian class insulated from wealth and family ties — was its own catastrophe, and I will get to that. But the diagnosis was precise. In any system where power is available to those who seek it, the seekers will be disproportionately those for whom power is the point. The navigator does not campaign. The navigator navigates.
I watched a version of this problem play out in a gorilla troop at Diergaarde Blijdorp, Rotterdam Zoo, during the period that started all of this — the same months, the same walks with my older son, the same slow rearrangement of my understanding that eventually became this book. The zoo houses a group of western lowland gorillas, and during the months I visited, the troop's silverback was an older male — calm, heavy, unremarkable to most visitors. He did not display. He did not charge the glass. He sat. When a juvenile screamed, he walked over. When two females argued over a resting spot, he positioned himself between them — not aggressively, just present — and the dispute dissolved. The visitors wanted spectacle. They wanted the chest-beating, the roaring, the demonstration of power they had seen in documentaries. The silverback gave them nothing. He was not performing authority. He was exercising it, which is a different thing entirely, and the difference is the subject of this chapter. We have, I think, confused the two. We reward the performance and ignore the exercise. And then we wonder why our leaders behave like performers.
In 1982, Frans de Waal published Chimpanzee Politics: Power and Sex among Apes, based on years of observation at the Royal Burgers' Zoo in Arnhem, the Netherlands. The book documented the power struggles among three adult males — Yeroen, Luit, and Nikkie — with a specificity that shocked readers who had assumed that politics was a human invention. The chimpanzees formed coalitions, broke them, reconciled after conflicts, withheld support strategically, and retaliated against perceived betrayals — not in the heat of the moment, but days later, when an opportunity presented itself. De Waal coined the phrase "alpha male" in this book, though the concept has since been so thoroughly vulgarised by popular culture that it bears restating what he actually meant by it.
An alpha male, in de Waal's observation, was not simply the strongest or most aggressive individual. The most stable alpha males — the ones whose tenure lasted — were coalition builders. They shared food. They intervened in disputes on behalf of the weaker party. They groomed widely, not just upward. They consoled losers of fights. They policed conflict rather than initiating it. The best alpha was, in de Waal's terminology, a "control role" — an individual whose authority derived not from fear but from the social services it provided to the group. Luit, during his period as alpha, was precisely this kind of leader. He mediated conflicts. He protected juveniles. He maintained order through presence rather than violence. The group was stable under Luit. Does this description remind you of any leader you have recently voted for? Or does it sound like someone who would never run for office?
Then Nikkie and Yeroen formed a counter-coalition. They challenged Luit. The power shifted. And on one night in 1980, Nikkie and Yeroen attacked Luit in a coordinated assault that left him with deep puncture wounds across his body, several fingers and toes bitten off, and his testicles torn out. He bled to death. The competent leader was killed by the ambitious ones. The parallel is not subtle.
De Waal spent the rest of his career — he died in 2024, having transformed the field of comparative psychology — documenting the qualities that predict stable leadership in primate groups. The pattern held across species, across settings, across decades of observation. The best leaders were not the most aggressive. They were the most socially intelligent. They built broad coalitions. They distributed resources widely. They de-escalated tension. They were, in the zoological sense, competent. The individuals who seized power through aggression held it briefly and violently, and the group suffered measurably under their tenure — higher cortisol levels, more injuries, less social grooming, reduced play among juveniles. The enclosure got worse when the wrong animal was in charge. Not because the wrong animal was evil. Because the traits that drive an individual to seize power — aggression, dominance-seeking, low empathy, willingness to form exploitative alliances — are precisely the traits that make an individual unsuitable for the role once it is obtained. The system selects for exactly the wrong qualities. What would happen if we designed it to select for the right ones?
This is the self-selection problem. The political scientist Brian Klaas, in his 2021 book Corruptible: Who Gets Power and How It Changes Us, quantified it for humans. Klaas reviewed the research on psychopathy and leadership and found that psychopaths — individuals who score above the clinical threshold of thirty on the Psychopathy Checklist — appear in the general population at a rate of approximately one in five hundred. In a study of aspiring corporate managers, the rate was one in twenty-five. A twenty-fold overrepresentation. The mechanism is not mysterious: the traits that define psychopathy — grandiosity, manipulativeness, superficial charm, absence of remorse, willingness to deceive — are also the traits that predict success in competitive hierarchical systems. The psychopath does not merely tolerate the campaign trail, the boardroom politics, the relentless self-promotion that modern leadership requires. The psychopath thrives in it. The system selects for the disorder. We built a selection mechanism that filters for the precise personality type we should be filtering against.
