Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

phys.org

Post reply on HN
Print view View original

91–100 of 117 posts

Re: 'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

#91

Post by adrianratnapala »

kbenson wrote: post #74
jzl wrote: post #55

I've seen a couple of oblique references to this in other comments, but I'll go a little deeper: if you want to truly understand the prisoner's dilemma scenario and how it relates to human, or more generally, animal behavior, there is incredible coverage of the topic in Richard Dawkins' The Selfish Gene. He pioneered the idea of running PD "competitions" in an attempt to discover the best evolutionary strategy for PD…

> - The agent must have a good chance of competing against the opponent more than once. I was actually thinking about this the other day. I was wondering how much the population size and matching algorithm in the iterated Prisoner's Dilemma matter to the strategy outcome. I suspect it's quite a lot, given that if you never play the same opponent again, strategies that defect often should come out quite well, but if y…

It's not crucial that you meet the same partner (opponent) again. It's enough for you to be in a population where there is a good chance that the partner will cooperate if you do. For some definition of "good chance" the right strategy is to try to cooperate yourself and then fall back on something else if it fails.

And this is interesting because it gives us a toy model of moralistic "Do unto others" (as opposed to rational exchange).

Re: 'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

#92

Post by adrianratnapala »

People have been doing Prisoner's Dilemma with different payoff amounts since the dawn of PD, and of course this can result in qualitatively different strategies.

But sticking to just the rules, rather than the results. Is is the snowdrift game just a payoff score tweak? Or is there some structural difference that I missed?

Re: 'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

#93

Post by ykler »

adrianratnapala wrote: post #92

People have been doing Prisoner's Dilemma with different payoff amounts since the dawn of PD, and of course this can result in qualitatively different strategies. But sticking to just the rules, rather than the results. Is is the snowdrift game just a payoff score tweak? Or is there some structural difference that I missed?

In snowdrift, in the case where your opponent is uncooperative, it is better for you to be cooperative.

Re: 'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

#94

Post by Smaug123 »

kobeya wrote: post #85
lisper wrote: post #61

Earlier quoted context omitted.

You're right. In fact, upon reflection, it is the PD which is not really a dilemma but rather a paradox. The logically correct move in PD (assuming a purely utilitarian quality metric) is always clear: if the game is non-iterated (or iterated with a known horizon) then defect. Otherwise play tit-for-tat. This is true no matter what your opponent does. In the SD the logically correct move depends on your prior on what…

That resolution only holds if you assume the other player to be equally rational in the PD.

In fact the strategy is simpler if the other player is exactly as rational as you: simply cooperate. The other player, being exactly as rational as you, will follow the same reasoning as you do to come to the same conclusion, and will do exactly the same thing as you. I would always cooperate with a copy of myself, for instance.

https://en.wikipedia.org/wiki/Superrationality

(Of course, the assumption "my opponent is exactly as rational as I am" never holds in real life.)

Re: 'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

#95

Post by tedsanders »

adrianratnapala wrote: post #91
kbenson wrote: post #74

Earlier quoted context omitted.

> - The agent must have a good chance of competing against the opponent more than once. I was actually thinking about this the other day. I was wondering how much the population size and matching algorithm in the iterated Prisoner's Dilemma matter to the strategy outcome. I suspect it's quite a lot, given that if you never play the same opponent again, strategies that defect often should come out quite well, but if y…

It's not crucial that you meet the same partner (opponent) again. It's enough for you to be in a population where there is a good chance that the partner will cooperate if you do. For some definition of "good chance" the right strategy is to try to cooperate yourself and then fall back on something else if it fails. And this is interesting because it gives us a toy model of moralistic "Do unto others" (as opposed to…

Can you explain why cooperation still works when you don't repeat partners? (Assuming I've interpreted you correctly.)

The reason cooperative strategies can work is that actions depend on previous actions. If you are not meeting the same partner, then your partner's actions will not depend on your previous actions, and you will always do better by defecting.

Re: 'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

#96

Post by adrianratnapala »

tedsanders wrote: post #95
adrianratnapala wrote: post #91

Earlier quoted context omitted.

