In the previous instalment, we looked at the formulaic AI think-piece β the six-paragraph article that walks you from fear to reassurance to motivational platitude, so predictable in structure that a machine can produce it in four seconds. We called it human slop: content technically written by a person, but so templated that authorship is beside the point.
That was the informational variety. This is the intellectual one.
THE SHARING INSTINCT
How Cooperation Built β and Destroyed β the World We Know
CHAPTER 1: A Story From Easter Island
In 1722, when the Dutch navigator Jacob Roggeveen became the first European to set foot on Easter Island, he encountered a puzzle that would haunt Western science for centuries. Scattered across the barren, treeless landscape stood nearly nine hundred enormous stone statues β the moai β some weighing over eighty tonnes, staring silently inland from the coast. The island's few thousand inhabitants lived in what Roggeveen described as "wretched" conditions. They had no timber, no metal tools, and no apparent means of having carved, transported, or erected the monoliths that towered above them.
The mystery seemed straightforward: how did they build them? But the deeper question β the one this book is really about β is: why did they stop?
For decades, the dominant theory was ecological suicide. The islanders, so the story went, felled every last tree to build the log rollers needed for transporting moai, collapsing their own ecosystem in a frenzy of competitive monument-building. It was a parable for our times: a civilisation that consumed itself. Jared Diamond made it the centrepiece of his 2005 book Collapse, and the story entered popular consciousness as a cautionary tale about resource depletion.
But recent archaeology has complicated this picture almost beyond recognition. Pollen analyses suggest the deforestation began centuries before the moai era, likely driven by Polynesian rats that ate the palm seeds. The islanders didn't fell the forests in a fit of monumental vanity; the forests were already failing when the statue-building began. And the moai themselves, far from being symbols of competitive excess, appear to have served a cooperative function β marking territories, codifying social agreements, and managing the distribution of scarce resources among competing clans.
The statues, in other words, were not the problem. They were the solution. Easter Island didn't collapse because its people cooperated too little. It collapsed β to the extent that it collapsed at all, which is itself debated β because the structures of cooperation that had worked for centuries couldn't adapt fast enough when external conditions changed.
This distinction β between the instinct to cooperate and the fragility of cooperative structures β is the subject of this book.
CHAPTER 2: The Potlatch Problem
In the winter of 1921, Kwakwaka'wakw chief Dan Cranmer hosted one of the largest potlatches in recorded history on Village Island, off the coast of British Columbia. Over several days, he gave away pool tables, sewing machines, canoes, Hudson's Bay blankets by the hundreds, and β by some accounts β several motorboats. Under Canadian law, the ceremony was illegal. The authorities arrested forty-five participants.
To the colonial administration, potlatch was irrational β a bizarre ritual of competitive gift-giving that impoverished its participants. To the Kwakwaka'wakw, it was the economy itself: a system for redistributing wealth, settling debts, establishing rank, and encoding social memory. The blankets weren't wasted. They were data.
What makes potlatch relevant to our story is not its exoticism but its familiarity. Look carefully and you'll find potlatch logic operating in almost every human society that has ever existed β including our own. The venture capitalist who funds a dozen failing startups to land one winner is engaging in a form of competitive redistribution. The academic who publishes freely, building reputation rather than charging for access, is operating within a prestige economy structurally identical to the one Cranmer navigated on Village Island. The logic of the potlatch is the logic of cooperation under uncertainty: distribute resources widely, accept losses, and trust that the network will return value over time.
The difference is that we've stopped recognising it as such.
CHAPTER 3: The Axial Age and the Scaling Problem
Between roughly 800 and 200 BCE, something remarkable happened across multiple civilisations simultaneously. In Greece, China, India, Persia, and the Levant β societies with little or no contact with each other β a new kind of thinking emerged. Philosophers and religious leaders began, for the first time, to articulate universal ethical principles: Confucian reciprocity, Buddhist compassion, Greek rationalism, Zoroastrian moral dualism, Hebrew monotheistic justice.
The German philosopher Karl Jaspers called this the Axial Age, and it has been a staple of grand historical narratives ever since. The usual explanation is ideational: human thought simply matured. But there's a more structural reading. Each of these societies had recently crossed a threshold of scale β through urbanisation, trade, or conquest β that made older, kin-based cooperation untenable. You can sustain a gift economy in a village where everyone knows everyone. You cannot sustain it in a city of fifty thousand strangers.