Klaas interviewed over five hundred leaders across the spectrum — presidents, philanthropists, rebels, cultists, dictators — and the finding that emerges from his work is not that power corrupts, though it does, but that power attracts the corruptible. The self-selection happens before the individual enters office, before the first vote is cast, before the campaign begins. The person who looks at the political process — the fundraising, the media management, the relentless public performance of conviction, the willingness to simplify complex realities into slogans — and thinks "yes, this is for me" is, statistically, not the person you want making decisions about your healthcare, your children's education, or whether the country goes to war. And yet here we are. This is how we select the keepers of our enclosure. Does that not strike you as extraordinary?
Every social species solves the governance problem. This is worth establishing, because the tendency in political thought is to treat governance as a uniquely human achievement — the product of philosophy, constitution-writing, and Enlightenment reasoning. It is not. Governance is the management of group resources and group conflict, and every species that lives in groups manages both, whether it has a prefrontal cortex or not.
Honeybees, as Thomas Seeley documented in Honeybee Democracy in 2010, make collective decisions about nest sites through a process that is, in structural terms, more democratic than any human system. When a swarm must choose a new hive location, several hundred experienced scout bees fly out independently to evaluate potential sites. Each scout assesses cavity volume, entrance size, sun exposure, and other quality indicators. She returns and performs a waggle dance — the same communication system Karl von Frisch decoded, which I described in the previous chapter — and the vigour and duration of her dance correspond to the quality of the site she has found. A scout who has found a mediocre site dances briefly. A scout who has found an excellent site dances vigorously. Other scouts visit the advertised sites, evaluate them independently, and return to dance for the ones they judge best. Over hours or days, the dances converge. No single bee decides. No bee campaigns. No bee deceives. The decision emerges from the aggregation of honest, independent assessments, and the swarm moves to the site that has attracted a critical threshold of support. Seeley described five principles of effective swarm decision-making: minimise the leader's influence, rely on open debate, seek diverse options, aggregate independent assessments, and use quorum thresholds. The bees follow all five. Most human democracies follow none. What does that tell us about the sophistication of our governance, measured against an insect's?
Elephant herds are led by matriarchs — the oldest female in the group — and the research on what makes them effective leaders is among the most instructive in the zoological literature. Karen McComb and colleagues at the University of Sussex published a study in the Proceedings of the Royal Society B in 2011 demonstrating that the age of the matriarch directly predicted the group's capacity to assess predatory threats. Groups led by older matriarchs — sixty years and above — responded more appropriately to the roars of male lions, which represent a greater threat than female lions. Groups led by younger matriarchs under-reacted to male roars, a potentially fatal miscalibration. During the severe drought in Tanzania's Tarangire National Park in 1993, elephant family groups that migrated out of the park to find alternative food and water sources had lower calf mortality — and these groups were disproportionately led by older matriarchs whose memory of previous droughts, decades earlier, guided the decision. The matriarch leads not because she is strongest, not because she campaigned, not because she defeated rivals. She leads because she has lived the longest and remembers the most. Her authority is earned through survival and validated through outcomes. No elephant has ever voted for a matriarch. No matriarch has ever sought the role. The role sought her.
And then there are wolves, whose governance mythology I need to correct because it has been used to justify human political hierarchies that bear no resemblance to anything wolves actually do. In 1970, L. David Mech published The Wolf: Ecology and Behavior of an Endangered Species, in which he described wolf packs as dominance hierarchies led by "alpha" males and females who maintained their position through aggression. The book was hugely influential. The "alpha wolf" concept entered popular culture and has been used ever since to naturalise authoritarian leadership — the idea that nature itself selects for dominance, that the strong should lead, that hierarchy is biological destiny. Mech spent the next three decades trying to undo what he had done. In 1999, he published a paper titled "Alpha Status, Dominance, and Division of Labor in Wolf Packs" in which he stated, with the clarity of a scientist correcting his own error, that the alpha concept did not apply to wolves in the wild. His original observations had been made on captive wolves — unrelated individuals thrown together in enclosures, like strangers in a prison yard. They formed dominance hierarchies because they had no existing social bonds. In the wild, Mech discovered through years of fieldwork on Ellesmere Island in the Canadian Arctic, wolf packs are families. The "alphas" are simply the parents. The "hierarchy" is simply the age structure of a family unit. The pack is led by the breeding pair not because they won a fight but because they are the mother and father. Mech asked the scientific community to stop using the term "alpha" entirely. The community complied. Popular culture did not. We prefer the myth. The myth of the dominant alpha flatters something in us — something that wants to believe that power should belong to the strongest. Nature disagrees.
