The Bonobo and the Atheist does something truly rare, which is to come up with an interesting way of viewing morality. ‘What is right?’ is the most pored-over and divisive question in all of philosophy, the book mostly ignores it. The best parts of it contain zero commandments. Instead Frans De Waal asks ‘Why do we think [what is right] is right?’ (that’s my wording not his).
The book focuses on both chimp and bonobo societies, but the commonalities between their behaviour are such that De Waal frequently swaps between them. The sex-heaviness of bonobo lives seems to be irrelevant to the main thesis beyond putting in perspective that not all ape behaviour is aggressive fighting. This is mostly unimportant because the altruism shown between captive chimps was enough to make this point.
De Waal spends a lot of the book arguing against a few different scientific orthodoxies. These are all the standard pet hypothesis privileging. He also spends a lot of time talking about religion (as was typical for 2013) but I didn’t find those sections interesting so I won’t be discussing them.
The book is fundamentally about evolutionary psychology, which is notoriously tricky and easy to misunderstand, so unfortunately I think it’s necessary to spend the next section laying out definitions. Sorry.
Altruism, Altruism, and Altruism
De Waal uses the word altruism to mean different phenomena. While he describes this distinction he doesn’t codify it in language. I’m going to do that for him:
R-altruism is the giving of resources (like food, aid, etc.) to another, which may or may not improve reproductive prospects for your genes
G-altruism is a a subset of r-altruism which specifically harms your own genes’ prospects for reproduction
E-altruism is a cluster of similar emotional responses similar to the ones in a human brain, which lead to r-altruism. The cluster is somewhat centred on empathy, which De Waal (mostly) uses to mean a negative emotional reaction to another individuals pain that is specific to the type of pain experienced by the other.
These are different types of thing. E-altruism is a neurological pattern, r-altruism is a broad cluster of behaviour, g-altruism is defined by evolutionary fitness.
Examples:
A bee stinging an intruder to the hive and dying is r-altruism because it protects other bees including the queen. But it’s not g-altruism as the bee’s own genes (which it shares with the rest of the hive) are helped out. It’s also not e-altruism as the neuronal processes causing it more closely resemble what humans would call “aggression”, “fear”, or “anger”.
A tribal politician cynically helping another’s popularity in order to win themself a better position after a coup is r-altruism, but it’s not g-altruism as it clearly helps their own reproductive success, and since the motive was purely self-benefit it’s not e-altruism either.
A soldier sharing cigarettes with their comrade is r-altruism. It’s probably also e-altruism. But if the behaviour increases camaraderie and trust, and therefore their fighting ability and chances of surviving the war, it’s not g-altruism.
Raising a stranger’s child is r-altruism, g-altruism, and e-altruism. This is perhaps the archetypical example
Causes and Causes
As this book discusses evolutionary biology, it’s also worth thinking about proximate versus ultimate cause-ness.
Proximate causes (p-causes) occur within an organism’s lifetime, including things like “instinctual” and “learned” responses.
Ultimate causes (u-causes) are evolutionary causes. Often these are conflated because they’re similar in phrasing.
Examples:
If a herbivore runs away from a predator, the proximate cause of running is “the predator”, and the ultimate cause of the running is also “the predator”. But to be more specific, the proximate cause of running is “Neurons in the herbivore’s brain pattern-matched to this predator and activated the running process downstream” whereas the ultimate cause is “Herbivores which run from objects which resemble this predator have more offspring on average than ones which don’t.”
From the bee-sting example. The proximate cause of a bee stinging an intruder is aggression/anger/fear hormones in its nervous system. The ultimate cause is that beehives full of bees which defend them most effectively go on to produce more new queens and drones, and all the bees in a hive share a great deal of genetic information.
The Book’s Core Claims
The main thing I get from the book is the following set of claims about the evolution of morality:
Receiving r-altruism is beneficial
Social organisms will evolve reciprocal r-altruism (From 1)
Under certain reciprocity conditions, more r-altruism is generally advantageous genetically and is selected for
Infant learning (or perhaps other things) lead to mirror neurons
The most efficient way to implement r-altruism in a brain which already has the above other-modelling circuits is with empathy
E-altruism p-causes g-altruism but this is outweighed by the evolutionary benefits of r-altruism
Moral preferences over actions not affecting us arose from e-altruism as a result of the benefits of group cohesion
Claims 1-5 discuss why e-altruism could have evolved at all. Claims 6-10 cover the specifics of why it evolved the way it did, and how “morality” evolved. This presents the story of g-altruism as a case of the general inner-alignment failure in animal brains, one driven by energetic constraints on primate brains. Empathy is just the most efficient way of implementing the sort of r-altruistic behaviour which is beneficial in reciprocal social societies.
Claims 1 and 2: Reciprocal r-altruism
These are basic tenets of the evolution of behaviour. Observations of the world and modelling of various cooperative strategies show that social organisms evolve reciprocal r-altruism even when they do not live in kin-related groups.
Claim 3: More r-altruism is generally advantageous under certain circumstances
In an evolutionarily “stable” scenario, we’d expect the optimal level of r-altruism to be around the average of the population, same as for other traits. If this weren’t the case then the population would evolve to be more (or less) r-altruistic.
