This month the University of California Press published the Encyclopedia of Theoretical Ecology. I have an article in it, written with Richard McElreath and Pete Richerson, on the “Evolution of Cooperation.” The article’s intended audience are undergraduate biology majors and graduate students with backgrounds in evolutionary biology or ecology, but it might also be relevant to social scientists interested in evolutionary game theory. You can read the article, in its entirety, here. Below is a brief summary.
In the article, we try to give an overview of work in the evolution of cooperation and also alert the reader to some common misunderstandings, listed below. If you keep some of these ideas in your back pocket, you can pull them out at parties conferences to impress your friends colleagues! (YMMV):
- Not all observed cooperative behavior is altruistic. Many cooperative behaviors are mutualisms or coordination or threshold cooperation. (The article explains what we mean by these terms with a simple model of cooperative hunting in wolves.) Theorists and empiricists tend to focus on models of altruism, but it is important to determine what model of cooperation is most appropriate to the question being asked. However, this is often hard to do. (See also Tim Clutton-Brock and Brian Skyrms.)
- Kin-selection and group-selection are not different mechanisms of selection, they are different ways of accounting for fitness effects in models of selection. This argument will not be new to regular readers of this blog, so I will not repeat it here. However, I should point out that our encyclopedia article was finished in summer 2010, before the huge dust-up over Nowak, Tarnita, and Wilson’s infamous anti-kin-selection article and before the recent anti-group-selection writings of Dawkins and Pinker. It turns out that our claim that “old debates concerning the ‘correct’ style of analysis have largely faded in biology” was, in retrospect, sadly premature. However, in editing the page proofs, we were able to sneak in a reference to this excellent review of the debate by Lion, Jansen and Day.
- All models where altruism evolves are based on mechanisms of positive assortment – that is mechanisms where altruists are sufficiently more likely to interact with each other than with non-altruists. Plausible mechanisms of assortment are limited dispersal (biologists sometimes call this population structure or viscosity and social scientists sometimes call this clustering), signaling (green beards and kin recognition), and book-keeping (direct and indirect reciprocity).
- Tit-for-Tat is not an evolutionarily stable strategy in the iterated prisoner’s dilemma. In fact, there are no evolutionarily stable strategies in the interated prisoner’s dilemma. In fact, there are no evolutionarily stable strategies in any sufficiently iterated game. This is an old result (see Boyd and Loberbaum), but many social scientists seem unaware of it. My guess is that this is because Axelrod’s Evolution of Cooperation is still the most common introduction to cooperation in the social sciences and, while Axelrod’s book should be required reading for anyone interested in cooperation, it is particularly confusing on this point. I will elaborate on this further in a future post
- In humans, cultural inheritance is more likely to generate altruistic behavior in large groups than genetic inheritance. This is because it is much easier to maintain cultural variation between groups than genetic variation in the face of migration (i.e., the children of immigrants tend to adopt the cultural norms of their new group, but they still have copies of their parents’ genes).
It occurs to me that the points in the above list will not be clear to many readers. This is just a taste! If you are interested in more elaboration and examples, give the actual article a try!
Thanks for posting the “evolution of cooperation” article; it was very clear and useful. I’m curious: The article was written too early to reference the recent Press/Dyson paper on the Iterated Prisoner’s Dilemma. What would you have changed, if anything, in the article’s discussion of IPD if you had known about it?
The short answer: I would not have changed anything.
The longer answer: The Press/Dyson paper is pretty technical and outside the scope of the encyclopedia article and does not contradict anything we wrote.
They consider an IPD between two players, but the stability results we discuss involve the evolution of strategies in a large population. This is an important distinction since strategies that do better in a two-player interaction do not necessarily do well when in a population.
For a simpler example, a strategy that always defects (ALLD) will beat TFT every time in a two-player interaction. But TFT (under many conditions) will have a higher average score than ALLD if there are enough other TFT players around in the population.
Press/Dyson, perhaps confusingly, use the term “evolutionary strategy” to describe one or their player-types, but their model is outside evolutionary game theory. For example, I doubt their first proof (that long-memory strategies reduce to short-memory strategies) would hold in a population-level context.
I read the paper. The most obvious problems were errors of omission:
The paper makes no mention of manipulation. It seems like a large omission. Cooperation through manipulation by other humans and by cultural symbionts is commonplace.
The paper makes no mention of evolutionary legacy hypotheses either. Our ancestors evolved in a band of close friends and relatives which we had extended reciprocal relationships with. Humans are adapted to that environment – not the modern world where we are surrounded largely by unrelated strangers.
The discussion of signalling in the paper is all wrong. Humans engage in “virtue signalling” engaging in public altruistic acts to improve their repuatations and to prove how fine and trustworthy they are to onlookers. It’s a pretty big deal in terms of motivating gifts and pro-social acts, and it has nothing to do with green beards or kin recognition.
On a more positive note, both kin selection and reciprocity are treated more favourably here than by some of these authors in the past. It seems as though some progress is being made.
Obviously, many things are left out. The article had a fixed length and was intended as an introduction of the key points for an audience interested in modeling in theoretical ecology. We had to pick and choose what are considered important to that audience.
Given the audience, the article focuses primarily on non-humans except for the very short discussion of culture at the end. An article with more of a focus on humans might have more about reciprocity models with costly signalling and discussion of evolutionary legacy/great mistake hypotheses.
But I am glad you liked our discussion of kin selection and reciprocity.
Thanks for posting the “evolution of cooperation” article; it was very clear and useful. I’m curious: The article was written too early to reference the recent Press/Dyson paper on the Iterated Prisoner’s Dilemma. What would you have changed, if anything, in the article’s discussion of IPD if you had known about it?