Sapiens Part 1: The Cognitive Revolution

Standard advice for how to read a book goes something like this:

  • First, skim the entire book, try to get the big picture.
  • Next, read the book quickly, getting a medium-level view.
  • Finally, actually read the entire book carefully.

Sensible enough. The basic theory here is that we need a conceptual framework in which to organize information. This is hard to build if we just consume detail after detail in a linear fashion.

But if that’s true, then why are books always written as detail after detail in a linear fashion? Shouldn’t the book explicitly provide the same content with different levels of detail?

Reading Sapiens by Yuval Harari, I decided to try creating these other levels. This post is my summary of part one of that book. And also a summary of that summary.

Sapiens, Part 1. Summary of Summary

Many different species of humans have existed since 2.5mya. For most of that time, humans were not dominant over other species. The key differentiators were giant brains, walking on two legs, complex social structures, tools, and cooking.

Sapiens evolved in East Africa 150kya, In the cognitive revolution 70kya, Sapiens evolved language. This made it possible to describe complex ideas and to gossip. Gossip made it possible to cooperate in groups of up to 150. Invented myths like religions made it possible to cooperate in even larger groups. These abilities to cooperate made Sapiens more powerful than animals or other humans.

After the cognitive revolution, Sapiens quickly overran Africa, Asia, and Europe, eliminating all other humans. Then, after inventing boats, clothing and snowshoes, Sapiens reached landmasses no other humans had ever seen. They settled Australia 45kya and America 14kya. Wherever they went, most large animals soon went extinct. This happened because those animals had not evolved to avoid humans, and couldn’t cope with the large-scale fires Sapiens set.

We only have a superficial picture of forager life after the cognitive revolution. Foragers lived in intimate, mostly nomadic bands of a few dozen to a few hundred. They needed to master a huge number of skills. Foragers were extremely fit due to varied diets and constant exercise. This era is still our “home” with the diet, exercise, and social structure we evolved for. It was common to die suddenly from childbirth, accidents, or violence.

Sapiens, Part 1. Summary

Here we are, with our plumbing and satellites and quantum mechanics. How did we get here, and why are we like we are?

Big picture, our history is as follows:

  • 13,500,000,000 years ago – The universe comes into being.
  • 3,800,000,000 years ago – Molecules form on Earth to create life.
  • 2,500,000 years ago – Humans evolve
  • 150,000 years ago – Homo sapiens evolve
  • 70,000 years ago – The cognitive revolution.
  • 12,000 years ago – The agricultural revolution.
  • 500 years ago – The scientific revolution.

The first part of Sapiens tells the story of how humans evolved from other animals, how we Sapiens evolved from other humans, and how we took over the planet.

There have been many types of humans (species of the genus homo) for a long time. Humans first evolved in East Africa from other apes 2.5 mya and settled settled Europe, North Africa, and Asia by 2 mya.

Different populations evolved in different areas, adapted to local conditions. In Europe, Homo neanderthalis were bulkier and adapted to the cold. In East Asia, Homo erectus were the first capable of using fire and hunting in groups. These survived for 2 million years. On Flores island in Indonesia, Homo floriensis weighed at most 25kg. In 2010 in a new species, Homo denisova was discovered in a cave in Siberia. At the same time East Africa produced more species: Homo rudolfensis, Homo ergaster, eventually Homo sapiens. Critically, until 10kya, there were many different human species, all around at the same time.

These archaic humans weren’t particularly dominant. They lived in constant fear of predators. Still, there were two major differences to other animals.

First, humans have massive brains. They are only 2-3% of mass, but consume 25% of energy, compared to around 8% for other apes. (Incidentally, neanderthals had bigger brains than Sapiens.) Evolution reduced the size of muscles to “pay” for this energy expenditure.

The value of our gigantic brains today seems obvious. But 2mya the most salient outcomes were simple tools like flint knives and pointy sticks. Was this really worth it? Or were there more subtle benefits? This apparently remains a mystery.

Second, humans stand upright. This allows us to see further and frees arms for stuff like throwing rocks. We developed extra nerves and muscles to allows us to perform intricate tasks. Humans always had simple tools.

Standing has downsides. We arguably haven’t fully adapted to it, hence backaches and stiff necks. Further, standing requires narrower hips, which means extra deaths in childbirth. This selected for women who gave birth earlier. Human babies became undeveloped and helpless. This then contributed to social abilities, since lone mothers couldn’t survive.