The point is not that animals are better at governance than humans. The point is that across every social species studied — primates, insects, elephants, canids, cetaceans — the same pattern emerges: the best leaders are not the ones who sought the role. The best leaders are the most experienced, the most socially connected, the most competent at the specific task the group needs performed. Leadership, in nature, is a function. It is not an identity. The matriarch does not think of herself as a leader. The scout bee does not campaign. The gorilla silverback does not give speeches. They perform the role because the role needs performing, and the group recognises their competence through direct, long-term observation.
Human governance began the same way.
Christopher Boehm, an anthropologist at the University of Southern California, spent decades studying political organisation in small-scale human societies. His 1999 book Hierarchy in the Forest: The Evolution of Egalitarian Behavior presented a thesis that is, to anyone familiar with de Waal's chimpanzee work, immediately recognisable: human hunter-gatherer societies were egalitarian not because the humans in them lacked the desire to dominate, but because the group actively prevented domination. Boehm called this a "reverse dominance hierarchy" — a system in which the rank and file band together to suppress any individual who attempts to accumulate disproportionate power.
The mechanisms were consistent across the ethnographic record. If a hunter returned with a large kill and began to boast, he was mocked. If a skilled individual began to issue commands, he was ridiculed. If someone attempted to claim a disproportionate share of resources, the group withdrew cooperation — stopped sharing food, stopped assisting in tasks, stopped including the offender in social activities. In extreme cases, the transgressor was exiled or, in the most extreme cases documented in the ethnographic literature, killed. The anthropologist Richard Lee, working with the !Kung San of the Kalahari in the 1960s and 1970s, described what he called "the insulting of the meat" — a practice in which a successful hunter's kill was deliberately belittled by the group. "You mean you dragged us all the way out here for this thin bag of bones?" a companion might say of an excellent kill. The purpose was not cruelty. It was governance. The mockery prevented the accumulation of prestige that could, unchecked, become the accumulation of power. We had, for most of our history as a species, a built-in immune system against tyranny. What happened to it?
Leadership in these societies was situational. The best tracker led the hunt. The most experienced forager led the gathering expedition. The elder with the deepest knowledge of water sources led the migration. No single individual held authority across all domains, and no authority persisted beyond the situation that required it. When the hunt was over, the tracker became an equal member of the group. When the migration was complete, the elder sat down with everyone else. There was no permanent political class, no career in governance, no individual whose identity was "leader" rather than "member who is currently leading this specific activity because they are the most competent at it."
Boehm's insight was that this system did not emerge from the absence of hierarchy. It emerged from its active suppression. The desire to dominate is real — it is visible in chimpanzees, in children, in every playground and every boardroom. What the hunter-gatherer societies achieved was not the elimination of that desire but its containment through constant, collective vigilance. The group watched. The group noticed. The group responded. And the group could do this because it was small enough — one hundred, one hundred and fifty, two hundred people — for every member to observe every other member's behaviour directly.
The reverse dominance hierarchy worked because it operated within the Dunbar threshold. It worked because everyone knew everyone. And the question we must now ask is: what happens when that condition is removed?
And then, as with every system examined in this part of the book, the village became a town, and the town became a city, and the city became a civilisation, and the mechanism broke.
At band scale, governance is automatic. The animal knows the leader. The animal has watched the leader for years — in conflict, in scarcity, in grief, in celebration. The animal has seen the leader share food when food was scarce, intervene in disputes with fairness, make decisions that benefited the group rather than the decision-maker. The evaluation is direct, continuous, and based on years of embodied evidence. The animal does not need a campaign. The animal has been observing the campaign — the real one, the one measured in daily behaviour rather than speeches — for its entire life.
