Cultural Evolution

Here's a puzzle: European explorers — some of the smartest, most educated people from one of the most technologically advanced civilizations in history — routinely starved to death in environments where indigenous peoples thrived. They starved in the Arctic, where the Inuit considered the land among their richest hunting grounds. They starved in jungles, surrounded by edible plants. One group nearly perished in Alabama before being captured by local Indians. These weren't stupid people. They simply didn't have the culture.¹

The anthropologist Joseph Henrich argues that what separates humans from other apes isn't raw intelligence — it's our capacity for cumulative cultural transmission. Individual humans are bad at inventing survival techniques from scratch. What we're spectacularly good at is copying, refining, and faithfully passing on complex bodies of knowledge across generations. The difference between a clever ape and a technological civilization isn't a bigger brain; it's a bigger collective brain, one distributed across social networks and maintained by high-fidelity cultural learning.²

The Complexity You Can't Reinvent

Consider how the Inuit hunt seals. You find a breathing hole in the ice, verify by smell that it's active, assess the hole's shape with a curved piece of caribou antler, set a down indicator, then plunge a harpoon with a detachable tip tethered by sinew line. The harpoon shaft is made of driftwood (too scarce for fires), the rear spike of polar bear bone (you also need to know how to kill polar bears). To cook the seal, carve a lamp from soapstone, render blubber for oil, make a wick from a specific moss species. For water, identify old sea ice by color and texture, melt it with your soapstone lamp.

No single person figured this out through reason. The Fuegian arrow-making process involves fourteen steps, seven tools, and six different materials, including choices that seem arbitrary (gnarled chaura wood when straighter options exist, feathers specifically from the left wing for right-handed archers). Every hunting-gathering culture has knowledge of this complexity. And here's the thing that should unsettle anyone who prizes rationality: nobody in these cultures really understands why most of these steps are necessary. The knowledge is causally opaque — it works, but the causal chain linking each step to the outcome is invisible to the practitioners.¹

Why Tradition Beats Reason

This is where Henrich's argument gets uncomfortable. For essentially all of human history, the most dangerous thing you could do was think for yourself.

The Naskapi of Labrador decided where to hunt caribou by reading patterns of cracks in heated shoulder bones — a form of divination. This sounds insane, but it's actually a near-optimal strategy. Hunting is a matching-pennies game: hunters want to find caribou, caribou want to avoid hunters. Any predictable bias in hunter decision-making (returning to previously successful spots) can be exploited by prey. Divination provides a crude randomizing device, bypassing the cognitive biases that make humans terrible at generating truly random sequences.¹

The Romans used bird augury to decide when and where to attack. Generals risking thousands of soldiers on the basis of weird birds sounds crazy — until you realize that in warfare, unpredictability is a strategic advantage. If you're generally predictable (and Scott Aaronson's experiments show that you are — a simple program can predict your "random" choices 70% of the time), outsourcing decisions to random oracles may genuinely be the best available strategy.¹

Then there's manioc. This tuber contains cyanide. Amazonian peoples developed a multi-step, multi-day processing technique — scraping, grating, washing, separating, boiling, sitting for two more days, then baking — that removes the toxins. The process is so laborious that a "reasonable" person might experiment with shortcuts. They'd find that simple boiling removes the bitter taste and prevents acute symptoms. Everything would seem fine. Twenty years later, chronic cyanide poisoning would begin manifesting as goiter and neurological damage, in a way so causally distant from the dietary shortcut that nobody would connect the two. When the Portuguese brought manioc to Africa without the processing techniques, chronic cyanide poisoning became — and remains — a serious health problem, centuries later.¹

The lesson is stark: in evolutionary terms, reason is dangerous. Evolution seems to have "wished" for creatures smart enough to learn traditions but not smart enough to question them. Since it couldn't find that design, culture itself evolved the next best thing — elaborate mechanisms for suppressing independent reasoning. Witch-hunting, religious dietary laws, fairy tales where violating obscure prohibitions brings catastrophe. These aren't bugs in human culture. They're the immune system that keeps the knowledge base intact.¹

The Wheel Experiment: Proof by Ramp

Derex, Boyd and colleagues turned this argument into a laboratory demonstration. The setup: a wheel with four spokes resting on an inclined ramp, with movable weights on each spoke. Participants adjusted the weights to make the wheel roll down as fast as possible, arranged in "transmission chains" of five — each person could see what the previous person did before taking their own turns.³

The wheel got faster. Averaged across fourteen chains, speed improved steadily from 123.6 m/h in the first generation to 145.7 m/h by the fifth. Cultural evolution was working. But when each participant was tested on their understanding — asked to predict which of two wheel configurations was faster — the fifth generation was no better than the first. Performance improved; comprehension didn't budge.⁴

The second experiment was the killer. They let each generation leave a written theory for the next person. Surely explicit causal reasoning would help? It didn't. In some respects, theories actually made things worse. Participants who inherited a theory tended to believe it and stopped exploring alternatives. Of 56 theory-receivers, some got inertia-related theories, some got energy-related theories, some got theories so wrong as to be irrelevant — and the theories showed no convergence toward correctness over generations. People got better at the task while their explanations stayed bad, and having an explanation made them less curious.⁴

This is the experimental proof of what the manioc and Naskapi cases argue from history: cultural evolution doesn't need anyone to understand what they're doing. Faithful copying of successful behavior plus a little experimentation is enough. Causal understanding is a bonus, not a requirement — and when it arrives prematurely, it can actually slow things down by narrowing the search space. The implications for how we think about Cognitive Biases sharpen further: what looks like irrational deference to tradition may be the only strategy that works when the causal structure is too complex for individual reasoning.