Cooking and controlling fire is a first major gap between humans and other species. Various human species mastered fire by 300kya. This provided defense, light, and heat. But cooking was the biggest benefit. It killed microorganisms, allowed a larger variety of foods, and greatly reduced time spent eating. Evolution could spend less energy on teeth and intestines.

These archaic humans were not at the top of the food chain. Yes, they had large brains, tools, and social structures. But they were weak, and constantly afraid of predators. They survived off of plants, insects, small animals. The specialized in eating bone marrow, left over after other animals had killed something and literally picked the bones clean. Only 400kya they started to hunt large animals.

Not until Sapiens did humans climb to the top of the food chain. Harari gives the following poetic and somewhat baffling remark:

Most top predators of the planet are majestic creatures. Millions of years of dominion have filled them with self-confidence. Sapiens by contrast is more like a banana republic dictator. Having so recently been one of the underdogs of the savannah, we are full of fears and anxieties over our position, which makes us doubly cruel and dangerous. Many historical calamities, from deadly wars to ecological catastrophes, have resulted from this over-hasty jump.”

By 150kya Homo sapiens evolved in East Africa. These looked just like modern people, but weren’t cognitively the same. They moved out of East Africa 70kya and quickly took over all Eurasia. (Africa and Asia 60kya, Europe and Australia 45kya, and America 15kya). Traditionally, it was debated if they interbred or replaced the other humans already in those places. In 2010 scientists collected enough neanderthal DNA to compare to modern humans. Europeans probably have 1-4% neanderthal DNA. Aboriginal Australians might have 6% Homo denisovan DNA! So these weren’t quite distinct species.

It’s unclear exactly how the process of Sapiens displacing other humans went. There could have been conflict and genocide. Or, perhaps the other species slowly retreated and were out-bred. Regardless, almost all other humans were gone by 30kya. The drawf-like humans in Flores finally vanished when Sapiens arrived 12kya.

What happened to Sapiens 70kya that caused them to take over the world? The answer is the cognitive revolution: They evolved language and culture.

Before the cognitive revolution, Sapiens had left East Africa, but were driven back by Neanderthals. Then, between 70k and 30kya they invented boats, oil lamps, bows and arrows, needles and art and then took over Europe, East Asia, and Australia.

This all happened because of language. Why did language happen to Sapiens, rather than Neanderthals, or some other species? It seems to be simply because of accidental mutations.

Are Sapien language capabilities really unique? Yes. Other animals can communicate by making sounds. Indeed, some birds are much better at making sounds than you. However, they have a finite vocabulary, things like “hawk!” or “enemy monkeys”. Sapiens can express an infinite number of sentences with distinct meaning.

There are two theories for why language evolved. The first is that it was for compound propositions like “There is a beehive behind the tall bush a five-minute walk down the river”. The second is that language evolved for gossip. Since we are so social, it’s crucial to keep track of what’s going on with all the other relationships in your band. You can’t observe it yourself, so talk to others. Still today much — or even most? — human communication is some form of gossip.

Gossip is powerful. However, the number of relationships in a group is the square of the number of people in it. At some point, our ability to keep track of all the relationships breaks down. This seems to happen around 150 individuals. Still today we have many organizations (family businesses, military companies) that work well up to this size with minimal internal organization. But beyond that, gossip doesn’t seem to suffice.

After language came fiction: Myths, gods, and religions. These really are unique to humans — try getting a monkey to trade you a banana for some virgins in the afterlife. At first, this seems strange. Isn’t it harmful to believe in false things? Why would evolution favor that? In fact, fictions are tremendously useful. They allow cooperation at a larger scale than gossip. For a very large groups to cooperate, there needs to be an agreed set of rules. Kings have no true divine rights. But if we believe they do, then it is possible for countries to exist. Even if kings are greedy, they want to avoid bands fighting each other, which is a huge win. So myths can reduce internal conflict. Similarly, large international corporations don’t “really” exist. But because we believe they do, thousands of people can work together to accomplish things.

Mythical beliefs allow Sapiens to adapt rapidly, bypassing genetic evolution. When convenient we are eerily quick to discard divine kings and adopt human rights. It’s not important if mythical beliefs are *true*. What matters is the behaviors they instill in us.