The important part is where the equilibrium population level of r-altruism lies. It probably depends on how many social interactions occur, and how complex they are. Vampire bats have simple “grudger” cooperation in the prisoner’s dilemma-like game of food sharing, because they only roost together and share food, and don’t engage in more complex social behaviours.
Chimpanzees and bonobos have social structures involving political alliances and hierarchies, for which reciprocal r-altruism is important at all levels. De Waal gives behavioural observations containing evidence that grooming, food sharing, etc. are important for apes to maintain allies and therefore power. This likely makes the equilibrium level of r-altruism in such a population much higher.
Claim 4: E-altruism p-causes r-altruism in nonhuman apes
Based on claim 3, the ancestors of chimpanzees, bonobos, and humans experienced a period of evolution where r-altruism was positively selected for. The question is whether or not this r-altruism was caused by e-altruism.
By this I mean whether or not the neurological machinery for r-altruism in apes ~10 million years ago was similar to the neurological machinery for r-altruism in humans today (which we’ve called e-altruism).
Frans De Waal has an excellent take on evolutionary psychology which I appreciate greatly:
I personally adhere to a different law of parsimony, according to which, if two closely related species act the same under similar circumstances, the mental processes behind their behaviour are likely the same, too.
I will go into more detail later but I think nonhuman apes have empathy.
Claim 5: R-altruism can u-cause e-altruism
Here we complete the full-circle of evolutionary biology. Certain social structures + reciprocity cause r-altruism to be advantageous. R-altruism is caused by e-altruism in the ancestors of apes. The genes for e-altruism are selected for.
Claims 1-5 are the short version of the story. They explain how e-altruism might emerge in a population. The rest of the claims cover why e-altruism is more likely to evolve than other p-causes of r-altruism, and what happens after.
Philosophy
I’ll take a short interlude to point out that this book is, somewhat, a work of philosophy. While it makes no grand moral judgements, it does attempt to supply a basis for morality. For the author, morality is almost always predisposed to the two Hs of Helping or at least not Hurting others. Morality is here also presented as a product of evolution, both up to humans, and in humans. In humans it can also be a product of cultural evolution. This brings him to the following conclusions:
Morality is not universal
Introspection is fundamental for moral reasoning
Emotions are more important than reason as a foundation
This is a wholly descriptive attitude to morality, which is very interesting as morality itself is prescriptive.
De Waal does dip into prescriptivism sometimes though, and he does this somewhat inconsistently. There is a blurring between describing how apes evolved empathy to drive our behaviour, to judging moral philosophies based on his own views, or based on empathy. He doesn’t think the is/ought distinction is something to worry about, gliding between “this is how morality evolved” and “this is what is moral” without being clear about it. I dock him points for this lack of clarity.
Frans De Waal is very unhappy with utilitarianism. He describes the “maximum happiness principle” which I don’t think anyone subscribes to anymore, at least nobody I know would follow his suggestion of putting Prozac in the water supply.
He takes a dig at the core effective altruism philosophy of not truly feeling the scale of your actions, and yet doing the maximum good anyway. He likens this to being a loving wife to someone out of a sense of duty, rather than love.
He mentions the ingroup/kin bias of empathy in a positive light, giving examples such as Peter Singer paying private aides to take care of his elderly mother, or Jean Valjean bringing bread to his starving daughter rather than giving it to children in the street. Both of these he considers to be examples of good.
This reminds me of the Quaker principle of morality as an inner light. Empirically this has a pretty good track record, but it doesn’t address what to do when someone genuinely thinks the right thing to do is go on a crusade, empathy be damned. De Waal’s method of empathy-based morality also tends to fail us when we must deal with outgroups, or in far-mode thinking.
I think De Waal is mostly interested in discussing morality with people who are already (mostly) good. I don’t think he has a particularly strong interest in telling others how to be good in the first place, or in deep logical analysis of morality, so the philosophical parts of this book are some of the weakest. It is probably also the case that his meta-ethical preferences are different from mine, so what I see as bugs in his moral code might be features to him.
Back to the Claims
Claims 6 and 7: Other modelling and mirror neurons
These claims are about characteristics seen as pre-adaptations to e-altruism via empathy. Pre-adaptations are traits which are useful in and of themselves, and also make it easier to evolve a different trait. For example grasping hands and mobile shoulders are good for moving through trees, but make it easier to evolve tool usage.
Since other modelling and mirror neurons are suggested as pre-adaptations it would be good to see evidence of both of them in monkeys, which implies their presence in the last common ancestors of monkeys and modern apes.
Social monkeys can predict the behaviour of others. De Waal gives the example of subordinate males in a troop of macaques mating with females while the dominant male isn’t present. There is evidence of varying degrees of “depth” in other-modelling amongst monkeys. I’m confident that social monkeys model one another.[1][2]
Mirror neurons are also documented in monkeys. De Waal suggests they are important factor for infant learning.[3]
Whether or not De Waal is correct about the u-causes of other modelling and mirror neurons, it seems very likely that the ancestors of apes had both other-modelling and mirror neurons.
