I'm not a game theorist (and I'd welcome an opinion/correction from one), but I don't think the original paper's conclusion is very interesting, or has much bearing on real life.
Their result (that jokers, defectors, and cooperators will cycle) depends on some bizarre features of the joker:
- Jokers don't damage other jokers. This is why jokers drive out defectors.
- Jokers don't benefit from public goods. This is why jokers don't arise when there are lots of cooperators.
- The public good benefits that jokers forgo are redistributed back onto the cooperators. This is why cooperators flourish in joker populations: they produce a benefit for a large population, which then gets focused back on their small population.
This last feature is the really weird one. No public good I can think of can be redistributed this way [1]; indeed, goods that _can_ be efficiently reallocated like this tend to be naturally modeled as private goods.
Consider a public good: say, clean air. In this model we would have cooperators, who go out of their way to keep the air clean; defectors, who prefer air to be cleaner, but save effort by polluting; and jokers, who are indifferent to air pollution, produce a large amount of pollution, and somehow transfer the health benefits of clean air back onto the cooperators. Thus, if we added a single cooperator to a population of jokers, the cooperator would get a massive health benefit from not polluting, because of all the jokers "not consuming" the benefit. This is clearly nonsensical.
In short, the paper's conclusions follow from its premises, but its premises have nothing to do with any real situation that I can think of [2].
[1] There are a lot of public goods that degrade as more people use them, but not in a way that matches the math in this paper. Can anyone think of an example where this paper's conclusion would hold?
[2] Of course one can say that all game theory is an abstraction, which is true, but it still proceeds from a simplified model of reality, rather than totally arbitrary assumptions. This paper in particular would be getting no attention if the "joker" strategy didn't have a compelling real-world analogue. Furthermore, the problem here is not that their model is too simple, but that it adds weird, artificial features without explanation or justification.
(EDIT: Changed "small amount of pollution" to "large amount of pollution", as jokers do more damage than defectors.)
Their result (that jokers, defectors, and cooperators will cycle) depends on some bizarre features of the joker:
- Jokers don't damage other jokers. This is why jokers drive out defectors.
- Jokers don't benefit from public goods. This is why jokers don't arise when there are lots of cooperators.
- The public good benefits that jokers forgo are redistributed back onto the cooperators. This is why cooperators flourish in joker populations: they produce a benefit for a large population, which then gets focused back on their small population.
This last feature is the really weird one. No public good I can think of can be redistributed this way [1]; indeed, goods that _can_ be efficiently reallocated like this tend to be naturally modeled as private goods.
Consider a public good: say, clean air. In this model we would have cooperators, who go out of their way to keep the air clean; defectors, who prefer air to be cleaner, but save effort by polluting; and jokers, who are indifferent to air pollution, produce a large amount of pollution, and somehow transfer the health benefits of clean air back onto the cooperators. Thus, if we added a single cooperator to a population of jokers, the cooperator would get a massive health benefit from not polluting, because of all the jokers "not consuming" the benefit. This is clearly nonsensical.
In short, the paper's conclusions follow from its premises, but its premises have nothing to do with any real situation that I can think of [2].
[1] There are a lot of public goods that degrade as more people use them, but not in a way that matches the math in this paper. Can anyone think of an example where this paper's conclusion would hold?
[2] Of course one can say that all game theory is an abstraction, which is true, but it still proceeds from a simplified model of reality, rather than totally arbitrary assumptions. This paper in particular would be getting no attention if the "joker" strategy didn't have a compelling real-world analogue. Furthermore, the problem here is not that their model is too simple, but that it adds weird, artificial features without explanation or justification.
(EDIT: Changed "small amount of pollution" to "large amount of pollution", as jokers do more damage than defectors.)