A Neanderthal would probably beat a Sapien in a 1-on-1 fight. But in a group of hundreds, Sapiens could cooperate much more effectively. Sapiens traded over long distances. Neanderthals never did. Sapiens had hunting techniques involving dozens– for example, surround an entire heard and chase them into a gorge. Neanderthals hunted in at most small groups.

To summarize, what new abilities did language give us?

New AbilityWhat Became Possible
Transmit larger quantities of information about the surrounding world. Plan and carry out complex actions like hunting bison.
Transmit larger quantities of information about social relationships.Cohesive groups of up to around 150 people possible.
Transmit information about myths, like spirits, nations, companies, human rights.Cooperation of groups even larger than 150 people. Also enabled evolution of social behavior, at a rate much faster than genetic evolution.

Here’s a cute way of putting it: Imagine an equal number of chimps will fight an equal number of Sapiens. Neither has any weapons or tools. Suppose the fight is:

  • 3 vs. 3: This fight would probably be a close thing.
  • 15 vs. 15: The Sapiens would be favored. They could make a plan and communicate on the fly.
  • 1000 vs 1000: The Sapiens would be insanely favored. They could invent tactics and movement patterns the chimps would be helpless to counter.

The cognitive revolution is where “biology” becomes “history”. We cannot understand the French revolution in terms of genes and hormones. While biology sets the parameters, the main show is the social evolution of behavior. If you want to understand what happens, you must follow the dynamics of the invented fictions.

What happened after the cognitive revolution? Sapiens quickly spread over Africa, Europe and Asia, displacing all other humans. Then they found previously untouched landmasses, Australia and America.

Sapiens settled Australia 45kya. This is an accomplishment, as it was necessary to repeatedly cross channels, some hundreds of km wide, and then survive in a different ecosystem at the end. Probably Sapiens in Indonesia developed ocean-going boats for fishing and trading, but there’s no proof. In principle, people might have swam to Australia? This is unlikely. No other mammal ever crossed from Asia to Australa.

Then, everything died. Before Sapiens, there were giant birds, lizards that looked like dragons, and kangaroos and wombats that weighed more than 2 tons. Within a few thousand years, 23/24 Australian animals that weight 50kg or more were extinct. There were climactic changes around the same time, but these are probably not the main cause. Ocean animals were fine.

How did all these animals go extinct? There are three possibilities:

1. Maybe the animals were very easy to kill because they didn’t evolve alongside archaic humans for millions of years like the animals in Africa, Europe and Asia did. Giant wombats probably didn’t need to fear any other animals the size of Sapiens. Genetic evolution was too slow to save them.

2. Maybe humans transformed the environment using fire agriculture. It was common to burn forests to create grasslands– much better for hunting. Eucalyptus trees are robust to fire. These were rare before Sapiens came, but common after.

3. Maybe the animals were very unlucky to hit climate changes and Sapiens at the same time. Perhaps they could have adapted to one or the other, but not both at the same time.

Sapiens got to America around 16kya. This was even more impressive. Yes, they walked over from Asia because sea levels were low. But they had to adapt to the extreme conditions on Northern Siberia, where no other humans ever survived. They had to invent snowshoes, clothing, new weapons and hunting techniques. They couldn’t settle south of Alaska until around 14kya when global warming melted glaciers. By around 12kya, Sapiens had settled the southern tip of Argentina. This is incredibly rapid — moving south, they had to adapt over and over again to different climates.

Again, after Sapiens got to America, most of the large animals died. There were originally Mammoths, mastatons, bear-sized rodents, giant lions and cats, herds of horses and camels, and 6m tell 8 ton ground sloths. Overall, the Americas lost 84/107 large mammals. Many of these had lived for tens of millions of years, but were extinct within 2k years of Sapiens’ arrival. Some survived on Carribean islands for a while, but then disappeared after settled those islands.

Madacascar was only reach 1.5kya, Same story.

There are tons of islands throughout the world reached at various times. Same story.

What was life like after the cognitive revolution?

Why do must of us eat too much high-calorie food? It’s generally agreed that’s because we had little access to sweet food in our evolutionary history. You should gorge of ripe fruit if you almost never find it.

It’s hard to learn much about forager life. For example, were our forager ancestors monogamous or communal? Was there private property? Were there nuclear families? Was there philosophy or music?