Claim 8: E-altruism is the “best” implementation of r-altruism
This is the claim in the book for which the evidence is weakest. I can forgive this since it’s basically impossible to test empirically. It’s also one of the most interesting claims.
The argument is the following: when you have mirror neurons and are already modelling others, the easiest (in the sense of the most likely to evolve) way to cause generally r-altruistic behaviour is to plug your models of others’ emotions directly into your own reward circuitry.
There are other ways to implement r-altruism. Some sort of Machiavellian reciprocity, especially with the starting-point of helping (I’ll scratch your back until you refuse to scratch mine) also works. De Waal mentions that chimp r-altruism is highly dependant on reciprocity. If a chimp A gets in a fight with chimp B, and their friend chimp C doesn’t aid them, chimp A will often ignore chimp B to punish chimp C instead. He implies that bonobo society is less reciprocal in this sense, but does not state it explicitly.
Also central to the implementation of e-altruism is the “[r-]altruism feels good” hypothesis. This is the idea that apes (including people) enjoy helping others, and is perhaps De Waal’s most important claim:
In academic circles, too, it is impossible to avoid the image of out-of-control animals. This is critical in relation to moral evolution, because the opposite of morality is that we just do “what we want”, the underlying assumption being that what we want is not good.
Empathy provides a means for r-altruism to be pleasant. A really good investigation of empathy vs Machiavellian reciprocity would compare the emotional effects of helping low-ranking vs high-ranking individuals. This would help to test whether our emotions obey the “kiss up, kick down” rule. This hasn’t been done.
Claim 9: G-altruism from e-altruism
If e-altruism makes us feel good when we help others, then when helping (i.e. being r-altruistic towards) others is easy, people will help others. De Waal reports a number of cases of g-altruism amongst apes that he’s observed (and even monkeys) such as helping out older members of the community who are too infirm to make a good political ally, or a male carrying the infant of a low-ranking female around, when the female was injured and unable to carry the infant herself.
I think De Waal’s belief is that empathy is in some sense “by default” switched on for humans and apes (at least towards others of our own species), and only under certain circumstances (like conflict or perceived “cheating”) is it turned off.
Humans practice g-altruism regularly. Whether or not cases of g-altruism in apes are common is unclear. The hypothesis of g-altruism as a spillover of e-altruism makes sense. The question is why the existing off-switch for empathy isn’t hasn’t evolved to stop such cases.
There are costs for implementing complex algorithms in the brain. And the “optimal” algorithm to prevent all g-altruism might well be too complex. The algorithm would have to distinguish between those who genuinely are too weak to ever help in future, those who are temporarily ill, those who are weak but have stronger allies, juveniles, etc. and this might just not be possible. The costs of such an algorithm might outweigh the benefits of the lack of resources given.
There might be a sort of time limit for sufficient levels of Machiavellian empathy-deactivation to occur. Once enough individuals in the population have sufficient care for weaker individuals and a tendency to punish third-party actions (see claim 10), then being unempathetic to the weak could be punished too much to be advantageous, even without complexity costs.
This part, like claim 8, has a lot of “might” and “could”. This is one of the areas which I’m still slightly confused about. I have a model which can explain why the costs of g-altruism are outweighed, but I don’t know if this is the only route evolution could have taken. If I saw another species evolve e-altruism but then take the Machiavellian route (of evolving highly complex empathy-deactivating algorithms to prevent any g-altruism), I wouldn’t at all be surprised. I would be surprised if it turned out 95% of species with empathy-like emotions ended up evolving highly complex off-switches for it to prevent the majority of g-altruism.
Claim 10: “Morality” from e-altruism
De Waals refers to one-on-one morality as a social code which determines how individuals treat one another directly, and community-level morality as how individuals think about the rules as applied to others. He also suggests that one-on-one morality is the result of a mixture of empathy and physical enforcement by third parties. I think this makes no sense by his own definition so I’ll ignore his distinction, although I may have misunderstood his position.
Group hierarchies amongst chimps are maintained by physical force, even when that force isn’t currently in use. When politics are involved, one must care about harm done to one’s allies, so successful individuals will care about conflicts in which they are a third party. I think this is sufficient pre-adaptation for further cognition relating to conflicts involving only others.
De Waals suggests that group cohesion also makes it advantageous for individuals to enforce rules or mediate fights. It is apparently quite common for older females or former alpha males to spend time mediating between fights in the group. Tit-for-tat fairness concepts plus empathy provide a pretty reasonable mechanism for this all to evolve.
Experiments and Epistemology
Animal behavioural research is difficult from an epistemic perspective. Humans are deeply biased towards empathizing and projecting. For an example consider the chimp “sign language” experiments.
When a researcher has spent their whole life studying the animal in question it’s unsurprising that they’ll be fond of their subjects (it would be a bit sad not to be). This makes it reasonable to suspect bias!
