DanRollins
asked on
Prisoner's Dilemma
I have recently come across several references to this (Prisoner's Dilemma) and I'm finding it fascinating. Even my daughter mentioned it (by a differnet name) in regards to a college Sociology class -- she won a $10 prize by "doing the obvious" and defecting -- she couldn't understand why the other members of the class didn't come to the same conclusion!
If you don't know what I'm talking about, please see
Prisoner's dilemma (Wikipedia)
http://en.wikipedia.org/wiki/Prisoner%27s_Dilemma
The "iterated prisoner's dilemma" is the interesting variation.
It is discussed extensively in "The Selfish Gene" by Richard Dawkins. It shows how an equilibrium can be (often must be) reached -- "unselfish" cooperation ends up being the stable thing to do, even when every participant is only out for his own greed.
The "PD" also came up in an SF series I read by Kim Stanley Robinson ("Forty Signs of Rain," et. al). It applies PD to social contexts and uses the term "to defect" -- meaning to not cooperate -- in a way that is being seen more and more often in common vernacular.
It was also mentioned in an episode o the TV show "Numb3rs."
I thought it might be interesting to discuss this, partly because of the novely to being able to cross post in these three areas...
Philosophy and Religion -- cooperation is a key philosophical concept; evolution
Puzzle and Riddles -- PD is game theory
Math & Science -- Related scientific concepts such as "Nash Equilibrium" cellular automata, etc.
Some discussion points:
* Have you ever run across the idea of the PD in common media -- your
reading, TV shows, whatever?
* The idea's been around for a while: Did you study it in college?
* Name some practical uses for the "esoteric" game-theory concept
If you don't know what I'm talking about, please see
Prisoner's dilemma (Wikipedia)
http://en.wikipedia.org/wiki/Prisoner%27s_Dilemma
The "iterated prisoner's dilemma" is the interesting variation.
It is discussed extensively in "The Selfish Gene" by Richard Dawkins. It shows how an equilibrium can be (often must be) reached -- "unselfish" cooperation ends up being the stable thing to do, even when every participant is only out for his own greed.
The "PD" also came up in an SF series I read by Kim Stanley Robinson ("Forty Signs of Rain," et. al). It applies PD to social contexts and uses the term "to defect" -- meaning to not cooperate -- in a way that is being seen more and more often in common vernacular.
It was also mentioned in an episode o the TV show "Numb3rs."
I thought it might be interesting to discuss this, partly because of the novely to being able to cross post in these three areas...
Philosophy and Religion -- cooperation is a key philosophical concept; evolution
Puzzle and Riddles -- PD is game theory
Math & Science -- Related scientific concepts such as "Nash Equilibrium" cellular automata, etc.
Some discussion points:
* Have you ever run across the idea of the PD in common media -- your
reading, TV shows, whatever?
* The idea's been around for a while: Did you study it in college?
* Name some practical uses for the "esoteric" game-theory concept
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
Excellent example, mrider01!
And each guy is also thinking... If I can convince the other guys to go for the other girls, then I can take a shot at the blond!
But apply the "iterated PD" to the situation: If the technique of ignoring the blond worked once or several times in the past, then each participant would be more inclined to cooperate. Occasionally, someone would "defect" (go for the blond) and thus harm the group. That guy would not be invited to the next outing.
And more subtle...
If this became a standard at all bars, then an equilibrium might be reached, where the blond got so few dates that she became "just one of the pack" (would be selected randomly rather than first) and she'd end up with her "fair share" of dates.
And even more subtle...
Somebody would realize that the best-looking girl was only being selected randomly rather than first, and would see a defection-for-profit opportunity...
And each guy is also thinking... If I can convince the other guys to go for the other girls, then I can take a shot at the blond!
But apply the "iterated PD" to the situation: If the technique of ignoring the blond worked once or several times in the past, then each participant would be more inclined to cooperate. Occasionally, someone would "defect" (go for the blond) and thus harm the group. That guy would not be invited to the next outing.