Rational Imitation

The copying that drives cultural evolution isn't mindless. Even chimpanzees show a surprising sensitivity to why a demonstrator chose a particular action. In a clever adaptation of a study originally designed for human infants, Buttelmann and colleagues showed enculturated chimps an experimenter operating devices with unusual body parts — pressing a button with his forehead, pushing a box with his foot, sitting on a lid. In the "hands free" condition, the experimenter had his hands available but chose the unusual method anyway. In the "hands occupied" condition, he used the unusual method because his hands were holding something.⁵

The chimps copied the unusual action significantly more in the hands-free condition — exactly the same pattern as 14-month-old human infants. The logic, in both species: if someone uses an unusual method when they could have used an easier one, there must be a reason, so copy the method. If they're forced into it by circumstances that don't apply to you, just get the result however you can. This is rational imitation — selective copying that evaluates the demonstrator's intentions before deciding what to reproduce.⁵

This matters for the cultural evolution story because it shows that the building blocks of high-fidelity cultural transmission — attention to others' choices, inference about hidden reasons, selective reproduction of intentional actions — exist in our closest relatives. What they lack isn't the capacity for rational imitation but the capacity for cumulative culture: the ratchet that lets each generation add to what the last one built. The wheel experiment showed that humans achieve cumulative improvement without understanding what they're doing. Chimps achieve understanding of others' actions without cumulative improvement. The human trick was combining both — rational imitation plus faithful transmission plus occasional blind variation — into a system that generates complexity no individual could design.

The Collective Brain

If cultural generation happens at some low rate through trial and error, then the fidelity of transmission determines whether a society advances, stagnates, or regresses. Henrich calls this the "collective brain" — your community's total capacity for maintaining and innovating cultural knowledge. Larger, more interconnected populations have fancier tools and technologies, not because individuals are smarter, but because the collective pool is bigger.²

There's archaeological evidence for this. The Polar Inuit, cut off from other Inuit populations, began losing valuable tools and technologies despite having the same-sized individual brains. Their collective brain was severed. The inverse is also visible: at Olorgesailie in Kenya, around 320,000 years ago — right at the dawn of Homo sapiens — archaeological finds show obsidian tools traded from sources 100 kilometers away, carefully processed pigments, and sophisticated Middle Stone Age technology that replaced million-year-old Acheulean hand axes. These changes coincided with environmental instability, which pushed human groups to build broader social networks. More connections meant a bigger collective brain, which meant a cultural ratchet that could actually turn.⁶

This reframes the whole question of human uniqueness. We didn't become special because we got smart. We became special because we got social in a way that let cultural knowledge accumulate faster than it decayed. The brain adaptations that distinguish us from other apes aren't primarily about logic — they're about imitation, social learning, and the transmission of complex skills. Human toddlers perform about equally with chimpanzees on memory, logic, and spatial reasoning tasks. Where they absolutely destroy chimps is in learning from other people.¹

Religion as Technology

This framework transforms how you think about religion. Bret Devereaux, a historian of ancient warfare, argues that polytheistic religions were fundamentally practical — not about moral belief, but about practical knowledge. The Roman didn't care about theology in the abstract; he needed to know which gods existed, what they wanted, and what rituals to perform to keep them happy. The proof that a ritual worked was that the community still existed. If the ritual had been performed for generations and the society survived, then by definition the ritual was effective.⁷

This is exactly the trial-and-error cultural evolution that Henrich describes, wearing religious clothing. And it's why ancient religions were both fiercely conservative (if it works, do it exactly the same way forever) and paradoxically open to innovation (if a crisis suggests the gods are angry, try something new; if the crisis abates, add it to the portfolio of things-that-work). The system could absorb change without anyone noticing things had changed — a property most modern religions with fixed texts lack.⁷

Henrich extends this to the rise of "big gods" — moralizing deities with power over the afterlife. For most of human history, people believed in gods that were whimsical and weak. Big gods emerged because societies that adopted them could scale cooperation to larger groups. It's intergroup competition: if your set of rituals and beliefs lets you out-compete neighboring groups, your religion spreads. The earliest monumental architecture is always religious — temples and tombs — suggesting that shared ritual was the scaffolding on which complex societies were built.²

