The Rise of Cooperation
How Cooperation Can Evolve in a Cheater’s World
Whether you’re a free-loading virus or a meat-stealing monkey, selfishness pays. So how could cooperators survive in a cheater’s world? Thomas Flatt, a postdoctoral research associate at Brown, was part of a group that created a theoretical model that neatly solves this dilemma, which has stumped evolutionary biologists and social scientists for decades. The trick: Keep the altruists in small groups, away from the swindling horde, where they multiply and migrate.
It’s a truth borne out in biology and economics: Selfishness pays. Viruses can steal enzymes to reproduce. Tax evaders can take advantage of public services to survive and thrive. And, according to game theory, the cheats win out over the altruists every time.
Yet cooperation is a hallmark of human society, allowing for the creation of everything from the local grange to the United Nations. Cooperation can also be found in the animal world. Lions hunt in packs. Ants and bees create colonies. So how could cooperation evolve in a cheater’s world?
It’s a paradox called the “tragedy of the commons,” a conflict between individual interests and the common good that has stumped scientists for generations. Now a trio of researchers, including a Brown University biologist, has created a unique theoretical model that can explain the rise of cooperation. Under their model, altruists not only survive, they thrive and maintain their numbers over time. The work appears in the Proceedings of the Royal Society B: Biological Sciences.
“What’s exciting about our approach is that is so simple and so general – in principle it can be applied to explain cooperation at all levels of biological complexity, from bugs to humans,” said Thomas Flatt, a postdoctoral research associate in Brown’s Department of Ecology and Evolutionary Biology. “It’s also exciting because cooperation is a critical notion in so many disciplines, from biology to sociology. Yet its existence and persistence doesn’t always make sense. Now we have a new mechanism that explains when cooperation can work.”
Timothy Killingback, a mathematician at the College of William & Mary, led work on the model. It’s based on public goods games, a staple of game theory and a simple model of social dilemmas.
Under the typical public goods game, an experimenter gives four players a pot of money. Each player can invest all or some of the money into a common pool. The experimenter then collects money thrown into the pool, doubles it and divides it amongst the players. The outcome: If every player invests all the money, every player wins big. If every player cheats by investing a just few dollars, every player reaps a small dividend. But if a cooperator squares off against a cheater – with the altruist investing more than the swindler – the swindler always gets the bigger payoff. Cheating, in short, is a winning survival strategy.
Under the new model, the team introduced population dynamics into the public goods game.
Players were broken into groups and played with other members of their group. Each player then reproduced in proportion to the payoff they received from playing the game, passing their cooperator or cheater strategy on to their offspring. After reproduction, random mutations occurred, changing how much an individual invests. Finally, players randomly dispersed to other groups, bringing their investment strategies with them. The result was an ever-changing cast of characters creating groups of various sizes.
After running the model through 100,000 generations, the results were striking. Cooperators not only survived, they thrived and maintained their numbers over time. The key is group size.
“In our model, you can get groups of different sizes – and cooperators seem to flourish in smaller groups,” Flatt said. “In these smaller groups, the high investments of cooperators begin to pay off. Cooperators have a higher level of fitness, so they reproduce at higher rates. This allows them to get a toehold within a group, then dominate it, then send their offspring to spread their altruism elsewhere.”
The model created by Killingback, Flatt, and Jonas Bieri, a Swiss population biologist and computer programmer, is unlike any other. It relies solely on population dynamics to explain the evolution of cooperation. Most other models assume more complicated mechanisms such as kin selection, punishment and reciprocity. Some of those mechanisms require cognition, so those models can only be applied to humans and higher-order animals.
The paper is available online or may be downloaded in pdf from the June 22, 2006, issue of Proceedings of the Royal Society B: BiologicalSciences.
The Swiss National Science Foundation and the Roche Research Foundation funded the work.
10 Comments on "The Rise of Cooperation":
I think this was one of Axelrod's original conclusions too -- a small group of cooperators, working mostly together, can succeed in a world of defectors.
The concept is not new though. The nouns "cooperative", and perhaps even "commonwealth", show that people caught on to this strategy a long time ago.
As for groups that have tried it and succeeded, the Korean Businessmen's Association in Canada is a good recent example.
The Jews in Europe over the last 15 centuries are another. Which leads to the next question -- If you and your friends successfully cooperate in a world of cheaters, won't the less successful cheaters resent it, and make up conspiracy theories to explain your success, and then use the power of the masses to wipe you out? The words "ghetto", "final solution", and "Protocols of the Elders of Zion" suggest that it really happens.
The novelty here is that of application.