And more subtle...
If this became a standard at all bars, then an equilibrium might be reached, where the blond got so few dates that she became "just one of the pack" (would be selected randomly rather than first) and she'd end up with her "fair share" of dates.
And even more subtle...
Somebody would realize that the best-looking girl was only being selected randomly rather than first, and would see a defection-for-profit opportunity...
ASKER
... He might *pretend* to select randomly (i.e., let others select first) then jump in on the Blond. He'd need to be careful to avoid detection.
Dawkins' book describes a number of scenarios where the balance would (or did) shift back-and-forth. And a predictable result is an increase in the ability to be deceptive.
In another excellent book
"The Red Queen: Sex and the Evolution of Human Nature" by Matt Ridley
I thought the most interesting musings (always de-emphasized for lack of proof) were around the idea that the ungainly and evolutionarily-risky size of the human brain is due in part (or even *mainly*) to the extreme advantage of being able to deceive one's mate into helping you care for your offspring while you simultaneously mate with others who are likely to produce more (or better) offspring.
And what follows is the (perhaps more important) ability to *detect* such deception. When there is an "arms race" like that, evolutionary selection pressure can happen relatively quickly.
Dawkins' book describes a number of scenarios where the balance would (or did) shift back-and-forth. And a predictable result is an increase in the ability to be deceptive.
In another excellent book
"The Red Queen: Sex and the Evolution of Human Nature" by Matt Ridley
I thought the most interesting musings (always de-emphasized for lack of proof) were around the idea that the ungainly and evolutionarily-risky size of the human brain is due in part (or even *mainly*) to the extreme advantage of being able to deceive one's mate into helping you care for your offspring while you simultaneously mate with others who are likely to produce more (or better) offspring.
And what follows is the (perhaps more important) ability to *detect* such deception. When there is an "arms race" like that, evolutionary selection pressure can happen relatively quickly.
Well, it seems that now we're going to start redefining the 'game' or 'strategy'. If all the guys are discussing their strategy, and they all agree to go for the girlfriends (which is what I think we ultimately are talking about) then I think a better way to do things would be to cycle through who gets the blonde, when we start dealing with Iterated PD. Or does a 'strategy' by definition not change? Could the everyone in the group each have the strategy of "Every fifth group of girls that enters will have the blonde that I can hit on," given that there are five guys. But in the end, everything comes down to the rule of PD, being that everyone follows their strategy and does not stray away from it.
Then, if people stop following the agreed 'strategy,' is when you enter the 'arms race' scenario, which is truely a lose-lose situation. So you have to take a look at which you value more: having yourself succeed, or having others lose... which then rolls back to the 'selfish gene'...
Then, if people stop following the agreed 'strategy,' is when you enter the 'arms race' scenario, which is truely a lose-lose situation. So you have to take a look at which you value more: having yourself succeed, or having others lose... which then rolls back to the 'selfish gene'...
ASKER
A successful variation of the "Tit-for-Tat" strategy is one that randomly (but rarely) defects when the opponent cooperates and randomly (but rarely) cooperates in response to a defection.
The idea is to break up a lose/lose chain against an opponent who uses a similar basic strategy.
BTW, the other reason I asked this question is because I so rarely get to type "Tit" in these discussions :-)
The idea is to break up a lose/lose chain against an opponent who uses a similar basic strategy.
BTW, the other reason I asked this question is because I so rarely get to type "Tit" in these discussions :-)
Ditto RobinD. It does not work in practice. Games are never really balanced fairly anyway.
- bye -
Forget Prisoners for there time, let them do there time then think about them.
So I'm getting some points for being the only person to add anything meaningful to this post? ;-)
ASKER
Sure! Everyone who contributes in an interesting way to discussions I've started will get a share of the points. Now if only I could think of a way to use EE Points as the reward in a PD-based game... :-)
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
In that case, I am dissing all the other somethings.