At civilisation scale, governance is abstract. The animal knows the leader's name. It has seen the leader's face on a screen. It has heard the leader's voice, processed through microphones, edited, curated, delivered in thirty-second segments optimised for the attentional architecture described in the previous chapter. The animal has never watched the leader resolve a conflict. It has never seen the leader share food with a rival. It has never observed the leader comfort a grieving member of the group. It has no embodied evidence of the leader's character — only mediated representations of a performance designed, by professionals, to simulate the qualities that the animal's neurology is calibrated to assess through direct observation. We evaluate our leaders the way a herring gull chick evaluates its parent — by the signal, not the substance. And the signal, as we learned in the previous chapter, can be faked.
I vote. I have voted in every general election since I was eligible. I do so because I believe the franchise matters — because the alternative to imperfect democracy is not better democracy but something far worse. But I want to describe what the act actually involves, stripped of the civic mythology. I walk to a building I have never entered for any other purpose. I stand in a queue with neighbours I do not know. I am handed a piece of paper listing the names of people I have never met. I have not watched any of these people navigate a crisis. I have not seen them with their families. I have not observed their behaviour under pressure, in private, over years. I have seen a photograph, a slogan, and — if I have been diligent — a manifesto written not by the candidate but by a team whose job is to make the candidate electable. I mark my X. I leave. The person I have chosen may serve for years, making decisions that affect my children's education, my healthcare, my safety, my shelter, my taxes, the conditions of war and peace — and I have selected them using less information than a chimpanzee has about the lowest-ranking member of its troop. This is our system. This is what we call democracy. And if that description makes you uncomfortable, I would ask: does the discomfort come from the description, or from recognising what we have been doing all along?
This is not an argument against democracy. It is an observation about the distance between what the animal evolved to do and what the animal is asked to do. In a group of one hundred and fifty, the organism selects leaders through the most rigorous evaluation method available in nature: years of direct behavioural observation. In a nation of sixty-seven million, the organism selects leaders through a process that would not pass the minimum standards of a zoo welfare assessment.
The self-selection problem compounds at scale. In a small group, anyone might lead — the role is temporary, situational, and available to whoever is most competent for the task at hand. There is no career path in governance because governance is not a career. It is a chore. Like cleaning the latrine or standing watch at night, it is performed because it must be performed, and the person performing it looks forward to the moment when someone else takes over.
Modern governance is not a chore. It is a profession. In the United Kingdom, the London School of Economics documented the steady rise of what it termed "career politicians" — individuals who entered politics directly from university, often through internships and research assistant positions, and who have never held a job outside the political sector. One in seven British MPs, by a 2012 analysis, had never worked in any role that was not directly connected to party politics. They moved from student politics to parliamentary assistantships to safe seats to ministerial positions, accumulating power without ever having been subject to the constraints — the accountability, the feedback, the consequences — of working within the systems they were elected to govern. They govern our hospitals without having waited in our waiting rooms. They govern our schools without having taught in our classrooms. They govern our economy without having lived on our wages.
The United States presents the same pattern in starker terms. An analysis by the Center for Responsive Politics found that in the 116th Congress, the majority of members were millionaires — a status shared by approximately one percent of the American population. The median net worth of a senator was approximately 1.7 million dollars. The median net worth of an American household was ninety-seven thousand three hundred dollars. The people writing tax policy were seventeen times wealthier than the people the tax policy affected. The people deciding healthcare legislation had access to healthcare plans that the general population could not obtain. The people determining education funding sent their children to schools that cost more per year than the median annual income of the families whose children attended the schools being funded. Does this arrangement sound like governance of the people, by the people, for the people? Or does it sound like something else entirely?
This is not corruption in the conventional sense. Most of these individuals were not breaking laws. The problem is structural: the system selects for individuals who have the resources, the connections, the social capital, and the personality type to navigate a process — fundraising, media management, party loyalty, relentless public performance — that bears no resemblance to the task the role was designed to perform. A gorilla troop does not select its silverback through a competition in which candidates perform chest-beating displays before an audience of spectators who have never met them. A gorilla troop selects its silverback through years of direct observation. The silverback emerges. The politician campaigns. These are not the same process, and they do not select for the same traits.
Brian Klaas, again: "We have designed systems that attract the wrong people to power, that test the wrong traits, and that provide the wrong feedback once the person is in charge." The self-selection problem is not a bug in the system. It is the system. The system was not designed to select for competence. It was designed — or rather, it evolved, without design — to select for ambition, charisma, fundraising ability, media performance, and the willingness to reduce complex policy to slogans. These are the traits the system tests. These are the traits it rewards. And these are the traits that predict, with remarkable consistency, the wrong kind of leader. We test for one thing. We need another. And then we are surprised when the leaders we select are excellent at campaigning and terrible at governing.