Why is it so hard to know much? One major difficulty is that artifacts from wood, bamboo, and leather rarely survive. We mistakenly think of the “stone” age. In any case, foragers moved frequently, carrying everything on their backs, so they surely had few possessions anyway.

Another strategy is to study modern foragers. However, this is misleading. First, they are all influenced by neighboring societies. Second, the surviving forager cultures are mainly in areas very poor for agriculture, such as harsh climates. This would greatly mislead as to, say, population density. Third different forager societies are different from one another. Aboriginal australian clans had hundreds of different languages, religions, customs.

What do we actually know?

Animals. The dog was domesticated by 15kya, and perhaps much earlier. Dogs were used for fighting, hunting, and as an alarm system. Dogs probably evolved “accidentally” since those that best met human needs and/or manipulated humans got extra care and food. There were no other domesticated animals.

Society. People lived in intimate bands of a few dozen to a few hundred. There was no privacy or loneliness. There was some contact between different bands, but it was minimal. Most never saw someone outside their band for months on end. There was only trade in “prestige” items like pigments, not in things like fruit or meat.

Movement. Bands roamed back and forth over the same territory of dozens or hundeds of square kilometers. They occasionally explored new lands, due perhaps to conflict, demographic pressure, or the environment. If a band split every 40 years and moved 100km, you’d get from East Africe to China in 10k years, roughly what happened. Rarely, with great food sources, you’d have seasonal or even permanent camps.In particular, there were fishing villages.

Minds. A huge amount of knowledge was needed: What you can eat, what is poisonous, what is a cure, the progress of seasons, how to deal with animals, how to deal with weather, how to make weapons, traps, or clothes. Sapiens mastered techniques for drying, smoking, freezing food. Everyone needed to know basically all of this, whereas today we can hyper-specialize. There is some evidence that our brains have *shrunk* since the age of foraging.

Bodies. The forager diet provided ideal nutrition and exercise. This, after all, is what we adapted for. The diet was highly varied, which provided all nutrients. This was helpful when one source failed. People had incredible physical fitness due to constant exercise. Life expectancy was 30-40 years, though some made it to their 80s. There were few infectious diseases. Most of these came from domesticated animals and only emerged after agriculture. However, there was high child mortality. Any random accident could kill you. It was common to abandon or even kill the old or disabled, or unwanted babies.

Violence. It’s hard to estimate how violent this time was. We can do things like study skeletons and see how many clearly died due to violence. This has been done in different places, giving estimates of 0.25%, 4.5% or 40% ! (This misses soft-tissue damage, so is surely an under-estimate.) So violence was probably very common in some places, and less common in others.

Spirituality. Most believed in various types of animism, that everything is alive and has feelings, including things like rocks. But we don’t know much. (Incidentally, are you familiar with Panpsychism?)

This is the age we evolved for. It provided the diet, exercise, and social structure that is still “home” for us. But you could die at any time from violence or a random accident. We don’t know a tremendous amount about how societies worked, or how common violence was.

Then, we started farming and pretty much everything got worse.


What does washing your hands actually accomplish?

(Probably a lot?)

Disease experts are constantly telling us how important it is to wash our hands. But how much does it actually matter? This is surprisingly hard to answer.

In handwashing’s favor: We have very good reasons to believe it should work. Soap destroys the integrity of cell membranes, killing them. So regular soap kills bacteria and fungi. It doesn’t kill viruses but helps wash them off your hands.

Against handwashing: There are quite few randomized controlled trials that actually test the hypothesis of “if ordinary people washing their hands more they get sick less”. And those trials that exist are tricky to interpret.

What seems to be the most ambitious trial by far took place outside Karachi in 2002-2003. They assigned 300 households each as controls, to regular soap and to antibacterial soap. They visited the soap households weekly to give out soap, and to encourage people to wash their hands.

They found that in soap households, children younger than 5 were ~50% less likely to get pneumonia or diarrhoea, and ~33% less likely to get impetigo (a skin infection). Here’s a chart for children under 15:

There are various other reviews of different studies. These typically find reductions in illness between 20% and 33%, depending on the disease and if they are studying high or low-income countries.