And call for the game to be over immediately.
:-P
And call for the game to be over immediately.
:-P
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
Odd/even probably can't work unless we could prevent people from posting several times in a row.
Perhaps this:
Cooperate= "Say something nice about the previous poster"
Defect= "Say something rude about the previous poster"
In a way, we see a standard escalation whenever we see a forum flamewar. Everything is on-topic and on-track, when somebody takes offense... And now you've suddenly got two posters shouting obscenities at one another... (tat! Tat! TAT! TAT! TAT!!!)
Perhaps this:
Cooperate= "Say something nice about the previous poster"
Defect= "Say something rude about the previous poster"
In a way, we see a standard escalation whenever we see a forum flamewar. Everything is on-topic and on-track, when somebody takes offense... And now you've suddenly got two posters shouting obscenities at one another... (tat! Tat! TAT! TAT! TAT!!!)
I was imagining a near equilibrium where rather than try and hog all the positions, all would slow down to let the prize be shared more equally. It would only take two people posting hard and furiously to put each other off.
Robin,
Consider this to be me putting you off. Or as Dan noted it, something "rude".
Unfortunately, I'm not sure how much we'll be able to disagree on the rules of a relatively standard game. Other than joking about the stupidity of other people's rules, I don't think there will be anything but pseudo-heated debate. But I'll do whatever I need to do to get points. Tell me how to defect, then sit back and watch.
Consider this to be me putting you off. Or as Dan noted it, something "rude".
Unfortunately, I'm not sure how much we'll be able to disagree on the rules of a relatively standard game. Other than joking about the stupidity of other people's rules, I don't think there will be anything but pseudo-heated debate. But I'll do whatever I need to do to get points. Tell me how to defect, then sit back and watch.
'course the other problem is that unlike a tv game where the money goes up the further they get, in here the points get spread thinner and thinner until there isn't so much of an incentive to try to win them.
>something "rude". - I'm ignoring your comment will that do?
>something "rude". - I'm ignoring your comment will that do?
I'm not sure how "PD" this is, you'll have to let me know, but how about if all the points go to the fourth person to post after this message. If less than four people post, then the points get split among the three or less.
But acually, it would be ridiculus for three people to split it, because then a fourth person would steal the points. So in reality, the only split should be a two-way.
Clearly, the points should go to the first person who came up with a way to use PD to distribute points.
Nah, I think Ozo should get them. I mean, come on, he posted some links. Nothing says "I really care!" like posting a wikipedia article. And I see you have your game on, gilbar by saying defecting and saying something rude about other posts.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
In most PD variations, cooperating (keeping silent) gets a walk for both you and your partner-in-crime. The whole point is that there is a true dilemma:
Cooperate: If everybody else does, then everybody wins
Defect: Regardless of what anyone else does, you lose, but to a lesser degree.
A dilemma comes when there are two options, usually where both of them have drawbacks. In this case, the drawbacks are different -- hard to weigh against one another -- because there is no way to know what the other guy will do.
If you could be certain that he'd cooperate, then it's a no-brainer: You should cooperate.
But how sure are you?
For instance, if you are 90% certain, then the extra reward needs to cover that 10% risk.
When money is used as the incentive, then the calculations are easier than when a jail term is involved. Even then, you might be weighing apples against oranges -- a $1000 reward is not the same to Bill Gates as to some homeless junky in need of a fix.
Cooperate: If everybody else does, then everybody wins
Defect: Regardless of what anyone else does, you lose, but to a lesser degree.
A dilemma comes when there are two options, usually where both of them have drawbacks. In this case, the drawbacks are different -- hard to weigh against one another -- because there is no way to know what the other guy will do.
If you could be certain that he'd cooperate, then it's a no-brainer: You should cooperate.
But how sure are you?
For instance, if you are 90% certain, then the extra reward needs to cover that 10% risk.