I want to hold the compassionate frame, because it would be easy, at this point, to slide into contempt — to treat democracy as a failure, to mock the voters, to despair at the politicians. This book does not do contempt. Every system examined in Part Two began as a good impulse. Money was an attempt to extend cooperation beyond the trust network. Justice was an attempt to protect the vulnerable from the powerful. Education was an attempt to transmit knowledge across generations. Media was an attempt to share information across the species. Each one worked, in some form, at village scale. Each one distorted at civilisation scale.
Governance follows the same arc.
The good impulse is ancient and visible across cultures. When a !Kung San band mocks its best hunter, the impulse is democratic: no one should have disproportionate power. When the Haudenosaunee (Iroquois) Confederacy established the Great Law of Peace — a governance structure that influenced the framers of the American Constitution — the impulse was federal: multiple communities governing themselves while coordinating on shared concerns. When the citizens of Athens, in the fifth century BCE, assigned most magistracies by lot rather than election, the impulse was egalitarian: sortition — random selection — prevented the wealthy and the eloquent from monopolising office. Aristotle, who was not a democrat, acknowledged the logic: election, he wrote, is oligarchic. The lot is democratic. The Athenians used complex allotment machines called kleroteria to select their council members, their magistrates, and their jurors randomly from the eligible citizen body. Most magistracies had one-year terms. No individual could serve on the council more than twice. The system was explicitly designed to prevent the formation of a professional political class.
The Athenians understood the self-selection problem twenty-four centuries before Brian Klaas named it. They built machines to solve it. We forgot.
But Athens excluded women, enslaved people, and foreigners — the majority of its population. The democratic impulse was genuine. The execution was monstrous. And this is the pattern, again and again: the impulse toward collective decision-making, toward distributed power, toward governance that serves the governed rather than the governor, is correct. It is biologically correct — it is how our species governed itself for two hundred thousand years. The execution fails not because the impulse is wrong but because the scale overwhelms the mechanism. Democracy at village scale, where every citizen knows the candidates, is powerful. Democracy at nation scale, where no citizen knows the candidates, is a performance. We feel the difference. We have always felt it. The vague dissatisfaction with politics that most of us carry — the sense that something is not working, that the system does not represent us, that the choices we are offered do not include the choice we actually want — is not cynicism. It is our neurology telling us that the mechanism has exceeded its operating parameters.
Monarchy tried to solve the scale problem by concentrating authority in a single lineage. It produced occasional competent rulers and frequent catastrophes, because hereditary succession is a lottery and the traits that make a good sovereign — wisdom, restraint, genuine concern for the welfare of millions one will never meet — are not genetically heritable. Oligarchy distributed power among a propertied elite, which solved the single-point-of-failure problem but introduced the class-interest problem: the governing class governed in its own interest. Democracy expanded participation to all citizens, which solved the class-interest problem in theory but reintroduced the self-selection problem: the citizens who seek office are not representative of the citizens who do not.
Each expansion — from monarchy to oligarchy to democracy — widened the circle of participation. Each was a genuine moral advance. Each inherited the structural problem of the one before it and added a new one. Democracy is the best system humans have tried, and I mean this without irony. It is also, when examined through the zoological lens, a system that asks the animal to do something its neurology is not designed for: evaluate strangers at a distance, using mediated information, in order to entrust them with decisions that affect every dimension of the animal's life. We are trying to do something our brains were not built to do. That is not a moral failing. It is a design mismatch.
If you wanted to find the closest modern approximation to small-group governance operating at a national scale, you would likely arrive in Switzerland. The Swiss political system is, by the standards of modern nation-states, unusual enough that most political scientists treat it as a category of one.
The Federal Council — the Swiss executive — consists of seven members elected by the Federal Assembly, each heading a federal department. The presidency rotates among the seven annually. The president has virtually no powers beyond those of the other six councillors. There is no prime minister. There is no head of state in the conventional sense. The executive is a committee. The idea that one person should embody the authority of the nation is, in the Swiss framework, structurally prevented.
