Page 1 of 1

The evolution of cooperation

Posted: Mon Sep 03, 2018 6:55 am
by Renée Bagslint
I'm currently reading The evolution of cooperation by Robert Axelrod, having been inspired to do so after Hannh Fry's excellent TV programme on The Joy of Winning. It has an interesting discussion about how cooperation emerges (or doesn't) in a population playing to iterated Prisoner's Dilemma games. It occurs to me that it might have a useful application to online communities such as Wikipedia -- their success and more importantly failure. Any thoughts?

Re: The evoluton of cooperation

Posted: Wed Sep 05, 2018 6:02 am
by Proabivouac
Renée Bagslint wrote:
Mon Sep 03, 2018 6:55 am
I'm currently reading The evolution of cooperation by Robert Axelrod, having been inspired to do so after Hannh Fry's excellent TV programme on The Joy of Winning. It has an interesting discussion about how cooperation emerges (or doesn't) in a population playing to iterated Prisoner's Dilemma games. It occurs to me that it might have a useful application to online communities such as Wikipedia -- their success and more importantly failure. Any thoughts?
I'm not sure what kind of cooperation you have in mind, but the simplest example is the three revert rule: you cannot revert a particular thing three times, but someone else can, and you can revert things for them, too. So a revert trading system is inevitable. You could even have a revert trading board in which every participant is granted credits for performing requested reverts which can then be spent upon posting their own requests.

Re: The evolution of cooperation

Posted: Wed Sep 05, 2018 7:49 am
by Renée Bagslint
I was thinking more in terms of social interaction, bearing in mind that rules like 3RR emerged from a community consensus around what were initially unstructured social behaviours. Interestingly one question is the extent to which such community consensus is an ex post facto rationalisation of a largely unconscious process.

Axelrod ran various studies on IPD: the well-known early study showed that among about 60 competing strategies, TIT-FOR-TAT did best: this is the strategy that cooperates at first then simply plays back the opponent previous choice. In more complex studies, its advantage depends on the value of the future: that is, the "discount" rate applied to the value of future rewards compared to a reward today. In population studies, this is related to the frequency of interacting with the same player. He suggests that small populations, where players frequently meet each other, tend to favour strategies that learn to cooperate, whereas in larger populations the effect is weaker. He also notes that large population with territoriality can evolve cooperative strategies in isolated territories which can then spread.

In terms of the Wikipedia MMORPG, then, one could postulate that early on, with fewer players who encounter each other frequently, cooperative strategies would evolve; that as the number of players grows, less cooperative strategies emerge; and that isolated communities might evolve spearate strategies, some more cooperative. All of this seems to fit, but I guess the question is whether one can model it well enough to make predictions.

Re: The evolution of cooperation

Posted: Thu Sep 13, 2018 4:39 am
by Proabivouac
Renée Bagslint wrote:
Wed Sep 05, 2018 7:49 am
In population studies, this is related to the frequency of interacting with the same player. He suggests that small populations, where players frequently meet each other, tend to favour strategies that learn to cooperate, whereas in larger populations the effect is weaker.
Yes, absolutely. This is obvious from ethnology alone.