Chapter One: We Are Not Rational Actors

The dominant story of modern economics begins with a fiction. The fiction has a name, homo economicus, and for roughly two centuries it has shaped the way governments design policy, corporations structure incentives, and consultants build spreadsheet models. Homo economicus is a rational actor. He, and the pronoun feels appropriate, given the tradition's origins, makes decisions by calculating utility, maximising self-interest, and selecting the option that returns the greatest personal benefit. He is, in the language of the discipline, a preference-satisfaction machine.

The problem with this story is not that it is occasionally wrong. The problem is that it describes a creature that does not exist and has never existed. It misses the most fundamental thing about us.

We are not rational actors who happen to be social. We are social animals who, over time, built an economic system on top of our social needs, and then confused the system for the thing itself. The confusion has consequences, and those consequences are accelerating. To understand what is now happening to work, to belonging, to health, and to the communities that sustain human life, we first need to go back much further than the Industrial Revolution, much further than the invention of wage labour. We need to go back to what we actually are.

The Evolutionary Baseline

The anthropologist Robin Dunbar did not set out to change how we understand human society. He was, in the early 1990s, studying the relationship between primate brain size and social group size, a prosaic enough research question, when he noticed something striking. Across different primate species, there was a consistent correlation between the size of the neocortex and the average size of the social group the animal maintained. Larger brains were not correlated with better individual problem-solving. They were correlated with larger, more complex social networks.

The implication was significant. The human brain, the most metabolically expensive organ in the body, consuming roughly 20 per cent of our caloric intake despite constituting only 2 per cent of our mass, did not evolve primarily to think. It evolved primarily to manage social relationships. We are, in Dunbar's formulation, built for belonging.

When Dunbar applied his regression analysis to Homo sapiens, the predicted social group size came out at approximately 150. This number, now known universally as Dunbar's Number, is the cognitive limit of stable social relationships a human being can maintain at any one time. Not acquaintances, not contacts, not LinkedIn connections, but relationships involving genuine mutual knowledge and reciprocal obligation: the people you would call in a crisis, the people who know your history, the people whose absence you would notice and whose presence anchors you in the social world.

One hundred and fifty. The figure recurs across human history with uncanny consistency. The average size of neolithic villages. The basic unit of effective military organisation across multiple armies across multiple centuries. The approximate size of the Hutterite communities that, by long tradition, deliberately split when they exceed this threshold because they find that larger groups lose social cohesion. The average number of people invited to a British wedding. The typical size of a functional business team before it requires formal management hierarchies to hold together. The number keeps appearing because it is not arbitrary, it is the architecture of the human mind.

What Dunbar's research established is that sociality is not a luxury we add to our lives when economic circumstances permit. It is not a preference item, optional when times are lean. It is the core cognitive function for which our brains were designed. We have 150-person-shaped holes in us, and when those holes go unfilled, when the social network thins, when relationships break, when the tribe disperses, something at a neurochemical level begins to malfunction.

The neurochemistry is worth understanding in some detail, because it explains with uncomfortable precision why belonging is not metaphor. Oxytocin is a neuropeptide produced in the hypothalamus and released in social contexts, physical touch, eye contact, shared effort, the experience of being seen and recognised by another person. It binds to receptors across the brain, reducing cortisol (the stress hormone), dampening the threat-detection systems of the amygdala, and producing what researchers describe as a physiological sense of safety. When oxytocin flows, the nervous system shifts out of threat-response mode into what might be called rest-and-belonging mode: blood pressure drops, digestion improves, immune function strengthens, the heart rate settles.

The opposite is also true, and perhaps more important. When oxytocin production is chronically low, when social connection is absent or insufficient, the body behaves as though under sustained threat. Cortisol levels remain elevated. Inflammatory markers rise. The cardiovascular system operates under increased load. Sleep quality degrades. The immune system becomes dysregulated. Loneliness is not merely sad. It is, at the level of cellular biology, a state of chronic physiological stress that damages the body in measurable, documentable, ultimately fatal ways. We will return to the mortality statistics shortly. They are not metaphor either.