At first glance, 33% seems significant, but not overwhelming But I think two factors are missing from these studies:

  • Limited effectiveness of the intervention. These studies basically give people soap and encourage them to use it. How well does this work? I don’t know, but I bet a study that forced participants to wash their hands at gunpoint would have larger effect sizes. All the studies are really measuring is “how effective is repeatedly asking people to wash their hands”? But you aren’t deciding if you should be exposed in such a campaign– you are actually deciding if you should wash your hands.
  • Hand washing in the baseline. Just as these studies surely don’t take people to 100% handwashing, many in the control populations are already washing their hands. If we really wanted to check how important it is to wash hands (and we were horrible) we would force a group to avoid all hand washing. This would surely make hand washing look even better.

Then there’s another issue:

  • Externalities. When you wash your hands, you don’t just prevent yourself from getting sick, you prevent others from getting sick. Some studies try to account for this. The Karachi study made sure that the “no soap” and “soap” groups were all clustered in different neighborhoods. But, still, they were still only reaching a subset of the population. Other studies don’t deal with this at all. If you’re altruistic, this is probably another major factor. (It’s possible in principle that certain diseases could be eradicated just by handwashing if that drove transmission rates low enough.)

How much should we increase our estimates based on these ideas? It’s not totally clear, but it seems like handwashing would look a lot better if it were possible to fully account for them.

In terms of personal choices, it probably doesn’t matter. I’d say they push careful handwashing into the territory of something you’d be crazy not to do if you might be exposed to pathogens.

Society: Web of trust or sticks pointed everywhere?

People are strangely decent.

Like all good pessimists, I think I’m not not a pessimist– I’m just realistic. Perhaps because of that, I am impressed by how not-totally-evil humans are. If you assume that evolution is war, and we are the product of millions of years of evolution tuning us towards perfect selfishness, there’s a lot of odd stuff to see.

  • Say you go attack a random person on the street, and they start screaming for help. Random strangers may very well put their own safety at risk to help them. Certainly, they will consider responding, and feel guilty if they don’t.
  • Most of us face huge numbers of opportunities to steal. (From work, by cheating with automated checkout machines, at acquaintances parties, etc.) While there is plenty of theft, people also pass up many opportunities. If people made a cold-blooded calculation of “is the value of this good greater than the expected penalty?”, there would be much more theft.
  • Many fast-casual restaurants ask customers to clean up after themselves. There is no enforcement. Yet, most people do as asked.
  • Most people don’t like to lie, even to total strangers that they will never see again.

People appear to be constantly do things against their self-interest. Often, these behaviors are beneficial to society at large: they pro-social.

We evolved in small-scale societies with repeated interactions

From an evolutionary psychology view, these pro-social behaviors are a bit puzzling. Why do people behave like this? One answer is that we evolved in different circumstances. Take a small hunter-gatherer clan of 50 people. Just by watching each other and gossiping, these people can easily track how pro or anti-social the behavior of each member is. Social status can then be used to punish anti-social behaviors.

Thus, small-scale societies are good at making it in everyone’s interest to behave pro-socially. (If getting caught lying would lead your entire clan to scorn you… don’t lie.) In such circumstances people would evolve to be pro-social.

Modern people in large cities are still surprisingly non-evil.

In modern life, we face very different conditions. In a city of a million people, you can defect over and over again against strangers with no real harm to your reputation. Thus police, you say, and I agree. But there are still a huge number of defections that aren’t penalized. You can litter. You can cut in line. You can spill your soda on the floor of the subway. You can play your crappy music on a speaker on a crowded beach. Some people do these things. But they mostly avoid them.

Despite the weaker incentives towards pro-social behavior in modern life, people are still pro-social far beyond their incentives.

Is there evolutionary pressure towards behaviors that would make modern society impossible?

This worries me a little. If:

  1. society’s functioning depends on people doing the right thing most of the time, and
  2. we do the right thing because we evolved in circumstances where that was in our interest, and
  3. we no longer live in such circumstances,

then shouldn’t the current evolutionary pressure be away from those pro-social behaviors? Will we all inevitably become lowlifes who screw over strangers whenever we can? And might that mean that society can no longer function?

Counterargument 1: Common-sense morality ain’t that great.

One counterargument is that the above paints common-sense morality in far too positive a light. While it may be true that people were relatively good towards neighbors in small-scale societies, they were usually happy to murder people from other tribes whenever convenient. Similarly today, common-sense morality leads us to care more about the things that we happen to see in front of us. This may lead us to making grotesque moral errors by disregarding huge numbers of people we don’t see (e.g. in other countries), or issues that are less apparent to us (antibiotic resistance, existential threats).