On the other hand, you can go too far. De Waal recalls a student of his defending their thesis on reconciliation behaviour (chimps making up after fights). Despite being rat psychologists, the reviewers claimed with certainty that no such behaviour could have taken place. When he offered to take them to the zoo to observe the animals (this was the ’70s, so no video recordings) they replied: “What good would it do to see the actual animals? It will be easier for us to stay objective without this influence.”
I assume most criticisms of his research do not fall into self-parody in this way, but maybe he really does live in a world of cartoon behaviourists.
R-altruism at all?
De Waal spends a while showing his own evidence for apes being r-altruistic at all. Big chunks of the book are direct accounts of ape behaviour clearly meant to give the impression of them having some humanity. But he also includes a number of studies, including plenty that he’s done himself.
Chimps generally show r-altruism in “giving assistance tests” where they’re willing to help out a second individual (human or chimp) by giving them a tool which is out of reach to the second subject.
On what are called “prosocial choice tests” their results have been mixed. These consist of giving the test subject the choice between rewarding themself, or to reward themself and another individual of the same species. De Waal has gained positive results (apes often take the prosocial option) whereas previous studies did not.[4]
De Waal claims this is due to the previous studies using complex devices which the chimps couldn’t understand. His studies used a token exchange system, in which the chimps handed a human researcher a red token to reward just themself (the antisocial choice), or a green token to also reward a chimp in an adjacent enclosure (the prosocial choice).
Interestingly, more dominant chimpanzees chose the prosocial option more often, and that begging behaviours from the recipient chimp decreased the frequency of the prosocial choice. This is evidence against claims that chimp r-altruism occurs only under threat.
His study only included female chimps from the troop under investigation. I don’t know why this was done.
I think his claims about chimps exhibiting r-altruism under these circumstances are correct. They square with numerous food-sharing observations of wild chimpanzees, and with the results of giving assistance tests. Several types of monkey have also been shown to have a prosocial bias in prosocial choice tests.
Empathy in Apes
There are a couple of examples which the author uses to sell the concept of empathy in apes. One is of a female chimp who, upon seeing an old, ill chimp leaning against a concrete wall, pushed some bedding behind him in order to cushion his back. As this was the first time he had shown any sign of illness, De Waal claims that this required her to model how she would feel in his situation.
The second is that bottle-reared bonobos (a common thing at sanctuaries due to poachers killing adults for bushmeat) will often play with bottles and water to bottle-feed their own offspring, and other juveniles, in addition to nursing. Crucially to the claim of empathy, they take care not to put too much water into the young bonobo’s mouth. This could be because they can imagine that too much water in the mouth is unpleasant and could cause the juvenile to choke.
Both of these seem like relatively weak evidence, but based on De Waal’s argument about parsimony, I think it’s likely that apes have something neurologically similar to empathy.
Spot-check
The book was written in 2013, which is late enough that the author should have considered the replication crisis in psychology. Here are some random picks of articles cited in the book:
Aknin et al. “Giving leads to happiness in young children” A study suggesting that various conditions where children give treats to a puppet, or share toys, are happier than when they do not give, or do not share the toy. The most questionable part of this study is the usage of two people rating the children’s happiness based on video recordings as a method of assessing happiness, which the authors claim is standard and reliable. No direct replication has been attempted. The general impression I get from literature is that giving seems to make people happy, although almost all potential studies might be confounded by the fact that being happy makes people give.
Brosnan et al. “Monkeys reject unequal pay” This one is pretty famous, monkeys will do a task (handing researchers a token) when rewarded with cucumber slices, until they see another monkey being rewarded with grapes. Suggests monkeys have a concept of inequity. This has been reviewed here and replicated.
Evans et al. “Chimpanzees use self-distraction to cope with impulsivity” Chimpanzees are presented with a box that slowly fills with treats. They can grab the box at any time but that prevents any more treats from being put in. The longer they can hold themselves back, the greater the reward. In this experiment chimps were shown to be able to hold back longer when given toys to distract themselves with. Seems to be reasonable experimentally. Their data isn’t amazing. No replication has been attempted.
Here are a couple of papers which contradict claims made in the book:
Riedl et al. “No third-party punishment in chimpanzees” This suggests that chimpanzees actually do not punish third-party transgressions. I think this evidence is weaker than expected because they used theft as the “crime”, and chimpanzees do not respect ownership under certain circumstances, such as when the “owner” is away. Most accounts of chimpanzee interventions in third-party scenarios consist of physical aggression.
Jensen et al. “Chimpanzees are rational maximizers in an ultimatum game” Provides evidence against the abstract concept of fairness in chimps, at least to the same degree as in humans. Chimps will generally accept any offer in ultimatum games, without a “spiteful” punishment of low offers.
Epistemic health: significantly above average for a psychology-heavy pop-science book.
Pointing At Morality
I now feel I have a good understanding of the sort of thing that morality is. I feel I can see where it likely came from, and why I feel the way I do about certain things.
I see my own moral values as starting from a bunch of emotional value judgements over a bunch of things. A subset of this is a few strong preferences to things like consistency and universality, and respecting the value judgements of others. With these preferences, and my own reasoning abilities, I attempt to wrangle the rest into a better shape.