The Axial Age, on this reading, was not a spontaneous flowering of wisdom. It was an emergency firmware update. Societies that had outgrown tribal cooperation needed new software β universal principles, codified laws, impersonal institutions β to coordinate at scale. The philosophies that survived were the ones that solved the scaling problem.
This reframing changes how we understand the present. Our own era faces an eerily similar challenge. The cooperative structures that sustained the twentieth century β nation-states, international treaties, professional guilds, shared media environments β are struggling under the weight of globalisation, digital communication, and (ironically) the very interconnectedness they were designed to foster. We have scaled past our institutions again. And just like the societies of the Axial Age, we're scrambling for new ones.
CHAPTER 4: Trust Molecules and Trading Posts
In 2004, neuroeconomist Paul Zak ran an experiment that would make him briefly famous. He had subjects play a trust game β one player sends money to another, it's tripled, and the receiver decides how much to return β while measuring their oxytocin levels. The results were striking: receiving a show of trust triggered an oxytocin surge that made recipients more generous in return. Zak called oxytocin "the moral molecule" and went on a TED-talk circuit arguing that biology had hardwired us for cooperation.
The story is seductive. It's also significantly overstated. Subsequent replications have been mixed; oxytocin's effects are highly context-dependent, modulated by in-group/out-group dynamics, prior relationships, and cultural framing. The "moral molecule" narrative collapsed under the weight of its own simplicity β a pattern we'll see again.
But Zak's underlying question was the right one. If cooperation is so fragile β if every act of trust is a gamble that the other party won't defect β how did humans manage to build cooperative networks that spanned continents? The answer, drawn from economic history and anthropology alike, is: slowly, and with a lot of infrastructure.
Consider the fondaco β the medieval trading post that European and Arab merchants maintained in each other's port cities. These were not merely commercial buildings. They were trust architectures: physical spaces where reputation could be verified, disputes adjudicated, and contracts enforced in the absence of shared law. The fondaco solved the same problem that blockchain enthusiasts claim to be solving today β trustless cooperation between strangers β but it did so through social engineering rather than cryptographic proof.
Every lasting human institution, from the Hanseatic League to the modern university, is a fondaco in disguise: a structure that makes cooperation cheaper by reducing the cost of trust.
CHAPTER 5: The Cooperation Trap
If the previous chapters paint cooperation as humanity's great talent, this one must paint its shadow. Because the same instinct that builds cathedrals and trading networks also builds cartels, ethno-states, and armed alliances. Cooperation is not inherently virtuous. It is a tool, and like all tools, it serves the purposes of whoever wields it.
The mafia is one of the most effective cooperative institutions in modern history. So is the multinational cartel. So, for that matter, was the transatlantic slave trade β a staggeringly complex network of cooperation between European merchants, African brokers, plantation owners, and financial institutions, sustained across centuries and oceans. The cooperative structures that enabled it were, in a purely structural sense, as sophisticated as anything the Hanseatic League produced.
This is the cooperation trap: the features that make cooperative systems powerful β shared identity, mutual obligation, trust enforcement, collective action β are precisely the features that make them dangerous when directed inward. Every in-group implies an out-group. Every trust network has a boundary, and the treatment of those outside the boundary is often inversely proportional to the trust maintained within it.
Understanding this is not optional. Any serious account of human cooperation that stops at the heartwarming examples is not scholarship but advertising.
CHAPTER 6: Building Better Fondacos
So where does this leave us?
We are a species that cooperates compulsively, ingeniously, and often destructively. Our cooperative structures β from village reciprocity to international law β are powerful but fragile, effective but amoral, and perpetually outpaced by the scale of the problems they're meant to solve. The Easter Islanders didn't lack the instinct to cooperate. They lacked structures that could adapt when the ecology shifted beneath them. We're in the same position today.
But the Axial Age offers a counterpoint, and perhaps a reason for cautious optimism. When existing cooperative structures have failed at scale, human societies have occasionally managed to invent new ones β not through sudden enlightenment, but through slow, messy, institution-level experimentation. The challenge of our era is the same challenge Confucius and Solon faced: how to build trust architectures that work among strangers, at scale, under conditions of radical uncertainty.
We have done it before. We will need to do it again. And understanding how cooperation actually works β not the heartwarming version, but the full picture, with its cartels and its cathedrals β is where that work begins.
