Nice essay on the Prisoner's Dilemma Tournaments

[mouser]

I remember one of the first things that got me truly interested in artificial intelligence as a young kid was talking with my father about a Scientific American article on the "Prisoner's Dilemma" with my father.
It's a wonderfully rich and accessible mathematical/game-theory problem, made all the more exciting by the idea of having these tournaments that pit different strategies against each other.

https://egtheory.wor....com/2015/03/02/ipd/



And another more recent interesting discussion: http://www.scientifi...tion-into-question1/

Found on: http://researchblogging.org/

[rxantos]

A lot of oversimplification. It assumes that what drives an organism is only its rewarding itself.

For most the nearby pack is as important as themselves. They are your family. Thus your reward is having a stronger family.

Assume the classical prisoner dilemma:
If A and B decided not to talk, then A and B gets 1 year of other charges.
If A and B decided not to talk, then A and B get 10 years.
If A decides to talk, and B does not A gets free, B gets 20
If B decides to talk, and A does not B gets free, A gets 20


But what if your organization is important to you? Or what if deciding to talk means that you will get wacked by the other members of the organization? In real life this happens.

If your son is going to go to jail and you can prevent it by taking the blame? Many people will choose to help their son. The survival of the fittest model is flawed, as it does not take into account that not all that drives a living being is their own survivability. Sometimes the survivability of those around you (aka your family) means more to you.

[Renegade]

^ I think the point of the tournament is with iterations, and not single-shot strategies.

Instead of thinking of "years in prison", the point is to think in terms of X negative consequences and "defection" (the term used for the problem).

[mouser]

No one is claiming this is capturing all of the complexities of real life -- this is a very simplified model that has provided a surprisingly fertile and fascinating playground for research into some important issues.  That's how science works