[Book Review]: The Bonobo and the Atheist by Frans De Waal
The Bonobo and the Atheist does something truly rare, which is to come up with an interesting way of viewing morality. ‘What is right?’ is the most pored-over and divisive question in all of philosophy, the book mostly ignores it. The best parts of it contain zero commandments. Instead Frans De Waal asks ‘Why do we think [what is right] is right?’ (that’s my wording not his).
The book focuses on both chimp and bonobo societies, but the commonalities between their behaviour are such that De Waal frequently swaps between them. The sex-heaviness of bonobo lives seems to be irrelevant to the main thesis beyond putting in perspective that not all ape behaviour is aggressive fighting. This is mostly unimportant because the altruism shown between captive chimps was enough to make this point.
De Waal spends a lot of the book arguing against a few different scientific orthodoxies. These are all the standard pet hypothesis privileging. He also spends a lot of time talking about religion (as was typical for 2013) but I didn’t find those sections interesting so I won’t be discussing them.
The book is fundamentally about evolutionary psychology, which is notoriously tricky and easy to misunderstand, so unfortunately I think it’s necessary to spend the next section laying out definitions. Sorry.
Altruism, Altruism, and Altruism
De Waal uses the word altruism to mean different phenomena. While he describes this distinction he doesn’t codify it in language. I’m going to do that for him:
R-altruism is the giving of resources (like food, aid, etc.) to another, which may or may not improve reproductive prospects for your genes
G-altruism is a a subset of r-altruism which specifically harms your own genes’ prospects for reproduction
E-altruism is a cluster of similar emotional responses similar to the ones in a human brain, which lead to r-altruism. The cluster is somewhat centred on empathy, which De Waal (mostly) uses to mean a negative emotional reaction to another individuals pain that is specific to the type of pain experienced by the other.
These are different types of thing. E-altruism is a neurological pattern, r-altruism is a broad cluster of behaviour, g-altruism is defined by evolutionary fitness.
Examples:
A bee stinging an intruder to the hive and dying is r-altruism because it protects other bees including the queen. But it’s not g-altruism as the bee’s own genes (which it shares with the rest of the hive) are helped out. It’s also not e-altruism as the neuronal processes causing it more closely resemble what humans would call “aggression”, “fear”, or “anger”.
A tribal politician cynically helping another’s popularity in order to win themself a better position after a coup is r-altruism, but it’s not g-altruism as it clearly helps their own reproductive success, and since the motive was purely self-benefit it’s not e-altruism either.
A soldier sharing cigarettes with their comrade is r-altruism. It’s probably also e-altruism. But if the behaviour increases camaraderie and trust, and therefore their fighting ability and chances of surviving the war, it’s not g-altruism.
Raising a stranger’s child is r-altruism, g-altruism, and e-altruism. This is perhaps the archetypical example
Causes and Causes
As this book discusses evolutionary biology, it’s also worth thinking about proximate versus ultimate cause-ness.
Proximate causes (p-causes) occur within an organism’s lifetime, including things like “instinctual” and “learned” responses.
Ultimate causes (u-causes) are evolutionary causes. Often these are conflated because they’re similar in phrasing.
Examples:
If a herbivore runs away from a predator, the proximate cause of running is “the predator”, and the ultimate cause of the running is also “the predator”. But to be more specific, the proximate cause of running is “Neurons in the herbivore’s brain pattern-matched to this predator and activated the running process downstream” whereas the ultimate cause is “Herbivores which run from objects which resemble this predator have more offspring on average than ones which don’t.”
From the bee-sting example. The proximate cause of a bee stinging an intruder is aggression/anger/fear hormones in its nervous system. The ultimate cause is that beehives full of bees which defend them most effectively go on to produce more new queens and drones, and all the bees in a hive share a great deal of genetic information.
The Book’s Core Claims
The main thing I get from the book is the following set of claims about the evolution of morality:
Receiving r-altruism is beneficial
Social organisms will evolve reciprocal r-altruism (From 1)
Under certain reciprocity conditions, more r-altruism is generally advantageous genetically and is selected for
E-altruism p-causes r-altruism
Reciprocal r-altruism u-causes e-altruism (From 3 & 4)
Social living leads to brains which model others
Infant learning (or perhaps other things) lead to mirror neurons
The most efficient way to implement r-altruism in a brain which already has the above other-modelling circuits is with empathy
E-altruism p-causes g-altruism but this is outweighed by the evolutionary benefits of r-altruism
Moral preferences over actions not affecting us arose from e-altruism as a result of the benefits of group cohesion
Claims 1-5 discuss why e-altruism could have evolved at all. Claims 6-10 cover the specifics of why it evolved the way it did, and how “morality” evolved. This presents the story of g-altruism as a case of the general inner-alignment failure in animal brains, one driven by energetic constraints on primate brains. Empathy is just the most efficient way of implementing the sort of r-altruistic behaviour which is beneficial in reciprocal social societies.
Claims 1 and 2: Reciprocal r-altruism
These are basic tenets of the evolution of behaviour. Observations of the world and modelling of various cooperative strategies show that social organisms evolve reciprocal r-altruism even when they do not live in kin-related groups.