The most revelatory research into the biological basis of social need came from Naomi Eisenberger at the University of California, Los Angeles. Eisenberger designed a study in which participants played a simple online ball-tossing game called Cyberball with two other apparent players. In the critical condition of the experiment, the other players gradually stopped throwing the ball to the participant, who was effectively excluded from the game. The exclusion lasted only a few minutes. It was with strangers the participant would never meet. It involved something as trivial as a video game.

Nevertheless, when Eisenberger placed her participants inside a functional MRI scanner and observed what happened in their brains when exclusion occurred, she found something that has since reshaped our understanding of social pain. The brain regions that activated in response to social exclusion were not the ones associated with emotional regulation or cognitive processing. They were the anterior cingulate cortex and the anterior insula, precisely the regions that activate in response to physical pain.

The brain, Eisenberger's work demonstrated, does not distinguish between the pain of a broken bone and the pain of social rejection. Both register in the same neural machinery because, from an evolutionary standpoint, they are the same category of threat. For a social species operating in a world in which individual survival was essentially impossible without group membership, exclusion from the group was a sentence of death. The brain learned, over millions of years of selection pressure, to process social rejection with the same urgency it processes a hand pressed to a hot stove. The signal is: this is dangerous. React now.

This is the baseline from which any serious analysis of work, belonging, and the present crisis must begin. We are not rational actors who prefer social connection when convenient. We are social animals for whom exclusion from the group activates pain pathways, for whom belonging is mediated by the same neurochemical systems that regulate physical threat, for whom the quality of social relationships is literally a matter of life and death. The economic system was built on top of these biological facts. It did not replace them or transcend them. It merely organised them, temporarily, in a particular institutional form, and that form is now changing faster than we have imagined possible.

What Work Actually Provides

When the academic Marie Jahoda published Employment and Unemployment in 1982, she was working from research she had conducted five decades earlier. In the late 1920s and early 1930s, she had studied the community of Marienthal in Austria, a village whose sole employer, a textile factory, had closed, leaving almost the entire working-age population unemployed. What she observed there, and what she spent the following decades theorising, was that the destruction wrought by unemployment was not primarily economic. The families of Marienthal did not starve. They received public assistance. They had food. What they lost was harder to quantify and far more devastating.

Jahoda's central contribution was to distinguish between the manifest function of employment and its latent functions. The manifest function, the thing we think work is for, is income. The latent functions are everything else: the things that work quietly provides, below the level of conscious awareness, that turn out to be essential to psychological wellbeing and social participation.

She identified five latent functions. The first is time structure, the way employment organises the experience of time, dividing the day into meaningful segments, dividing the week into working days and rest days, creating the rhythm that stops time from becoming an undifferentiated, unnavigable mass. The second is social contact, the regular, enforced interaction with people outside the family unit, the ambient human presence of the workplace, the low-stakes conversations that constitute the daily texture of social life. The third is shared goals and collective purpose, the sense of contributing to an enterprise larger than oneself, of effort that matters beyond the immediate household. The fourth is status and identity, the social recognition that comes from occupational role, the answer to the question "what do you do?" The fifth is regular activity, the simple imposition of a reason to be somewhere, doing something, that structures physical presence in the world.

When she surveyed the unemployed residents of Marienthal, what she found was that even men and women who hated their factory jobs, who experienced work as drudgery, who complained bitterly when employed, fell apart without it. They did not enjoy the leisure their unemployment provided. They became more passive. They left the house less. They lost track of time. They ate badly and slept erratically and stopped engaging in activities they had previously claimed to love. The absence of work was not the liberation they had imagined; it was a structural dismantlement of the architecture of a life.

Jahoda's framework has aged remarkably well, in part because the five latent functions map cleanly onto what neuroscience has since established about social and psychological need. But for the purposes of this argument, I want to focus on four of them in particular, developed and extended through the century of research that followed her original observations.