Göbekli Tepe offers a striking illustration. This site in southeastern Turkey features massive carved stone pillars, eleven thousand years old, erected by people who hadn't invented pottery or metal tools. There's no evidence of permanent settlement — no cooking hearths, houses, or trash pits. It appears to be a purely ceremonial site, built by hunter-gatherers before the advent of agriculture. The conventional narrative said farming came first, then surplus, then temples. Göbekli Tepe inverts this: the social coordination required to build the monument may have been what drove the development of settled agriculture, not the other way around.⁸

Self-Domestication and the Shrinking Brain

There's a physical signature of the collective brain hypothesis that nobody talks about enough: human brains have been shrinking. Over the past 20,000 years, the average male brain has lost about 150 cubic centimeters — a chunk the size of a tennis ball. The trend appears in China, Europe, Africa, everywhere we look. If it continues at this rate for another 20,000 years, we'll approach Homo erectus brain volumes.⁹

Three theories compete. The climate explanation (warmer planet, smaller bodies, smaller brains) doesn't account for the fact that comparable warming periods occurred many times before without shrinkage. The nutrition explanation (agriculture initially worsened diets) fails because the trend appears in Australia and southern Africa, where agriculture arrived only recently. The most interesting explanation is self-domestication: as complex societies emerged, the brain shrank because you didn't need to be as smart to stay alive. Individuals who couldn't survive alone could scrape by with the help of others — the first social safety nets.⁹

Richard Wrangham pushes this further. Every domesticated animal — all thirty of them — has lost brain volume compared to its wild ancestor, typically 10-15 percent. They also develop more gracile builds, smaller teeth, flatter faces, and a wider range of coloration. Sound familiar? Wrangham argues humans domesticated themselves through capital punishment. "Over the last 100,000 years, language became sufficiently sophisticated that when you had some bully who was a repeat offender, people got together and said, 'We've got to do something about Joe.'" Anthropological records suggest the death penalty was applied to at least 10 percent of young men per generation in traditional societies. Select against aggression for enough generations and you get the domestication syndrome — including smaller brains.⁹

The Siberian silver fox experiment provides the mechanism. Starting in 1958, selecting only the tamest foxes for breeding produced, within twelve generations, animals with white patches, floppy ears, curly tails, smaller skulls, and dramatically reduced aggression — the whole domestication package from selecting on a single behavioral trait. Wrangham suspects the shortcut is through regulatory genes that slow brain development relative to body development, yielding adults with more juvenile (and less aggressive) brains.

Brian Hare's comparison of dogs and wolves maps the cognitive consequences. Wolves, with larger brains, are better at independent problem-solving — more prone to flashes of insight, more persistent when stuck. Dogs excel at reading human social cues — tracking gaze and gestures, using humans as tools. The domesticated animal isn't dumber, but it thinks differently: less autonomous cognition, more social cognition. If the same tradeoff happened in humans, our shrinking brains may reflect not declining intelligence but a cognitive style increasingly optimized for the collective brain — less lone-wolf brilliance, more capacity for the cooperative information processing that Henrich's framework predicts.⁹

Lost and Found

One of the most unsettling implications of cultural evolution is that knowledge doesn't move in a straight line. Anton Howes, a historian of innovation, documents how ancient remedies that "obviously" didn't work turn out, on closer examination, to have worked perfectly well — for reasons nobody understood at the time. Terra sigillata, a clay from the Aegean island of Lemnos, was used as a cure-all for millennia. Nineteenth-century chemical analysis determined it was "just clay," and the practice was abandoned. Twenty-first-century DNA sequencing revealed that the ritual treatment of the clay — covering it with water and letting it stagnate — introduced a fungus producing bioxanthracene B, a compound with genuine antibacterial and antimalarial properties. Our ancestors' ritual procedure was, in effect, a form of antibiotic production.¹⁰

Mercury for syphilis worked along the same lines as chemotherapy — hopefully killing the disease before the cure killed the patient. Mechanical ventilation, invented to combat the (wrong) theory that noxious airs caused disease, inadvertently saved lives. Quarantine Lazarettos appeared on the Mediterranean long before germ theory. The pattern is consistent: culturally evolved practices often work for reasons completely different from the reasons their practitioners believe, and "modernizing" by stripping away the seemingly irrational parts can destroy the functionality.¹⁰

This connects directly to Cognitive Biases — our systematic errors in reasoning aren't just failures of individual cognition but features of minds designed to defer to cultural knowledge rather than reinvent it. It connects to Emergence — cultural evolution is itself an emergent process where complex, functional knowledge arises from simple interactions (copying, occasional mutation, selective retention) without any designer. It connects to the Bicameral Mind theory — if theory of mind itself is a cultural invention that spread through the Bronze Age Near East, then even the framework we use to have a mind is a product of the same cumulative cultural process that produced seal-hunting techniques and manioc processing. And it should make anyone working on Ai Alignment nervous: if human cultural knowledge is this opaque, this resistant to rational reconstruction, what are we losing when we try to formalize it into training data and reward functions?