Claim 3: More r-altruism is generally advantageous under certain circumstances
In an evolutionarily “stable” scenario, we’d expect the optimal level of r-altruism to be around the average of the population, same as for other traits. If this weren’t the case then the population would evolve to be more (or less) r-altruistic.
The important part is where the equilibrium population level of r-altruism lies. It probably depends on how many social interactions occur, and how complex they are. Vampire bats have simple “grudger” cooperation in the prisoner’s dilemma-like game of food sharing, because they only roost together and share food, and don’t engage in more complex social behaviours.
Chimpanzees and bonobos have social structures involving political alliances and hierarchies, for which reciprocal r-altruism is important at all levels. De Waal gives behavioural observations containing evidence that grooming, food sharing, etc. are important for apes to maintain allies and therefore power. This likely makes the equilibrium level of r-altruism in such a population much higher.
Claim 4: E-altruism p-causes r-altruism in nonhuman apes
Based on claim 3, the ancestors of chimpanzees, bonobos, and humans experienced a period of evolution where r-altruism was positively selected for. The question is whether or not this r-altruism was caused by e-altruism.
By this I mean whether or not the neurological machinery for r-altruism in apes ~10 million years ago was similar to the neurological machinery for r-altruism in humans today (which we’ve called e-altruism).
Frans De Waal has an excellent take on evolutionary psychology which I appreciate greatly:
I will go into more detail later but I think nonhuman apes have empathy.
Claim 5: R-altruism can u-cause e-altruism
Here we complete the full-circle of evolutionary biology. Certain social structures + reciprocity cause r-altruism to be advantageous. R-altruism is caused by e-altruism in the ancestors of apes. The genes for e-altruism are selected for.
Claims 1-5 are the short version of the story. They explain how e-altruism might emerge in a population. The rest of the claims cover why e-altruism is more likely to evolve than other p-causes of r-altruism, and what happens after.
Philosophy
I’ll take a short interlude to point out that this book is, somewhat, a work of philosophy. While it makes no grand moral judgements, it does attempt to supply a basis for morality. For the author, morality is almost always predisposed to the two Hs of Helping or at least not Hurting others. Morality is here also presented as a product of evolution, both up to humans, and in humans. In humans it can also be a product of cultural evolution. This brings him to the following conclusions:
Morality is not universal
Introspection is fundamental for moral reasoning
Emotions are more important than reason as a foundation
This is a wholly descriptive attitude to morality, which is very interesting as morality itself is prescriptive.
De Waal does dip into prescriptivism sometimes though, and he does this somewhat inconsistently. There is a blurring between describing how apes evolved empathy to drive our behaviour, to judging moral philosophies based on his own views, or based on empathy. He doesn’t think the is/ought distinction is something to worry about, gliding between “this is how morality evolved” and “this is what is moral” without being clear about it. I dock him points for this lack of clarity.
Frans De Waal is very unhappy with utilitarianism. He describes the “maximum happiness principle” which I don’t think anyone subscribes to anymore, at least nobody I know would follow his suggestion of putting Prozac in the water supply.
He takes a dig at the core effective altruism philosophy of not truly feeling the scale of your actions, and yet doing the maximum good anyway. He likens this to being a loving wife to someone out of a sense of duty, rather than love.
He mentions the ingroup/kin bias of empathy in a positive light, giving examples such as Peter Singer paying private aides to take care of his elderly mother, or Jean Valjean bringing bread to his starving daughter rather than giving it to children in the street. Both of these he considers to be examples of good.
This reminds me of the Quaker principle of morality as an inner light. Empirically this has a pretty good track record, but it doesn’t address what to do when someone genuinely thinks the right thing to do is go on a crusade, empathy be damned. De Waal’s method of empathy-based morality also tends to fail us when we must deal with outgroups, or in far-mode thinking.
I think De Waal is mostly interested in discussing morality with people who are already (mostly) good. I don’t think he has a particularly strong interest in telling others how to be good in the first place, or in deep logical analysis of morality, so the philosophical parts of this book are some of the weakest. It is probably also the case that his meta-ethical preferences are different from mine, so what I see as bugs in his moral code might be features to him.
Back to the Claims
Claims 6 and 7: Other modelling and mirror neurons
These claims are about characteristics seen as pre-adaptations to e-altruism via empathy. Pre-adaptations are traits which are useful in and of themselves, and also make it easier to evolve a different trait. For example grasping hands and mobile shoulders are good for moving through trees, but make it easier to evolve tool usage.
Since other modelling and mirror neurons are suggested as pre-adaptations it would be good to see evidence of both of them in monkeys, which implies their presence in the last common ancestors of monkeys and modern apes.
Social monkeys can predict the behaviour of others. De Waal gives the example of subordinate males in a troop of macaques mating with females while the dominant male isn’t present. There is evidence of varying degrees of “depth” in other-modelling amongst monkeys. I’m confident that social monkeys model one another.[1][2]
Mirror neurons are also documented in monkeys. De Waal suggests they are important factor for infant learning.[3]
Whether or not De Waal is correct about the u-causes of other modelling and mirror neurons, it seems very likely that the ancestors of apes had both other-modelling and mirror neurons.
