Tuesday, February 17, 2004

A short post on Retaliation

This will just be a short post on retaliation. My masters thesis deals with cooperation in game theory, what does this have to do with retaliation. Think sustainable cooperation is about always turning the ther cheek? Think again. Cooperation without retaliation in iterated prisoner's dilemma type games is a recipe for low utility, as it is so easily exploited. An eye for an eye and a tooth for a tooth, or tit-for-tat, however, also works poorly unless both agents have full and accurate knowledge about what the other agent is doing. In the presence of noise, it leads to Israel-Palestine or Northern Ireland style constant retaliation. What is needed are recoverable retaliation strategies, which retaliate for a short period of time over percieved wrongdoings, but then are capable of returning agents to mutual cooperation.

I've been having trouble finding evidence of the results I'm finding in my simulations/analysis in the real world, but then I just borrowed Peacemaking amongst Primates after following a link from slashdot on Animal Social Complexity
and it talks about exactly the stuff I've been dealing with. It talks about tolerance, in a similar way as I have. It talks about reconciliation, meaning a similar thing to my recoverable retaliation strategies. And it talks about "Good" Agression, and agression as a "Das sogenannte Bose" or so-called evil, in the same way as I concluded that retaliation was an essential part of sustaining cooperation, even in simulations with some amount of partner selection and reputation, where one might have thought these other factors might have mitigated the need for retaliation. In fact allowing retaliation would have over complicated the reputation system I'd developed, so I ditched the reputation system :-)

Anyway the main point is that I've found some real world ammunition to support some of the claims in my thesis in the field of game theory and software agents, from the unlikely field of primatology. But I think I'm quite used to finding interesting connections to all sorts of research now, working in the very fertile (IMO) area of agents and game theory.