Structure and Temporal Rhythm is the least romantic of the four but perhaps the most consequential in practice. The working day is a machine for creating meaning out of time. The alarm clock, the commute, the morning coffee, the nine o'clock start, the lunch hour, the afternoon slump, the end-of-day transition, these are not merely conventions. They are the scaffolding within which self-regulation, motivation, and a sense of forward movement become possible. Remove them, and time collapses inward. Ask anyone who has spent more than a month without regular employment, and they will tell you that the days became shapeless. Not the days themselves, Tuesday and Thursday still had 24 hours each, but their felt quality, their sense of progress, their capacity to contain anything worth doing. Time structure is invisible until it disappears, at which point its absence is felt as a kind of vertigo. This is not weakness of character. It is a description of the human mind as it actually functions, which requires external temporal anchors to regulate its own operation.

Identity is perhaps the most socially sophisticated of work's latent functions. The question "what do you do?" is the primary social locator in virtually every modern society. It is the first question asked at parties, on dates, in the first minutes of conversation with strangers. This is not idle curiosity. It is the rapid social classification system by which human beings establish the outlines of who they are talking to, what they might have in common, what social register to adopt, where the other person sits in the structure that organises collective life. Occupational identity is not the only form of identity, we are also parents, partners, members of communities, people with histories, but in modern Western societies it has become the primary organising axis around which other identities arrange themselves. To have no occupation is not merely to have no income. It is to lack the primary answer to the primary social question, which creates a particular form of social disorientation that goes well beyond the financial. The unemployed person who says "I don't know who I am any more" is not being theatrical. They are accurately describing a disruption to the identity infrastructure that work quietly maintained.

Community, the third latent function, is the one most often referenced when people discuss what they miss about work, and for good reason. It is the function that maps most directly onto Dunbar's architecture and Eisenberger's neuroscience. The workplace, at its best, is a ready-made community: a group of people who share physical space, common purpose, mutual knowledge, and the daily low-stakes interactions through which genuine social bonds form. These bonds are often dismissed as superficial, "just work friends," people say, but the research does not support that dismissal. The sociologist Mark Granovetter demonstrated in the 1970s that weak ties, the relationships we maintain with people we see regularly but do not know intimately, are in many ways more socially important than strong ties, because they connect us to information, opportunities, and social networks that our close circle does not contain. The colleague you exchange ten words with each morning over the communal coffee machine is not an intimate friend. She is, however, a node in the social fabric that makes you feel embedded in a community rather than adrift in an anonymous world. Lose enough of those nodes, and the fabric itself begins to disintegrate.

Purpose, the conviction that one's efforts contribute to something beyond oneself, that one's presence makes a difference to an enterprise that matters, is the fourth latent function and the one most directly implicated in the Sebastian Junger insight that runs through the social science of belonging like a thread: what humans cannot tolerate is not hardship but uselessness. Viktor Frankl observed, in circumstances of incomparable extremity, that human beings can endure almost any suffering if they can find meaning in it, and are destroyed by comfort if they cannot. The workplace provides purpose not through the grandeur of the work itself, most jobs, honestly considered, are not grand, but through the daily confirmation that one's skills are needed, one's presence expected, one's output consequential to the operation of something. Remove that confirmation, and what replaces it is not liberation but a particular kind of existential nausea: the suspicion that one is surplus to requirements, that the world would continue without interruption if one simply ceased to participate. This is what unemployment does to people who never wanted to be defined by their jobs. The job was not what they loved. The necessity of the job, the fact of being needed for it, was what sustained them.

Jahoda identified these functions in 1982. Decades of subsequent research have extended, refined, and repeatedly replicated her findings. They are not controversial within the academic literature. What is remarkable is how seldom they enter the public conversation about the future of work, in which the debate remains stubbornly focused on the manifest function, income, as though replacing wages with a universal basic income would resolve the crisis that AI displacement is creating. It will not. Income is the smallest part of what work provides. The larger parts are structural, social, psychological, and existential, and no transfer payment addresses any of them.

