4 ms·

It is not always wise advice. ("Now you see that evil will always triumph because good is dumb." - Dark Helmet) In an iterated prisoner's dilemma [1] where bo

It is not always wise advice. ("Now you see that evil will always triumph because good is dumb." - Dark Helmet) In an iterated prisoner's dilemma [1] where both players know the number of rounds that will be played, the optimal strategy is to always defect. When the number of rounds is unknown to both players, a good strategy was empirically shown to be "tit for tat" [2] - if your opponent is cooperating, then you cooperate too, otherwise you continue to defect as long as they do, and only resume cooperation after they resume playing nice. Obviously game theoretic models are just models, but they can be useful and illuminating. [1] https://en.wikipedia.org/wiki/Prisoner%27s_dilemma#The_iterated_prisoner.27s_dilemma https://en.wikipedia.org/wiki/Prisoner%27s_dilemma#The_itera... [2] https://en.wikipedia.org/wiki/Tit_for_tat https://en.wikipedia.org/wiki/Tit_for_tat

throwanem 10y ago
Perhaps Russell might have been wise to toss in something about not being overly literal, too.
Retric 10y ago
Iterated 1v1 prisoner's dilemma with a known number of rounds is effectively just a single round of prisoner's dilemma. As to strategy's it really depends on what the penalties are and if you can see other peoples rounds. AKA if Algo 3 can see when Algo 1 and 2 have their rounds, before having a round with 1 and 2. Another interesting version is escalating prisoners dilemma, where round X is worth more than round X -1.
shanusmagnus 10y ago
It's also useful to consider how often real life resembles the prisoner's dilemma, iterated or otherwise (in my experience, not often) and in the cases where it appears to, whether other factors (such as getting the reputation of being a dick) might swamp the advantage that is to be gained by betraying your partner. Of course you can always cook up a situation where any strategy is the best one; but in my real-life experience kindness tends to be a pretty good strategy, for a variety of realistic objective functions. (edit: grammar)

throwanem 10y ago

Perhaps Russell might have been wise to toss in something about not being overly literal, too.

Retric 10y ago

Iterated 1v1 prisoner's dilemma with a known number of rounds is effectively just a single round of prisoner's dilemma. As to strategy's it really depends on what the penalties are and if you can see other peoples rounds. AKA if Algo 3 can see when Algo 1 and 2 have their rounds, before having a round with 1 and 2. Another interesting version is escalating prisoners dilemma, where round X is worth more than round X -1.

shanusmagnus 10y ago

It's also useful to consider how often real life resembles the prisoner's dilemma, iterated or otherwise (in my experience, not often) and in the cases where it appears to, whether other factors (such as getting the reputation of being a dick) might swamp the advantage that is to be gained by betraying your partner. Of course you can always cook up a situation where any strategy is the best one; but in my real-life experience kindness tends to be a pretty good strategy, for a variety of realistic objective functions. (edit: grammar)