Well, this gave me an emotional lift. Cooperators can thrive as long as they stick together and keep out the freeloaders. Thanks for posting it, Carolyn! (Well, I suppose that still doesn't help when the freeloaders can pollute our shared planet, etc. Clearly the cooperators need to create a world dictatorship :-> Or colonize space.)
I see application of this work to "organizational behaviour". Cooperation may be the standard in a newly-formed company, but as the organization gets larger, cooperation becomes relatively less useful and cheating more so. This doesn't mean that everyone is better off cheating, but it does mean that one's political skills need to be good in order to defend oneself against cheaters. No wonder I don't like working at big corporations. The one time I got fired from a job it was at a company that started out as a great place to work, but as it grew internal politics arose, cheating became useful, and I happened to get skewered.
This post has been removed by a blog administrator.
Uplifting indeed. As a some-time corporate trainer, I encounter the cooperation vs. cheating dilemma on a regular basis. It is heartening to see a sound basis for the persistence of the cooperative spirit in biology. And yet...
One can't help but wonder what the downside of all of this must be. If we examine some species that are presumably high on the evolutionary scale of cooperation, we find a lot of insect colonies and the like. It is perhaps safe to say that the life of the colony (the common good) is of higher value than the life of any individual members. The fear is that highly evolved cooperative society is in some ways associated with a removal of individual identity. Specialization is still apparent in your societal role, but within that class, you are fixed and indistinguishable from all the others. I believe that such an outcome can arguably be traced back to the cooperative basis of their existence.
It also makes me think that in some ways we are all cheaters (oh horror) - it just depends on what level of rules you are looking at. After all, what is cheating? Doing things in a way that society does not think is 'right'? Thinking of yourself before the next person?When a person comes up with a novel solution to a problem, how many times does that involve a form of 'cheating' - of using the rules in ways in which they were not intended? Is it cheating when you are able to somehow shoot the grizzly who was intent on eating you for lunch?
By the way, for those of you who have heard of the game Cranium and enjoy it, I recommend you get a copy of Hoopla. It is a cooperative version of Cranium that for once doesn't feel like you are compromising fun for cooperation.
Here's a thought - and, to me, a potentially important one...
Rohan said that at one point he played the cheater and got burned. That kind of example aside, do you think that people are born cooperators or born cheaters? In other words, is it in the genetic (or early adoption) programming?
I have often thought of people I encounter as "givers" or "takers". I don't mean to imply that takers or cheaters are bad people. However, I can think of just about anyone I know, and I can classify them as one or the other, regardless of how nice or unpleasant they are. Is this an relatively constant and/or identifiable trait? I think it might be.
What do others think?
After running the model through 100,000 generations, the results were striking. Cooperators not only survived, they thrived and maintained their numbers over time. The key is group size.
“In our model, you can get groups of different sizes – and cooperators seem to flourish in smaller groups,” Flatt said. “In these smaller groups, the high investments of cooperators begin to pay off. Cooperators have a higher level of fitness, so they reproduce at higher rates. This allows them to get a toehold within a group, then dominate it, then send their offspring to spread their altruism elsewhere.”
The model created by Killingback, Flatt, and Jonas Bieri, a Swiss population biologist and computer programmer, is unlike any other. It relies solely on population dynamics to explain the evolution of cooperation. Most other models assume more complicated mechanisms such as kin selection, punishment and reciprocity. Some of those mechanisms require cognition, so those models can only be applied to humans and higher-order animals.
Alright. I have a few things I'd like to calrify, based on the above quote from the article, and some of the comments here.
First, the nature / nuture argument is popping up. Do traits need to be heritable, and if so are they recessive? These mathematicians did not speak to recessive, just that an individual passed along a successful strategy to its offspring. Why? Because they are not modeling the human genetic model here. A computer program - in this case a game strategy - can be copied as is, and moves into the next round of gaming, so whether a trait is recessive is moot. Nor do they comment on whether the parents are included in the next round or only offspring, though this would affect the next generation too. In other words, do parents get to reap their success directly (like investing in the future), or is the only result of success having offspring? What are they modeling if the latter? (Hint: how many children do you get to have. Period.) And we don't know what the weighting is on number of offpring per competivie success unit in the last round. There is a weighting. They mention that the more success parents get to have more offspring, that this is the measure of fitness in their model.
So there are lots of unmentioned variables that the original artical would clarify for us.
Moving on.
What does "the model explains" mean? This is an absurd sentence. IMO. The model models. And these guys are claiming with the use of common words like cheater and cooperator that the model models cheating and cooperating. :) Straight forward?
Let's look at that. What is cheating? In game theory, the word used is more often defecting and defector. I play a selfish strategy so I defect against the cooperators. Why did they use a different term, a more derogatory term?