The Belonging Deficit That Already Existed

In June 2025, the World Health Assembly, the supreme governing body of the World Health Organisation, representing 194 member states, passed its first-ever resolution on social connection. This is worth pausing on. The WHO has coordinated global responses to infectious disease, tobacco, antimicrobial resistance, and the opioid crisis. It had never before felt compelled to pass a formal resolution on the subject of human loneliness. The fact that it did is a measure of how severe the evidence had become.

The resolution was based on the findings of the WHO Commission on Social Connection, which had spent two years gathering data from across the world. Their headline finding was that loneliness and social isolation kill approximately 871,000 people per year, a figure comparable to tobacco-related deaths, and far exceeding deaths from road traffic accidents. One in six people worldwide experience persistent loneliness. Among adolescents and young adults, the figure is one in five. Among older adults, one in three are socially isolated, with the rate rising to one in four in lower-income countries.

These are not soft statistics about feelings. They are hard counts of mortality attributable, through the established physiological mechanisms discussed earlier in this chapter, to the chronic stress of social disconnection. The elevated cortisol, the inflammation, the cardiovascular strain, the immune dysregulation, all of it accumulates over time, and over time it kills.

The most rigorous quantification of this relationship came from the psychologist Julianne Holt-Lunstad, now at Brigham Young University, who conducted a pair of meta-analyses that together constitute the most comprehensive synthesis of the loneliness-mortality literature. The first, published in 2010, synthesised data from 148 separate studies encompassing more than 300,000 participants and concluded that people with adequate social relationships had a 50 per cent greater likelihood of survival over the study periods examined compared to those who lacked adequate social relationships. The second, published in 2015 and covering more than 3.4 million participants, examined specifically the effects of social isolation, loneliness, and living alone on mortality, and found that all three were associated with significantly increased mortality risk, social isolation increasing risk by 26 per cent, loneliness by 26 per cent, and living alone by 32 per cent.

To make the comparison concrete, Holt-Lunstad noted that the mortality risk associated with social isolation is comparable to smoking 15 cigarettes per day. This is not a figure of speech designed to make an abstract risk feel vivid. It is a literal epidemiological comparison, calculated on the same methodological basis as the smoking data. We have spent three decades building a global public health architecture around the harms of tobacco. We tax cigarettes, ban advertising, print graphic warnings on packets, fund cessation programmes, and treat the habit as a serious threat to national health. We have done essentially nothing of comparable scale or seriousness about the epidemic that, by the same measure, kills at the same rate.

In 2023, the United States Surgeon General, Vivek Murthy, published an Advisory, the mechanism reserved for matters of the most serious public health concern, declaring loneliness a national public health emergency. The 82-page document was remarkable in several respects. It did not hedge. It did not qualify. It stated plainly that loneliness had reached epidemic proportions in the United States, that the health consequences were severe and well-documented, and that the failure to address it constituted a collective abdication of social responsibility. Murthy's framing was explicitly structural: loneliness was not the product of individual inadequacy or personal failure but of the dismantlement of the social infrastructure within which human connection had historically occurred. The problem was not that people had stopped wanting to belong. The problem was that the institutions and spaces through which belonging had been organised were disappearing.

It is important to register the full weight of where this evidence places us. We are not describing a future risk, a belonging crisis that might materialise if AI displaces enough workers, if enough offices close, if enough communities fragment. We are describing a crisis that is already fully developed, already measurable, already killing people at industrial scale, and into which an additional, accelerating disruption is now arriving. The building was already on fire before anyone lit a new match.

Third Places That Disappeared

The sociologist Ray Oldenburg published The Great Good Place in 1989, in which he introduced the concept of the "third place", distinguished from the first place (home) and the second place (work) as the informal public gathering spaces that have historically served as the primary venues for community life: pubs, taverns, cafés, barbershops, community halls, the village square, the corner shop. Oldenburg's argument was that third places are not leisure amenities, an optional extra for societies prosperous enough to afford them. They are the infrastructure of civil society itself, the spaces within which social bonds form, political opinions develop, mutual aid organises, and the experience of community cohesion is generated and renewed.