Claim 8: E-altruism is the “best” implementation of r-altruism
This is the claim in the book for which the evidence is weakest. I can forgive this since it’s basically impossible to test empirically. It’s also one of the most interesting claims.
The argument is the following: when you have mirror neurons and are already modelling others, the easiest (in the sense of the most likely to evolve) way to cause generally r-altruistic behaviour is to plug your models of others’ emotions directly into your own reward circuitry.
There are other ways to implement r-altruism. Some sort of Machiavellian reciprocity, especially with the starting-point of helping (I’ll scratch your back until you refuse to scratch mine) also works. De Waal mentions that chimp r-altruism is highly dependant on reciprocity. If a chimp A gets in a fight with chimp B, and their friend chimp C doesn’t aid them, chimp A will often ignore chimp B to punish chimp C instead. He implies that bonobo society is less reciprocal in this sense, but does not state it explicitly.
Also central to the implementation of e-altruism is the “[r-]altruism feels good” hypothesis. This is the idea that apes (including people) enjoy helping others, and is perhaps De Waal’s most important claim:
Empathy provides a means for r-altruism to be pleasant. A really good investigation of empathy vs Machiavellian reciprocity would compare the emotional effects of helping low-ranking vs high-ranking individuals. This would help to test whether our emotions obey the “kiss up, kick down” rule. This hasn’t been done.
Claim 9: G-altruism from e-altruism
If e-altruism makes us feel good when we help others, then when helping (i.e. being r-altruistic towards) others is easy, people will help others. De Waal reports a number of cases of g-altruism amongst apes that he’s observed (and even monkeys) such as helping out older members of the community who are too infirm to make a good political ally, or a male carrying the infant of a low-ranking female around, when the female was injured and unable to carry the infant herself.
I think De Waal’s belief is that empathy is in some sense “by default” switched on for humans and apes (at least towards others of our own species), and only under certain circumstances (like conflict or perceived “cheating”) is it turned off.
Humans practice g-altruism regularly. Whether or not cases of g-altruism in apes are common is unclear. The hypothesis of g-altruism as a spillover of e-altruism makes sense. The question is why the existing off-switch for empathy isn’t hasn’t evolved to stop such cases.
There are costs for implementing complex algorithms in the brain. And the “optimal” algorithm to prevent all g-altruism might well be too complex. The algorithm would have to distinguish between those who genuinely are too weak to ever help in future, those who are temporarily ill, those who are weak but have stronger allies, juveniles, etc. and this might just not be possible. The costs of such an algorithm might outweigh the benefits of the lack of resources given.
There might be a sort of time limit for sufficient levels of Machiavellian empathy-deactivation to occur. Once enough individuals in the population have sufficient care for weaker individuals and a tendency to punish third-party actions (see claim 10), then being unempathetic to the weak could be punished too much to be advantageous, even without complexity costs.
This part, like claim 8, has a lot of “might” and “could”. This is one of the areas which I’m still slightly confused about. I have a model which can explain why the costs of g-altruism are outweighed, but I don’t know if this is the only route evolution could have taken. If I saw another species evolve e-altruism but then take the Machiavellian route (of evolving highly complex empathy-deactivating algorithms to prevent any g-altruism), I wouldn’t at all be surprised. I would be surprised if it turned out 95% of species with empathy-like emotions ended up evolving highly complex off-switches for it to prevent the majority of g-altruism.
Claim 10: “Morality” from e-altruism
De Waals refers to one-on-one morality as a social code which determines how individuals treat one another directly, and community-level morality as how individuals think about the rules as applied to others. He also suggests that one-on-one morality is the result of a mixture of empathy and physical enforcement by third parties. I think this makes no sense by his own definition so I’ll ignore his distinction, although I may have misunderstood his position.
Group hierarchies amongst chimps are maintained by physical force, even when that force isn’t currently in use. When politics are involved, one must care about harm done to one’s allies, so successful individuals will care about conflicts in which they are a third party. I think this is sufficient pre-adaptation for further cognition relating to conflicts involving only others.
De Waals suggests that group cohesion also makes it advantageous for individuals to enforce rules or mediate fights. It is apparently quite common for older females or former alpha males to spend time mediating between fights in the group. Tit-for-tat fairness concepts plus empathy provide a pretty reasonable mechanism for this all to evolve.
Experiments and Epistemology
Animal behavioural research is difficult from an epistemic perspective. Humans are deeply biased towards empathizing and projecting. For an example consider the chimp “sign language” experiments.
When a researcher has spent their whole life studying the animal in question it’s unsurprising that they’ll be fond of their subjects (it would be a bit sad not to be). This makes it reasonable to suspect bias!
On the other hand, you can go too far. De Waal recalls a student of his defending their thesis on reconciliation behaviour (chimps making up after fights). Despite being rat psychologists, the reviewers claimed with certainty that no such behaviour could have taken place. When he offered to take them to the zoo to observe the animals (this was the ’70s, so no video recordings) they replied: “What good would it do to see the actual animals? It will be easier for us to stay objective without this influence.”