What is cheating usually? Cheating usually means that you break the rules of the game, not that you follow them. I feel they are using a sleight of hand to make us feel (yes, feel) that the selfish strategy of defecting is actually cheating, that selfishness is cheating. Hmmm. Is it? Always?
And what is cooperating, which is the commonly used term in game theory? Cooperating in common speak is the sharing of resources and risks to achieve a joint goal. What is the joint goal in their model? I think it amounts in this article only to we'll both have more children.
Perhaps these guys only used two strategies (programs): cheating and cooperating. In other words, this is a society of clones, two types of clones. What about variation within those strategies? What happens if there are those who are more cooperative than others? What would it look like (in the model) if there was a continuum (or at least a broad and linear range) of amounts of resource and risk each parnet could put on the cooperative table? Let's say I can handle a 4 / 10, while you copperate at 7 / 10. And that we benefit each other consequently. Do you get a better payoff? Or is the payoff split equally? Etc.. Perhaps there is an optimal amount of cooperation, after which lesser amounts, or perhaps even defection ( -1 / 10) amounts make more sense (to the parent and the overall community total welfare).
And speaking of total community welfare, I am left wondering which combination of strategies over parents results in the maximal community success, rather than individual. Perhaps we need no more than 10% cooperators in the community overall for maximal success (success summed over each player in the iteration).
I could go on forever here it seems.
So I'll wrap up with explanation and the use of models. To the extent that a model can in fact be shown to be in one to one correspondance with a real world set of facts (a 1:1 isomorphism), then the model can be used to make predictions. Yay. Other factors in the real world are not included in the model, and aren't going to be predictable.
Which has always been my opinion of game theory... For any set of rules (a game theoretic model) you can show me, I, a sentient and fairly smart being, can decide to play outside the rules. Hence politics, diplomacy, manipulation, marketing, et al, and of course cheating.
Have fun.
The biggest evolutionary advantage that human beings have over all other species on the planet is language. I believe that language allows humans to create an intra-species memory bank. Every generation is capable of building on the previous generations work. Without this intra-species memory bank, stored in books, and other source of publication, we would have to relearn and reinvent everything in one lifetime.
Cooperation has been an advantage in economics, for a long time. For example in Adam Smiths book "Wealth of Nations", he outlines the concept of the "Division of Labour", which essentially means having factories or individuals be experts at making a specific product. This expertise leads more efficient production systems and therefore a more competitive economy.
YP - Welcome!
I'm reading John Barnes book, "The Sky so Big and Black", in which on pages 119-124 he outlines two stunning concepts:
1. Schools produce what the society actually expects from its future citizens. In other words, we get who we need, not who we'd like. And as societies vary over time, the role of schools and of education changes.
2. Societies in flux require individuals who are generalists, capable of acquiring new skills and areas of expertise regularly. Contrast with societies which are economically stable, which require individuals who are capable and happy to specialize and remain in one contributing role indefinitley. Are we in a time that requires generalists or a division of labour? Which way are we heading? Can we tell?
Hurray Carolyn!
I loved your latest review of the model in question.
I too had problems with the term "cheater", and even "defector" - both of which have varying derogatory connotations. My guess is that the research team was a bunch of cooperators! ;-) But to adopt common game theory parlance, let's switch to cooperator and defector.
I'd also have to totally agree that it can't be a black/white designation. Surely there are gradations on a C-D continuum.
Personally, I was not so interested in the nature/nurture mechanism, as wondering if the C-D score (if you will) on that continuum stays relatively constant for a given individual. That (it seems to me) might be a relatively easy thing to test for.
You are also correct in pointing out the limitations of game theory and modelling. I think what is primarily of interest to me about this work is not that they have created a predictive model, but that they point the way to finding the gaps in the previous models which say that cooperators should be quickly wiped out.
We have tended to focus on the human species in this thread, but if you try to roll things back to a simpler real world case, we should perhaps think of the bees and ants that the authors included in their original ponderings. It is intersting that you talk about clones, Carolyn. It seems to me that genetic diversity has been severely reduced in those species. And yet I have firsthand experience that while an entire bee hive may have the temperment of the queen, those temperments (and consequently behaviours) do vary from one queen/hive to the next. Which leads me to wonder: are there any bee defectors out there?! Or has that propensity been (for practical purposes - pardon the pun) wiped out in bees and ants?
As a final side note, I wonder what the implications are for AI research and development? Should the next generation of superior intelligence be based on cooperative or defector (I won't say defective) principles?
Post a Comment
<< back to .:. fuck decaf .:.