Below the federal level, the twenty-six cantons retain extraordinary autonomy. Each has its own constitution, legislature, government, and courts. And below the cantonal level, direct democracy operates with a frequency that no other modern nation approaches. Swiss citizens vote on federal matters three to four times per year, on questions ranging from immigration policy to railway infrastructure to foreign treaties. Any citizen can challenge any law passed by parliament by collecting fifty thousand signatures within one hundred days, triggering a national referendum. Any citizen can propose a constitutional amendment by collecting one hundred thousand signatures within eighteen months, triggering a popular initiative. The population does not merely elect representatives who make decisions. The population makes decisions. The mechanism is not a supplement to representative democracy. It is its constraint. The parliament proposes. The people decide.
This is closer to the band-level governance that Boehm described — closer to the situational, participatory, directly accountable system in which the animal evolved — than anything else at national scale. The Swiss system assumes, structurally, that the citizens themselves are capable of governance. It does not delegate decision-making to a professional class and then ask the citizens, once every four or five years, to evaluate that class's performance through a mechanism — the election — that provides no meaningful information about competence. It places the decisions themselves before the citizens, regularly, on matters of substance.
And it works, in many respects, remarkably well. Switzerland is consistently rated among the most stable, prosperous, and well-functioning democracies in the world. Its infrastructure is maintained. Its services are delivered. Its economy performs. Its political transitions are unremarkable — and unremarkable political transitions are, in the context of human history, one of the most remarkable achievements a society can produce.
But the Swiss exception comes with a footnote that the zoologist cannot ignore. Switzerland did not grant women the right to vote at the federal level until 1971 — decades after every other Western democracy. The canton of Appenzell Innerrhoden did not grant women cantonal suffrage until 1991, and even then, only because the Swiss Federal Court forced it, overruling the male-only Landsgemeinde vote. In a system where the citizens decide, the citizens decided, for seventy-five years after women's suffrage movements had succeeded elsewhere, that half the population should not participate. Direct democracy gives power to the people. It also gives power to the people's prejudices. The mechanism is only as good as the information and the empathy of those who use it.
Switzerland's domestic violence rates are, by European standards, high. Over twenty-one thousand cases were recorded in 2024, a six percent increase on the prior year. Approximately forty-two percent of Swiss women have experienced domestic violence. The rate of domestic homicide — deaths resulting from intimate partner violence — is higher in Switzerland than in many of its European neighbours. And Swiss suicide rates, while declining from historically elevated levels, have consistently been above the world average, with a rate of approximately eleven per one hundred thousand inhabitants. The country that comes closest to the small-group governance model is also, in important dimensions, not the sanctuary its political architecture might suggest. What does that tell us?
The Swiss model demonstrates something that the rest of this chapter has been building toward. The governance mechanism matters. Direct participation is better than distant representation. Small-scale accountability is better than large-scale abstraction. But no governance mechanism, however well designed, can substitute for the conditions it governs. The Swiss vote on policy. They do not vote on whether the economy should generate enough domestic violence to fill twenty-one thousand police reports a year. The economy generates it anyway, because the economy — like the justice system, the education system, the media environment — operates at a scale and complexity that no governance mechanism, however participatory, can fully penetrate.
Governance is not the enclosure. Governance is one feature of the enclosure. The best-governed zoo in the world still produces welfare failures if the enclosure itself is poorly designed. And that — the design of the enclosure itself — is what we must now turn to.
A zoo does not let its animals vote on enclosure management. This statement sounds, when applied to humans, authoritarian — and it would be, if the implication were that some class of humans should govern others without consent. That is not the implication. The implication is that governance, in good zoo practice, is not about who holds authority. It is about what the authority is for.
In modern zoo science, enclosure management is governed by welfare assessment — a systematic evaluation of the animal's behavioural and physiological indicators to determine whether its needs are being met. A paper published in Animals in 2022 by researchers at the University of Bristol documented the growing use of computerised welfare assessment tools that generate visual representations of welfare data across multiple domains: nutrition, health, social behaviour, environmental interaction, reproductive success, stereotypic behaviour, and stress indicators. The data inform management decisions. Enclosures are redesigned. Feeding regimes are adjusted. Social groupings are modified. Enrichment is introduced or withdrawn based on the animal's response. The animal does not vote. The animal is consulted — through its behaviour, through its cortisol levels, through its body condition, through the measurable indicators of its welfare. The keepers watch. The keepers measure. The keepers adjust.