His central observation was that third places are characterised by their accessibility, their informality, their habitual character, and their essentially egalitarian social atmosphere. They are the places you go not because you have an appointment but because you can. Where conversation is the primary activity and hierarchy is suspended at the door. Where regulars accumulate the kind of low-stakes, repeated social contact with one another that, over time, produces genuine social bonds. Where the phrase "a local" describes not merely geography but a form of membership, a social identity conferred by regular presence in a shared space.

What has happened to these places in the four decades since Oldenburg wrote is not difficult to document. In the United Kingdom, the pub has been the archetypal third place for at least three centuries, not merely a place to drink, but the village meeting room, the neighbourhood forum, the social infrastructure of working-class and middle-class life alike. Since 1990, approximately 40 per cent of all British pubs have closed. The closures have not been evenly distributed: they have fallen hardest on rural areas and post-industrial towns, precisely the communities least equipped to absorb the social loss, and they have accelerated in every period of economic pressure. In 2022 alone, more than 400 pubs closed permanently. What replaced them, in most cases, was nothing.

The story of pubs is the story of a broader institutional collapse. Church membership in the United States fell below 50 per cent of the population for the first time in recorded polling history in 2021, down from 76 per cent in 1947. Among adults under 30, fewer than 28 per cent belong to a church, a figure that would have been unimaginable to any previous generation. Trade unions, which at their peak provided not just labour representation but dense social networks, educational programmes, leisure facilities, and a profound sense of collective identity, have declined from covering more than half the British workforce in the 1970s to covering around 23 per cent today, with private sector coverage closer to 13 per cent. Civic organisations, Rotary clubs, Lions, the Women's Institute, have seen sustained membership decline for decades. Bowling leagues, amateur sports clubs, community choirs, neighbourhood associations: all showing the same pattern of contraction.

Robert Putnam catalogued the full extent of this collapse in Bowling Alone in the year 2000, documenting, through a meticulous synthesis of survey data spanning the previous three decades, how Americans had progressively withdrawn from virtually every form of civic participation and social connection. They signed fewer petitions, attended fewer meetings, belonged to fewer organisations, knew their neighbours less well, entertained at home less often, and spent less time with friends and family. Putnam's conclusion was that American social capital, the network of relationships, norms of reciprocity, and trust between individuals that makes collective action and community life possible, had eroded dramatically and systematically. He wrote this in 2000. It has not improved since.

It is important to understand what was lost when these institutions contracted. It was not merely a reduction in the number of agreeable places to spend an evening. The third place performed a specific social function that neither the home nor the workplace could replicate: it was where people encountered, with regularity and low social stakes, people who were different from them. The pub brought together the teacher and the builder, the young and the old, the long-established and the recently arrived. The trade union brought together workers from different trades, different backgrounds, different political instincts, held together by a common interest and a shared social venue. The church brought together people who might otherwise have no reason to occupy the same room. This enforced mixing, uncomfortable sometimes, occasionally revelatory, always socially generative, is the mechanism through which community cohesion is actually built. Not through the performance of togetherness but through the daily, mundane, unremarkable fact of sharing space with people unlike oneself.

When these spaces disappear, people do not simply rediscover community elsewhere. They retreat into smaller, more homogeneous, more defended social worlds, the household, the screen, the curated online community of people who already share their views and preferences. The world gets smaller. The social fabric gets thinner. And the capacity for the kind of loose, ambient, accidental belonging that once characterised healthy communities diminishes, quietly, without anyone declaring a crisis, until suddenly the WHO is passing resolutions and Surgeons General are writing 82-page advisories and researchers are cataloguing 871,000 deaths a year and asking where it all went wrong.

It went wrong steadily, over decades, in ways that were individually invisible but collectively catastrophic. The pub that closed. The church that emptied. The union branch that dissolved. The civic club that folded when its last founding members aged out. Each one a small loss, a minor local tragedy, easily attributed to changing tastes or economic pressure. Together, they represent the dismantlement of the social infrastructure of an entire civilisation.

The Gen Z Paradox

There is a generation that has grown up entirely within this diminished landscape, and they offer the most precise possible test of whether digital technology has compensated for what the physical social infrastructure lost. Generation Z, those born from the mid-1990s to the early 2010s, the first generation to have spent their entire adolescence online, provides the answer, and the answer is categorical.