I assume most criticisms of his research do not fall into self-parody in this way, but maybe he really does live in a world of cartoon behaviourists.
R-altruism at all?
De Waal spends a while showing his own evidence for apes being r-altruistic at all. Big chunks of the book are direct accounts of ape behaviour clearly meant to give the impression of them having some humanity. But he also includes a number of studies, including plenty that he’s done himself.
Chimps generally show r-altruism in “giving assistance tests” where they’re willing to help out a second individual (human or chimp) by giving them a tool which is out of reach to the second subject.
On what are called “prosocial choice tests” their results have been mixed. These consist of giving the test subject the choice between rewarding themself, or to reward themself and another individual of the same species. De Waal has gained positive results (apes often take the prosocial option) whereas previous studies did not.[4]
De Waal claims this is due to the previous studies using complex devices which the chimps couldn’t understand. His studies used a token exchange system, in which the chimps handed a human researcher a red token to reward just themself (the antisocial choice), or a green token to also reward a chimp in an adjacent enclosure (the prosocial choice).
Interestingly, more dominant chimpanzees chose the prosocial option more often, and that begging behaviours from the recipient chimp decreased the frequency of the prosocial choice. This is evidence against claims that chimp r-altruism occurs only under threat.
His study only included female chimps from the troop under investigation. I don’t know why this was done.
I think his claims about chimps exhibiting r-altruism under these circumstances are correct. They square with numerous food-sharing observations of wild chimpanzees, and with the results of giving assistance tests. Several types of monkey have also been shown to have a prosocial bias in prosocial choice tests.
Empathy in Apes
There are a couple of examples which the author uses to sell the concept of empathy in apes. One is of a female chimp who, upon seeing an old, ill chimp leaning against a concrete wall, pushed some bedding behind him in order to cushion his back. As this was the first time he had shown any sign of illness, De Waal claims that this required her to model how she would feel in his situation.
The second is that bottle-reared bonobos (a common thing at sanctuaries due to poachers killing adults for bushmeat) will often play with bottles and water to bottle-feed their own offspring, and other juveniles, in addition to nursing. Crucially to the claim of empathy, they take care not to put too much water into the young bonobo’s mouth. This could be because they can imagine that too much water in the mouth is unpleasant and could cause the juvenile to choke.
Both of these seem like relatively weak evidence, but based on De Waal’s argument about parsimony, I think it’s likely that apes have something neurologically similar to empathy.
Spot-check
The book was written in 2013, which is late enough that the author should have considered the replication crisis in psychology. Here are some random picks of articles cited in the book:
Aknin et al. “Giving leads to happiness in young children”
A study suggesting that various conditions where children give treats to a puppet, or share toys, are happier than when they do not give, or do not share the toy. The most questionable part of this study is the usage of two people rating the children’s happiness based on video recordings as a method of assessing happiness, which the authors claim is standard and reliable. No direct replication has been attempted. The general impression I get from literature is that giving seems to make people happy, although almost all potential studies might be confounded by the fact that being happy makes people give.
Brosnan et al. “Monkeys reject unequal pay”
This one is pretty famous, monkeys will do a task (handing researchers a token) when rewarded with cucumber slices, until they see another monkey being rewarded with grapes. Suggests monkeys have a concept of inequity. This has been reviewed here and replicated.
Evans et al. “Chimpanzees use self-distraction to cope with impulsivity”
Chimpanzees are presented with a box that slowly fills with treats. They can grab the box at any time but that prevents any more treats from being put in. The longer they can hold themselves back, the greater the reward. In this experiment chimps were shown to be able to hold back longer when given toys to distract themselves with. Seems to be reasonable experimentally. Their data isn’t amazing. No replication has been attempted.
Here are a couple of papers which contradict claims made in the book:
Riedl et al. “No third-party punishment in chimpanzees”
This suggests that chimpanzees actually do not punish third-party transgressions. I think this evidence is weaker than expected because they used theft as the “crime”, and chimpanzees do not respect ownership under certain circumstances, such as when the “owner” is away. Most accounts of chimpanzee interventions in third-party scenarios consist of physical aggression.
Jensen et al. “Chimpanzees are rational maximizers in an ultimatum game”
Provides evidence against the abstract concept of fairness in chimps, at least to the same degree as in humans. Chimps will generally accept any offer in ultimatum games, without a “spiteful” punishment of low offers.
Epistemic health: significantly above average for a psychology-heavy pop-science book.
Pointing At Morality
I now feel I have a good understanding of the sort of thing that morality is. I feel I can see where it likely came from, and why I feel the way I do about certain things.
I see my own moral values as starting from a bunch of emotional value judgements over a bunch of things. A subset of this is a few strong preferences to things like consistency and universality, and respecting the value judgements of others. With these preferences, and my own reasoning abilities, I attempt to wrangle the rest into a better shape.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3970702/
https://www.cell.com/cell-reports/fulltext/S2211-1247(20)30319-3?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS2211124720303193%3Fshowall%3Dtrue
https://pubmed.ncbi.nlm.nih.gov/15811234/
https://www.pnas.org/content/108/33/13847