The process is not democratic. It is not autocratic. It is empirical. The question is not "what does the animal want?" — because animals, including humans, often want things that harm them, as every previous chapter has demonstrated. The question is "what does the animal need, and are those needs being met?" The keepers are not rulers. They are maintenance staff. They do not govern the enclosure because they are superior to the animal. They govern it because someone must, and the governance is judged not by the keepers' satisfaction but by the animal's welfare. What if we judged our governments the same way? Not by GDP, not by election results, not by the stock market — but by the measurable welfare of the population? By sleep quality, social connection, mental health, physical health, creative expression, sense of purpose? Would any government on earth pass that assessment?
I am aware of the discomfort this analogy produces. The suggestion that human governance might learn from zoo management triggers an immediate objection: humans are not zoo animals. They have autonomy, dignity, rights, the capacity for self-determination. All true. But the observation stands: our species has designed systems for governing captive animals that are, in important respects, more empirically rigorous, more welfare-oriented, and more responsive to the animal's actual condition than the systems we use to govern ourselves. A zoo that managed its enclosures the way democratic nations manage their citizens — by asking the animals to select, from a set of candidates they have never met, an enclosure manager who would serve for four years with minimal accountability and whose performance would be evaluated not by welfare indicators but by the enclosure manager's ability to win the next selection — would lose its accreditation.
In 2016, Ireland established a Citizens' Assembly — ninety-nine randomly selected citizens, demographically stratified to represent the national population, tasked with deliberating on constitutional questions that the professional political class had been unable to resolve. The selection was by sortition, the same mechanism the Athenians used: random selection from the eligible population, adjusted for demographic representation. The assembly was given access to expert testimony, structured deliberation time, and facilitated discussion. It was asked to consider the question of abortion — a question so politically toxic in Ireland that elected politicians had avoided meaningful action for decades.
In April 2017, eighty-seven percent of the assembly's members voted to recommend amending the constitutional prohibition on abortion. Two-thirds voted to support access without restriction up to a specified gestational limit. The recommendation was passed to the Oireachtas, which called a referendum. In May 2018, sixty-six point four percent of Irish voters approved the amendment. A question that had paralysed the professional political class for a generation was resolved, within two years, by randomly selected citizens who had no political ambition, no donors to satisfy, no re-election to worry about, and no career incentive to avoid the difficult answer. What the professionals could not do in decades, ordinary citizens did in months. What does that tell us about the professionals?
The assembly also addressed climate change, producing recommendations that led to a Joint Committee on Climate Action, a parliamentary declaration of a climate and biodiversity emergency, and a government action plan. The mechanism worked not because the citizens were wiser than the politicians but because the citizens were free from the selection pressures that distort political judgment. They had no donors. They had no party line. They had no ambition to protect. They were, in the language of this chapter, not self-selected. They were randomly selected, which meant they were, for the first time in any modern governance process, representative — not in the electoral sense of "chosen by the majority" but in the statistical sense of "drawn from the population."
This is not a fringe experiment. Similar assemblies have been conducted in France, Belgium, Canada, the United Kingdom, and Germany. The French Citizens' Convention on Climate, convened in 2019, produced one hundred and forty-nine proposals, of which the government adopted many. Belgium's Ostbelgien region established a permanent citizens' council in 2019, randomly selected, with the authority to set the agenda for deliberative panels. The evidence, across every context in which sortition has been tried, points in the same direction: randomly selected citizens, given adequate information and structured deliberation, produce decisions that are more representative, more nuanced, and more responsive to long-term welfare than elected politicians operating under the pressures of fundraising, media management, and re-election. The data is consistent. The evidence is clear. The question is whether we are willing to act on it.
The Athenians knew this. They knew it so well that they built machines to implement it — the kleroteria, stone slabs with slots and tubes that randomised the selection of citizens for public office, ensuring that wealth, eloquence, and ambition could not game the process. And then the knowledge was lost — or rather, it was overridden by the Roman model of elected magistrates, which became the template for every Western democracy that followed. We inherited an electoral system from a civilisation that collapsed, and we treated it as though it were the natural order of things. It is not. It is a design choice. And design choices can be redesigned.
I said earlier that I vote. I want to expand on what that means, because the confession belongs here, in the chapter where governance is examined, and the lens must turn.