Counterargument 2: Society is the way it is because it works.

Another counterargument is that society is like it is for a reason. Suppose you decided to create a new fast-casual restaurant. The difference is, rather than just cleaning up your table when you are done eating, you should also operate the cash-register for yourself. This should be great — the restaurant can save money on cashiers and lower prices! Obviously this doesn’t exist because it wouldn’t work.

Maybe people aren’t systematically pro-social now, they are only mildly pro-social in certain limited circumstances. Society has evolved to exploit those behaviors where it can. If people weren’t pro-social in those ways, society would develop enforcement mechanisms, just like it has to our many existing imperfections.

Which society do we live in?

The question this leaves me with is, which of the following correctly describes society today. Both of the following seem plausible to me:

  • (The “society is held together by good actors” view.) Society is able to function because people mostly do the right thing. Yes, people lie, cheat, and steal. Yes, politicians are corrupt. Yes, we need police and tax inspectors and annual performance reviews. But these bad behaviors are limited. Most police could abuse their power more with little penalty. Tenured professors could neglecting teaching much more than they already do. Even corrupt politicians make some effort to choose good policies. At the end of the day, it’s just not possible to monitor everyone all the time. Fundamentally, society has no choice but to take people, give them responsibilities, and hope that they mostly do the right thing. Fortunately, they typically do.

  • (The “sticks pointed everywhere” view.) “Society” is another word for “all the tricks we’ve created to get us to stop defecting against each other all the time”. A few random exceptions aside, people usually do screw each other over given the chance. The non-defectors get bred out. Small-scale societies developed simple institutions to align individuals’ interests. Modern societies have come up with more and better tricks. There is no limit to our ability to improve these tricks. As we get better and better, evolutionary pressure will be ever more in the pro-social direction. The 20th century offered a clear demonstration of this in the triumph of markets over socialism.

Which of these is true, I don’t know.

It’s important to remember that while the first view might sound nice, it’s the grim view. If our decency is a remnant of our evolutionary past, evolution will soon “fix it” and civilized life shall perish from the earth. On the other hand, if everyone is already mostly screwing each other over whenever they can, then great — we can work on finding better mechanisms to promote cooperation.

Parfit Chapter 1: Rationality and Consequentialism Might Eat Themselves, But Let’s Not Make a Big Deal About It.

It’s impossible not to love a book that starts like this:

We are particular people. I have my life to live, you have yours. What do these facts involve? What makes me the same person throughout my life, and a different person from you? And what is the importance of these facts? What is the importance of the unity of each life, and of the distinction between different lives, and different persons?

That’s Derek Parfit’s Reason’s and Persons. This book is interesting, but difficult. That’s partially because the ideas are challenging, and partially because the book is written the way academic philosophers write books.

As an experiment, I’ve tried to extract some of the insights, and cast them in a more accessible way. This might be a failed experiment! I probably misunderstand some things, and trying to simplify the ideas might render them meaningless. You can judge this for yourself. Anyway, here’s my attempt at a summary of Chapter 1.


Self-interest theory (S) says that it is rational for a person to try to make their own life go “as well” as possible.

Consequentialist theory (C) says that each person should try to make “the outcome” in the world as good as possible.

There are different versions of these, depending on what “go well” and “the outcome” mean. S needn’t say to be selfish in the usual sense — a good life might include the welfare of loved ones. Similarly, the “outcome” in C can vary. Utilitarianism tries to maximize happiness minus misery. Other versions of C might consider how equally these are divided among people, or include principles like honesty.

This chapter takes these theories for a test drive. It considers some unusual situations, and asks what these theories say to do in them.

Kate the Writer

Kate is a writer, passionate about her books. She does not believe S. She works very hard, which she believes is a sacrifice of her own happiness. She writes so hard and so long she eventually collapses into exhaustion and depression.

Suppose we change Kate to believe S. Since she thought she was hurting herself by working so hard, she chooses not to do that anymore. She doesn’t find her work and life as meaningful as we used to.

Have we made her better off?

Takeaway: Belief in S may reduce someone’s circle of concern and thereby make them less happy.