You've probably read this book before.
Not this specific book β The Sharing Instinct doesn't exist; an AI wrote it in about ninety seconds. But you've read its siblings. You bought one at an airport. You saw the TED talk. Someone's uncle mentioned it at Christmas and got the thesis slightly wrong, and nobody at the table could tell because the thesis was designed to be slightly-gotten-wrong-at-Christmas.
The structure is always the same:
1. Big provocative question
"Why did civilisation develop unevenly?"
2. A surprising anecdote
Polynesian fishermen / obscure tribe / casino experiment / etc.
3. A bold universal claim
"Everything is actually caused by X."
4. A parade of loosely connected case studies
5. A satisfying grand theory
6. A hopeful or philosophical ending
"Understanding this helps us build a better future."
The Sharing Instinct follows this template beat for beat. Easter Island is the surprising anecdote. Potlatch is the "I bet you didn't know about this" case study. The Axial Age provides intellectual gravitas. Oxytocin supplies a science-flavoured interlude. Chapter 5 exists so the book can claim to have addressed the dark side. And Chapter 6 lands on the genre-mandatory note of cautious optimism, because every book that explains why the world is broken must end by suggesting we can fix it, or nobody buys it.
Swap "cooperation" for "geography" and you have Guns, Germs, and Steel. Swap it for "cognitive bias" and you have half of the behavioural economics shelf. Swap it for "networks" and you have the other half. The formula is so stable that it constitutes, in effect, a genre β and like all genres, it has conventions that readers have internalised to the point of invisibility.
The real problem is not that these books exist. It's how they get made.
Somewhere upstream of every pop-nonfiction bestseller is a body of actual scholarship: millions of pages of anthropological fieldwork, economic modelling, archaeological analysis, and peer-reviewed debate. This scholarship is dense, qualified, and riddled with disagreement. It has to be; that's how knowledge works. An academic paper on Easter Island's deforestation will spend thirty pages on pollen core methodology before cautiously proposing a partial conclusion hedged with six caveats.
Nobody buys that at an airport.
So the scholarship gets compressed. First by journalists, who summarise the findings and strip the caveats. Then by pop-nonfiction authors, who select the most narratively satisfying results across multiple fields and weave them into a single argument. Then by TED speakers, who distil the book into eighteen minutes with a personal anecdote and a memorable closing line. Then by social media, which reduces it further to a sentence, an infographic, or a quote card.
Anthropology (millions of pages, genuine results)
β
Journalism (summary; some nuance preserved)
β
Pop-nonfiction book (getting sloppy)
β
TED talk (getting sloppier)
β
Social media thread (pure signal loss)
Each layer removes nuance. Each layer increases reach. And each layer inherits more confidence than the evidence supports, because qualifications don't compress well. A paper that says "the evidence tentatively suggests X under conditions Y, though see Z for an alternative interpretation" becomes a book that says "X explains everything," which becomes a talk that says "X," which becomes a post that says "X!!!"
This is not a new observation β scholars have complained about popularisation for centuries. What makes it worth revisiting now is that AI has made the pipeline's structure impossible to ignore. The Sharing Instinct was generated by a language model in about ninety seconds. It produced a plausible pop-nonfiction book because the genre's conventions are so rigid that pattern-matching is sufficient. The template is the product.
AI didn't invent intellectual compression. Publishing did. The model just runs the same pipeline faster.
In the first essay, we argued that formulaic AI think-pieces are informational junk food: they satisfy the craving for legibility without providing real understanding. Pop-nonfiction slop is the intellectual equivalent. It satisfies a different craving β not "what should I do about AI?" but "why is the world like this?" β and it satisfies it the same way: with a feeling of insight that doesn't survive contact with the actual literature.
This is not the same as saying the books are wrong. Many of them cite real research and contain genuinely interesting material. Guns, Germs, and Steel drew on decades of serious biogeography. Thinking, Fast and Slow synthesised a Nobel-winning research programme. The underlying scholarship in these cases was substantial. The problem is what the genre does to that scholarship in the process of packaging it for a mass audience.
What it does, specifically, is three things.
First, it selects for narrative coherence over evidential strength. A case study gets included not because it's the most representative but because it's the most surprising. Easter Island is in every other pop-nonfiction book not because it's the best example of ecological collapse β many scholars argue it isn't one at all β but because giant stone heads on a barren island make for an irresistible opening chapter.