When money is used as the incentive, then the calculations are easier than when a jail term is involved. Even then, you might be weighing apples against oranges -- a $1000 reward is not the same to Bill Gates as to some homeless junky in need of a fix.
I'd thought that if you defected and he didn't then you'd walk, and if you both defected you'd get the same punishment. I guess I'm wrong.
But lets get to the important issue of hijacking this thread and making it be about mexican soap operas
But lets get to the important issue of hijacking this thread and making it be about mexican soap operas
ASKER
In many games of chance, the players must do constant risk/benefit calculations. For instance, in backgammon, you might move a checker to a place where it is in danger of being sent back to the start... That's a costly risk, so you weigh how likely it is to get "hit" against the benefits of having that checker at that location. The board is constantly changing; early in the game, being hit may not be much of a loss. Later, for instance, if most of the inner board is blocked, the risk is much more substantial.
Sometimes, you can do a very simple calculation; for instance: Three rolls (of 36) hit the checker -- call it a 12% chance of being hit. If hit, you lose, if missed, you win. When playing for $100, then that move "costs" you $12 but gains you $100. Some other move might cost you $50, so you select the one that is less costly. Or, some other move might "cost" only $3, but leaves you with a specific chance of leaving a much riskier situation later.
In the PD, you can also "do the math" but there is a huge unknown variable -- what will the other guy do? The beauty of the iterated PD is that it provides a means to adjust your calculations based on a history of previous encounters.
Sometimes, you can do a very simple calculation; for instance: Three rolls (of 36) hit the checker -- call it a 12% chance of being hit. If hit, you lose, if missed, you win. When playing for $100, then that move "costs" you $12 but gains you $100. Some other move might cost you $50, so you select the one that is less costly. Or, some other move might "cost" only $3, but leaves you with a specific chance of leaving a much riskier situation later.
In the PD, you can also "do the math" but there is a huge unknown variable -- what will the other guy do? The beauty of the iterated PD is that it provides a means to adjust your calculations based on a history of previous encounters.
Actually the way the Wikipedia article noted in the initial post describes the PD is similar to the one you are describing. It's a little more detailed than either walking or not walking. It is stated that if both people defect, then they each serve 5 years. If one person defects, they go free while the other gets 10 years. If they both cooperate, then they both serve 6 months. But you are still right when you say that regardless of what the other person chooses, you will be better off if you defect. If you were friends with the other prisoner, that would be a dilemma because you wouldn't want to put your pal away. But if you were both strangers, then I'd agree and say it's no dilemma at all. However, your TeleMundo situation is more of a damned-if-you-do/damned-if -you-don't .
ASKER
Just when I thought this thread was over... I saw the movie
The Dark Knight (http://www.imdb.com/title/ tt0468569/ )
..in which an awsome variant of Prisoner's Dilemma was portrayed. Without giving away a major spoiler, I can say that it had something even I had not thought of: Each minute that the clock ticks is, in a way, a separate game of an itterative PD.
My daughter is home from college and when we discussed the movie, she reminded me about the PD experiment that I mentioned at the top. She actually won $100 (not $10) so it was a serious challenge with a significant payoff.
The Dark Knight (http://www.imdb.com/title/
..in which an awsome variant of Prisoner's Dilemma was portrayed. Without giving away a major spoiler, I can say that it had something even I had not thought of: Each minute that the clock ticks is, in a way, a separate game of an itterative PD.
My daughter is home from college and when we discussed the movie, she reminded me about the PD experiment that I mentioned at the top. She actually won $100 (not $10) so it was a serious challenge with a significant payoff.
ASKER
Although we did get some good discussion here, I'll admit that I'm a bit disappointed in the lack of response to this question. To me, it's a fascinating topic, but I guess it's less so to other EE Members.
Anyway, thanks to all who participated.
Anyway, thanks to all who participated.
Thanks Dan
http://www.santafe.edu/research/publications/wpabstract/199901002
http://www.santafe.edu/events/abstract/519