The Desert Hitchhiker

Sam believes in S, and always does what’s best for him. He is is also unable to lie. He is driving through the desert when his car breaks down. Fortunately a stranger stops, and offers him a ride into town for $20. Sam would be thrilled to pay, but has no money in his wallet. Sam tells the stranger, “I don’t have money now, but I will pay you when we get to my house”.

But suppose the stranger gives the ride. After getting to Sam’s house, there would be no way to make Sam pay. At that point, Sam would to what’s best for him and refuse to pay. Since Sam is unable to lie, the stranger realizes this and leaves Sam out in the desert.

Would Sam be better off if he had not believed S?

Takeaway: There are situations where the goals of S are best achieved by making yourself behave in ways that S says are wrong.

Another Takeaway: I’ve talked to many people about this problem. Most respond that it’s trivial — the right answer is to be “overall” rational on a long time horizon. Perhaps. But if Sam got the ride, upon arrival at is house it would be irrational for Sam to pay. So if you want to use this defense, you need some definition of “rational” that is not Markovian. Good luck with that.

Schelling’s Answer To Armed Robbery

A psychopath breaks into your house. The police are 15 minutes away. The man orders you to open your safe full of gold. If you don’t he will kill your children.

If you give him the gold, he may kill everyone so that you can’t tell the police about him. But if you ignore him, he will probably hurt a child to show you he is serious. What should you do?

Fortunately, you are an expert on conflict strategy, and happen to have a special drug on hand, which you take. Suddenly you don’t mind in any way if your children are killed, or if you are tortured. The man threatens to kill your daughter, and you say “whatever.”

The man no longer has any power over you. His threats mean nothing against someone so irrational. In your drugged state, you will probably not remember what he looks like. His best bet is to immediately leave so as to minimize his chance of being caught by the police.

Was it rational for you to make yourself irrational?

Takeaway: If you predictably take the actions in your own self-interest, other agents can exploit this. The defense is to make yourself irrational.

The Firefighting Pact

You are self-interested. It will be very dry and hot tomorrow. Your neighbors are worried about fires and convene a meeting. They propose that everyone swear to help out if anyone’s house catches on fire. It’s annoying to put out a fire, but much more annoying to have your house burn down.

Your neighbors happen to have a highly accurate lie detector. Suppose a neighbor’s house were to catch on fire. At that point, it would be rational for you to flake on the agreement. After putting you into the lie detector, your neighbors realize they would not benefit from allowing you into the pact. They invite you to step outside and lock the door.

According to your own self-interest theory, would it be better for you to transform yourself into an “irrational” trustworthy person?

Takeaway: This is a reasonable explanation for why we evolved to be (somewhat) loyal and trustworthy. You can substitute repeated interactions for the lie detector if you like.

The Transparent People

You are part of a group of rational, self-interested and transparent people. You live together on an island, eating coconuts.

Tired of working so hard, Alice builds a machine. This machine will make her irrational in a carefully chosen way: she is rational except when it comes to fulfilling any threats. (Apparently, building such a machine is easier than just gathering the damn coconuts.)

After running the machine, Alice announces to the group “I will not be gathering any more coconuts. Either you gather coconuts for me, or I’ll burn down all the coconut trees.” It is now rational for all the other people to do Alice’s work for her.

Were Alice’s actions rational?

Takeaway: Suppose that people refused to give Alice coconuts. At that point, it would not be be rational for Alice to burn down the trees. (She would starve!) So Alice’s gambit stems from the ability to do irrational things.

Another takeaway: To avoid Alice’s tyranny, you should have foreseen this situation. You should have constructed a machine that made everyone threat-ignorers — rational except that you ignore threats even when it is against your interest to do so.

Too Many Do-Gooders

Currently, people derive great happiness and meaning from certain “selfish” desires like their kids or enjoying dinner. The desires are not agent-neutral, and so not compatible with C. One day, aliens drop a virus onto the planet that transforms everyone into pure “do-gooders” only focused on the average good in the world. In order to achieve this, the aliens needed to reduce most of these desires.

Would those aliens have done the best thing for the planet?

There is a tradeoff: all the happiness people get from those desires might be lost. This could be better than the current world, on net. However, total happiness might be higher if many people were left with some of these desires, so they were pretty good but not pure do-gooders.

Takeaway: There are situations where the aims of C are best achieved if most people believe something else instead of C.