Second, it demands a unified theory. Academic disciplines live comfortably with partial, competing explanations. Pop nonfiction cannot. The genre requires a single X that explains everything, because a book called "Several Partially Overlapping Factors, Each With Significant Exceptions, That May Have Contributed to Uneven Development" does not sell. The bold universal claim is not an intellectual conclusion; it's a marketing requirement.
Third, it flattens disagreement. Any serious field contains active debates, unresolved questions, and scholars who think the dominant narrative is wrong. The pop-nonfiction template handles this by acknowledging the debate in a paragraph and then moving on, giving the reader the impression that dissent exists but has been weighed and found wanting. The Chapter 5 manoeuvre in The Sharing Instinct β "cooperation also has a dark side" β is a structural concession, not a genuine engagement. It exists so the book can say it addressed the objection.
The result is a product that feels like deep understanding. You finish the book and you feel you know why civilisation developed unevenly, or why people make irrational decisions, or how networks shape history. But what you actually have is a narrative β compelling, well-constructed, and almost impossible to fact-check without reading the same fifty papers the author read. The feeling of understanding and the fact of understanding are not the same thing, and the genre is engineered to blur the distinction.
These books promise to explain the world. Real scholarship usually explains why the world is harder to explain than you thought.
It would be easy at this point to slide into a familiar kind of contempt β the STEM-adjacent sneer at "soft" disciplines, the implication that the humanities are inherently unserious and anyone who reads popular nonfiction is a rube.
That is precisely the wrong conclusion.
The problem is not liberal arts. Anthropology, history, economics, and cognitive science are serious disciplines producing genuine knowledge. The problem is the packaging layer that sits between academia and the mass readership: the publishing formulas, the TED-talk conventions, the social-media incentives that reward bold claims and punish careful qualification. The scholars who spend years on pollen core analysis are doing real work. The system that turns their findings into a chapter called "A Story From Easter Island" is doing something else.
And the system is remarkably good at what it does. The pop-nonfiction template works β commercially β because it maps onto real cognitive needs. Humans genuinely do want to understand why the world is the way it is. We genuinely do think in narratives. We genuinely do find it satisfying when complex phenomena resolve into a single explanatory principle. The template doesn't succeed by tricking people. It succeeds by giving people exactly what they want, in the precise form they want it, at the cost of the nuance they didn't know they were missing.
This is, structurally, the same argument we made about formulaic AI think-pieces. Those articles persist not because readers are stupid, but because they deliver the feeling of understanding a fast-moving field without the discomfort of genuine uncertainty. Pop-nonfiction slop does the same thing on a larger scale and a longer timeline. The packaging is better; the mechanism is identical.
Just as in the first essay, the antidote is specificity β and a tolerance for messiness.
A book that genuinely engages with the Easter Island debate doesn't open with a cinematic anecdote and pivot to a grand theory. It opens with the disagreement: here's what Diamond argued, here's what Hunt and Lipo found, here's what the pollen record suggests, here's what we still don't know. It tells you what the evidence can and cannot support. It makes the uncertainty legible rather than smoothing it away. This kind of book exists β many excellent works of public scholarship do exactly this β but it doesn't dominate the bestseller lists, because uncertainty is a harder sell than revelation.
The question for readers is not whether to stop reading popular nonfiction. It's whether to treat the feeling of insight as the end of the inquiry or the beginning. The pop-nonfiction book is, at its best, an index β a pointer to real scholarship that you can follow if you choose to. At its worst, it's a substitute for that scholarship, and a reader who finishes it believing they now understand Polynesian ecology or medieval trade networks has confused the map for the territory.
AI hasn't changed this dynamic. But it has, once again, clarified where the problem lies. A language model that has never read a pollen core analysis, never visited Easter Island, and has no opinions about the Kwakwaka'wakw can produce a text that passes for pop-nonfiction scholarship β not because AI is brilliant, but because the pipeline had already reduced the work to a template. The producers of human slop had made their own output so formulaic, so mechanically reproducible, that a machine could learn the pattern and run it on autopilot. That is not a commentary on artificial intelligence. It is a commentary on what the publishing industry decided was good enough.
The author would like to include a list of real-life samples, but the editorial board decided it would provoke unnecessary legal conflicts and thus the list was removed.