It's not crucial that you meet the same partner (opponent) again. It's enough for you to be in a population where there is a good chance that the partner will cooperate if you do. For some definition of "good chance" the right strategy is to try to cooperate yourself and then fall back on something else if it fails. And this is interesting because it gives us a toy model of moralistic "Do unto others" (as opposed to…

Can you explain why cooperation still works when you don't repeat partners? (Assuming I've interpreted you correctly.) The reason cooperative strategies can work is that actions depend on previous actions. If you are not meeting the same partner, then your partner's actions will not depend on your previous actions, and you will always do better by defecting.

Now that I think about it, you are right that for PD, there has to be some one-on-one iteration too. I think my automatic mental model was that agents came together and played a few rounds with one partner, and then switch to some other partner.

In this case you can afford to be nice in the first round, and then cut your losses. But even that strategy is only optimal if most of the rest of the community is nice in the first round too. If it is full of defectors, then you should defect in the first round too.

Re: 'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

#97

Post by jessaustin »

kbenson wrote: post #74
jzl wrote: post #55

I've seen a couple of oblique references to this in other comments, but I'll go a little deeper: if you want to truly understand the prisoner's dilemma scenario and how it relates to human, or more generally, animal behavior, there is incredible coverage of the topic in Richard Dawkins' The Selfish Gene. He pioneered the idea of running PD "competitions" in an attempt to discover the best evolutionary strategy for PD…

> - The agent must have a good chance of competing against the opponent more than once. I was actually thinking about this the other day. I was wondering how much the population size and matching algorithm in the iterated Prisoner's Dilemma matter to the strategy outcome. I suspect it's quite a lot, given that if you never play the same opponent again, strategies that defect often should come out quite well, but if y…

Yes, if "sociopaths" are rare then most of their games will be against opponents who haven't played against them before. One suspects this also means that if the number of games goes up dramatically (say we move to the city and enter an occupation with lots of daily transactions), initial cooperation will decrease as sociopaths are seen more often.

Re: 'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

#99

Post by neilwilson »

tedsanders wrote: post #95
adrianratnapala wrote: post #91

Earlier quoted context omitted.

It's not crucial that you meet the same partner (opponent) again. It's enough for you to be in a population where there is a good chance that the partner will cooperate if you do. For some definition of "good chance" the right strategy is to try to cooperate yourself and then fall back on something else if it fails. And this is interesting because it gives us a toy model of moralistic "Do unto others" (as opposed to…

Can you explain why cooperation still works when you don't repeat partners? (Assuming I've interpreted you correctly.) The reason cooperative strategies can work is that actions depend on previous actions. If you are not meeting the same partner, then your partner's actions will not depend on your previous actions, and you will always do better by defecting.

Your partners talk to each other.

The problem with all economic simulations is that they assume extreme individualisation. Because that is what they want the world to be like.

In reality the world is a network. Reputation matters.

Re: 'Snowdrift' game tops 'Prisoner's Dilemma' in explaining cooperation (2007)

#100

Post by jzl »

iainmerrick wrote: post #71
jzl wrote: post #55

I've seen a couple of oblique references to this in other comments, but I'll go a little deeper: if you want to truly understand the prisoner's dilemma scenario and how it relates to human, or more generally, animal behavior, there is incredible coverage of the topic in Richard Dawkins' The Selfish Gene. He pioneered the idea of running PD "competitions" in an attempt to discover the best evolutionary strategy for PD…

Richard Dawkins didn't pioneer that idea, it was Robert Axelrod. The seminal book about the topic is Axelrod's "The Evolution of Cooperation": https://en.wikipedia.org/wiki/The_Evolution_of_Cooperation (Dawkins was a prominent advocate of Axelrod's work, though. And "The Selfish Gene" is definitely a great read in its own right.)

Thank you for the correction. I read the book about 10 years ago and completely forgot that he wasn't talking about himself. Finally had a chance to look at the chapter again now and indeed he is discussing Axelrod's competitions in great depth.
Post reply on HN