Clare’s Child

Clare loves her child. She has the chance to spend $20 buying her kid a wonderful dinner, or that same $20 buying a stranger a cure for a horrible disease. She buys her kid dinner.

You point out to Clare that this was wrong– the benefit to the stranger is much greater. Clare agrees this was wrong but says “Given how much I love my kid, it was impossible not to do that. It would be wrong to make myself love my kid. So while it was wrong, I can’t be blamed for doing it”.

Do you blame Clare?

Takeaway: This assumes it’s impossible for people to act contrary to their strongest desires. If that’s true, then “blameless wrong-doing” exists. If it’s just really hard to act against your strongest desires, then “low-blame wrong-doing” still exists.

The Obscene Film

One day, a man breaks into your house. He says “You must allow me to film you performing an obscene act, or I will kill your children. I will later use that film to blackmail you into committing minor crimes.” If you make the film, your kids are safe forever. You know that, while you could reject the blackmail, given your real personality, you probably wouldn’t.

Is it right for you allow him to make the film?

Takeaway: ??? Part of the question is if (1): it’s wrong to allow yourself to be induced to do wrong or (2) the wrong only occurs when you actually go ahead and do it. Still, I don’t think I fully understand what Parfit is getting at here.

Esoteric Theories

Suppose we believe in C, but we live in a situation where C’s goals are best achieved by us convincing ourselves to instead believe in some other moral theory D. We probably need to forget that we used to believe C. (Otherwise how could we accept the new theory?)

Years pass, and circumstances on earth change due to technology. Now C says that it would be best to adopt some other moral theory E. Because we forgot about C, we can’t do this.

On the other hand, suppose that a small number of people secretly kept the flame of belief in C alive. When circumstances change, they reveal the supreme truth of C. After everyone converts back to C, they realize that they should now convince themselves to believe in E. A small number are nominated to remember C, and to reveal it only when circumstances change again.

Takeaway: This is a defense of C. Even if C says most of us should believe something else, there are good reasons that some people should remember C.

Murder and Accidental Death

Ben is about to die. Right before he dies, he plans to murder Cathy. Meanwhile a forest fire is bearing down on Deb, who will die unless she is rescued. The lives of Cathy and Deb are equally valuable.

You have the power to do one of two things:

  • Convince Ben not to murder Cathy.
  • Recue Deb.

You have a 50% chance of convincing Ben not to do the murder, and a 50.01% chance of rescuing Deb. What should you do?

Takeaway: The question here is if murder is itself wrong, or is it just the effects of murder that are bad.


If you like S or C, these arguments are a bit disturbing. Yet, Parfit’s point does not seem to be that S and C are incorrect. To the contrary, he brings up these issues only to say that they aren’t that bad.

The argument is that a theory being true is different from a theory being practical. S and C never claimed that believing in them would be helpful for achieving their goals. Yes, it’s strange, but it doesn’t mean they aren’t true.

Say that it’s true that S tells you to make yourself believe something other than S. That will then cause you to take actions that are wrong, according to S. Parfit says “OK then, go ahead and change yourself.” S never promised not to tell you to believe something else.

The same story is true for C. Oddly, Parfit gives few arguments that C can tell you to believe in something other than C. Instead he mostly, assumes it’s true. Anyway, he again says “Yes. If you believe C, and that’s what C says, then do it.” He further argues (as in Esoteric Theories above) that it C probably won’t tell everyone to not believe C.

The main purpose of this chapter is actually to defend S and C. Their difficulties are brought up to show that they aren’t fundamental difficulties, but more interesting curiosities.

  • It can be rational to make yourself become irrational. In fact, you almost have do this to prevent other people taking advantage of you. Thus, rationality might tell you to change yourself so that you do things that are wrong, according to rationality.
  • If you believe in Consequentialism, this might mean you should try to change yourself to believe in something else instead. That, in turn, will lead you to do things that are wrong, according to Consequentialism.
  • Just because a theory tells you to believe something else, doesn’t mean it is wrong. Something being “true” is just different from something being “practical”.
  • There’s a perfectly good answer to the “paradox” that a theory might tell you to believe in something else instead: go ahead and do it. The theories never said that you shouldn’t. It’s OK.

PS: I will ostensibly write summaries like this for later chapters. If that’s relevant, you can subscribe by RSS or email below.