The psychologist Jean Twenge, who has studied generational trends in American mental health and behaviour for more than two decades, has documented in painstaking detail the mental health crisis that accelerated sharply among American teenagers around 2012, the year that smartphone ownership crossed 50 per cent among American adolescents and Instagram began its expansion into mainstream teenage life. In her research, Twenge found that rates of depression among teenage girls increased by 58 per cent between 2011 and 2016. Rates of anxiety, self-harm, and suicidal ideation all followed comparable trajectories. Rates of loneliness among teenagers and young adults reached levels that had not been recorded in any previous generation.

The timing, Twenge argues, is not coincidental. It corresponds precisely to the transition from a social life organised primarily around physical presence, hanging out, going places, sitting in rooms with other people, to one organised primarily around digital mediation. The teenagers who grew up after 2012 spend dramatically less time with friends in person than previous generations did at the same age. They go out less. They date less. They drive less. They are, by their own account, more connected to more people than any generation in history, and lonelier than any generation in history.

This is what Sherry Turkle meant when she observed that we sacrifice conversation for mere connection. The distinction she was drawing is precise and important. Connection, in the digital sense, is the maintenance of a communication channel, a text thread, a social media relationship, a notification that someone has acknowledged your existence. It is bidirectional and instantaneous and can be maintained at essentially zero cost to either party. Conversation, genuine, sustained, embodied exchange between people who are physically present to one another, is something categorically different. It involves eye contact, tone of voice, physical posture, the reading of micro-expressions, the negotiation of silence, the shared experience of being in a particular place at a particular time. It is, above all, reciprocally vulnerable in a way that digital connection is not. To be physically present with another person is to be visible to them in ways you cannot entirely control. You cannot edit yourself before sending. You cannot compose your reaction after seeing theirs. You are simply there, and they are simply there, and something that social scientists call co-presence, the irreducible, neurologically significant fact of sharing physical space with another conscious being, occurs.

That co-presence is what oxytocin-mediated social bonding requires. Not the communication. Not the information exchange. Not the maintenance of social awareness through a feed. The actual physical presence of another human being, whose heartbeat you can theoretically hear if you are close enough, whose breath you share if you are in the same room, whose attention is a resource they cannot simultaneously allocate elsewhere. The smartphone, for all its communicative power, cannot replicate this. A thousand followers cannot replicate this. A group chat of three hundred cannot replicate this. You cannot belong to a platform. You can only belong to people, and you can only truly belong to people in a place.

Gen Z is the generation that grew up without adequate access to places, the third places already contracted, the social institutions already weakened, the physical gathering infrastructure already thinned, and substituted digital connectivity for what the places had previously provided. The substitution failed. They are more anxious, more depressed, more lonely, more isolated than any recorded generation, despite being more digitally connected than any generation in history. This is not a puzzle. It is a confirmation, conducted at generational scale, of everything the neuroscience predicts. The brain required for belonging did not evolve to process Instagram. It evolved to manage face-to-face social relationships within a group of approximately 150 people, in the same physical space, where the social stakes were real and the mutual recognition was embodied and the oxytocin could actually flow.

The rise of AI companion tools, applications like Replika and Character.ai, which allow users to develop ongoing "relationships" with chatbots designed to be responsive, supportive, and emotionally attuned, represents the most acute expression of what happens when genuine belonging is unavailable. These tools are used by millions of people, many of them young, many of them lonely in ways that have no institutional solution within the landscape they inhabit. The tools are not without value; they can provide a degree of social scaffolding for people who lack access to the real thing. But they cannot provide the thing itself. You cannot belong to a chatbot. You cannot matter to an algorithm. The chatbot's support is not chosen, it is programmed. Its attention is not a limited resource allocated in your favour, it is a function running on a server that costs nothing to extend simultaneously to ten thousand other users. The irreducible element of belonging, the element that the neuroscience traces to oxytocin, that Eisenberger's research locates in the anterior cingulate cortex, that Dunbar's analysis roots in the evolutionary architecture of the primate brain, is the knowledge that another conscious being has chosen to attend to you, in their finite time, in a shared physical world. Software cannot provide this. It can simulate the form of social connection while being constitutively incapable of delivering the substance.