I live in the Netherlands. The Dutch system is, by global standards, reasonably functional — proportional representation, coalition government, a constitutional monarchy in which the monarch has no political power. I vote for a party that approximately reflects my values, from a list of candidates I have never met, on the basis of a manifesto I have partially read, in a system I partially understand. The person I vote for will join a coalition with parties whose positions contradict portions of what I voted for. The coalition will produce a governing agreement that no voter voted for, because no voter was offered the opportunity to vote for a governing agreement — only for a party, whose subsequent compromises are negotiated in rooms I will never enter, by people I will never speak to, on the basis of calculations I will never see. I accept this because the alternatives available to me within the current enclosure are worse. You accept it too. We all do. And then we call it self-governance.
But I notice something. When my older son has a problem at school — when a teacher mishandles a situation, when a peer conflict goes unresolved — I walk to the school. I speak to the teacher. I see the teacher's face. I assess the teacher's response. I evaluate, through direct observation, whether the teacher understands the problem and is likely to address it competently. I can do this because the school is small, the teacher is known to me, and the scale permits the kind of direct behavioural evaluation that my neurology was designed for. When the Dutch government mishandles a situation — when the childcare benefit scandal destroys twenty-six thousand families, when housing policy fails an entire generation, when healthcare waiting lists extend beyond the point at which the condition being waited for is treatable — I cannot walk to the government. I cannot see the minister's face. I cannot assess, through direct observation, whether the responsible individuals understand the problem or are competent to address it. I am asked, instead, to evaluate their performance through the same coloured boxes that the previous chapter identified as an information environment optimised for arousal rather than accuracy. Do you see the absurdity? We assess our children's teachers face to face. We assess our nation's leaders through a screen that is designed to make us angry.
The animal that selects its children's teacher through direct observation selects its government through mediated performance. The first process works. The second process produces, with regularity, outcomes that no welfare assessment would endorse.
This is the last chapter of Part Two. I want to draw the threads together, because they have been accumulating across five chapters and the pattern is now, I think, unmistakable.
Money was an attempt to extend cooperation beyond the trust network. It worked at village scale, where the tokens were backed by relationships between people who knew each other. It broke at civilisation scale, where the tokens were created from nothing by institutions the organism could not evaluate, and the organism pledged decades of its life to repay them.
Justice was an attempt to resolve conflict and protect the vulnerable. It worked at village scale, where mediators were known to all parties and reputations were enforced by social visibility. It broke at civilisation scale, where justice was administered by professionals the defendant had never met, in a language the defendant did not speak, using procedures the defendant could not evaluate.
Education was an attempt to transmit knowledge across generations. It worked at village scale, where children learned by doing, from adults they knew, in the context of real tasks. It broke at civilisation scale, where children sat in rows, memorised abstracted content, and emerged after thirteen years unable to feed themselves, resolve a conflict, or regulate their nervous systems.
Media was an attempt to share information across the species. It worked at village scale, where the storyteller was known, the information was bounded, and the organism's attention returned, after the story ended, to its own life. It broke at civilisation scale, where the information was continuous, commercially motivated, optimised for arousal, and delivered through a device that the organism could not stop interacting with.
Governance was an attempt to manage collective resources and resolve collective conflicts. It worked at village scale, where leaders were selected through direct observation, authority was temporary, and accountability was enforced by the proximity of every member to every other member. It broke at civilisation scale, where leaders were selected through mediated performance, authority became a career, and accountability was diffused across millions of voters who had no capacity to evaluate what they were being asked to evaluate.
Five systems. Five good impulses. Five beautiful beginnings. The same failure, five times. Do you see the pattern?
The failure is not in the impulse. The impulse — to cooperate, to be fair, to learn, to share knowledge, to govern collectively — is correct. It is biologically correct. It is how the animal lived for the overwhelming majority of its existence as a species. The failure is in the scale. Every one of these systems worked when the group was small enough for the animal's neurology to function — when the organism could know the people it depended on, observe them directly, hold them accountable through social visibility, and resolve disputes through personal mediation. Every one of these systems broke when the group exceeded the animal's neurological capacity and the systems were asked to operate across populations of millions, then billions, without any corresponding upgrade to the hardware. We scaled the software. We could not scale the brain.
The village became a civilisation. The hardware did not change. The systems that worked at village scale were stretched, abstracted, professionalised, and scaled until they bore no resemblance to the processes they had replaced. And the animal — the same animal, with the same brain, the same social cognition, the same trust architecture calibrated for one hundred and fifty faces — was asked to navigate the result. We are that animal. We are navigating that result. Every day. Right now.
What happened when the village became a civilisation?
Part Three answers.