The Convergence

This chapter has moved through five seemingly distinct territories: evolutionary biology, labour economics, public health epidemiology, institutional sociology, and generational psychology. The movement was not arbitrary. Each territory illuminates a different dimension of the same underlying reality.

We are social animals, not by preference but by neurobiological design, shaped by millions of years of selection pressure that made group membership the precondition for survival. Our brains are optimised for managing 150-person social groups. Our neurochemistry responds to social exclusion with the same pain signals it deploys for physical injury. Our bodies, under conditions of chronic social isolation, enter a state of sustained physiological stress that degrades every major system and ultimately kills us.

For most of the last century, the workplace was the primary institution through which modern societies organised the social life of adults. Not intentionally, this was never the explicit purpose of employment, but functionally, through the latent mechanisms Jahoda identified: the structure, the identity, the community, the purpose that work quietly provided alongside the wages that were everyone's conscious focus. The workplace was the container into which millions of social animals who had lost their villages, their churches, their civic clubs, and their extended families could reliably pour themselves and find, each morning, a reason to leave the house, a group of people who knew their name, and the daily confirmation that they were needed by something larger than themselves.

That container is now developing serious cracks. AI is dismantling the social infrastructure of work at a pace and scale that has no historical precedent. It is doing so into a society whose other belonging institutions, the third places, the civic organisations, the religious communities, the trade unions, have already been substantially eroded. And the generation that will bear the most acute consequences is already the loneliest ever recorded, having grown up in a world where the physical infrastructure of belonging was thin and the digital substitute was abundant, and having discovered that abundance of connection is not a cure for poverty of belonging.

This is the crisis. Not one crisis, compounding. Several crises, converging. And they are converging faster than most people, most policymakers, most corporate leaders, most economists, most technology optimists, are prepared to acknowledge.

There is, however, a counterpoint to this story that the data also insists upon. In the same period that traditional belonging institutions collapsed, one category of physical gathering has grown, has in fact grown at record pace. Fitness and wellness communities. Gyms, studios, running clubs, martial arts academies, swimming groups, cycling classes. US gym membership reached 77 million in 2024, an all-time record, up 20 per cent from 2019. Running clubs surged 59 per cent. One in four Americans now holds a fitness facility membership. And the research is clear that the growth is not primarily about physical health outcomes: members who feel part of a community are three times more likely to maintain their membership long-term, group class participants are significantly less likely to cancel than solo gym-goers, and when surveyed, members consistently identify social connection, the people they know there, the sense of belonging to a group, as the primary driver of their continued participation.

This is not a coincidence. It is a substitution, a spontaneous, decentralised, market-mediated substitution, in which millions of people who have lost the social infrastructure of the pub, the church, the union, and the neighbourhood are finding their tribe in a rather different kind of physical space. The logic is identical to the logic of the third place: regular, habitual, embodied presence in a shared physical environment, with the same people, around a shared purpose, generating through repetition the ambient social bonds that constitute belonging. The medium is different. The mechanism is the same.

The question this book asks, and will spend its remaining chapters attempting to answer, is what follows from recognising this. What are the obligations of those who operate these spaces? What is the opportunity? What would it look like to build physical fitness and wellbeing facilities that take seriously, at an institutional and operational level, the social function they are inadvertently already performing? And what happens to communities, to the belonging infrastructure of modern society, if the fitness industry rises to this historical moment, or fails to?

Those are the questions of subsequent chapters. The foundation, laid in this one, is simply this: we are not rational actors who happen to enjoy social connection. We are social animals who built an economic system on top of our social needs, mistook the system for the need, allowed the system to erode the older institutions through which the need had been met, and are now confronting the consequences at scale, with no clear plan.

We have never needed to understand what we are more urgently than we do right now.