RE: A Simple Explanation of The Prisoner's Dilemma

[nxtblg]

Funny: the example you give will only be accepted at face value by folks with less than a full ration of common sense. A more normal kind of fellow would add, "and the defector gets punished by the co-operator on the schoolyard."

That's the trouble with the standard prisoner's-dilemma game: it's sequestered from the processes of everyday life.

Not only in that way. Another criticism - taken up by the game-theory community when computers were added! - is that it's unrealistic because it's one-shot. Add computers to run what's called an "Iterated Prisoner's Dilemma", where players repeatedly match up against others for many rounds of the game, and you get this more common-sensical finding: the ideal strategy over the long term is Tit-For-Tat. Your first move should be co-operate, and your next move should be the same as the move that you yourself faced in the last round. If you got screwed over in Round 1, you should screw over in Round 2; if you got co-operation in Round 1, you should co-operate in Round 2; and so on...

I can't prove it, but I'm sure that the band Jefferson Airplane poked a hole in the standard PD model in a song that ribbed John Nash:

From the lyrics the chorus:

Tell me why, if you think you know why,
People love when there's no tomorrow
And still not cry when it's time to go,
And still not cry when it's time to go,
And still not cry when it's time to go?

[Italicized part is the poke-through of the standard PD model]

Comments

[smooth]

that's the trouble with the standard prisoner's-dilemma game: it's sequestered from the processes of everyday life.

It is, but it there is generally a bit of a disconnect (on both sides) between idealized mathematical models of useful things and the useful things themselves. These models are useful exercises and can teach us things but can't (or at least shouldn't) be mapped directly to practice.

the ideal strategy over the long term is Tit-For-Tat

It isn't quite that simple, there are other strategies that do better in different models such as forgiving tit-for-tat but again these are still models.

[nxtblg]

Thanks for your comments.

[thecryptofiend]

Funny: the example you give will only be accepted at face value by folks with less than a full ration of common sense. A more normal kind of fellow would add, "and the defector gets punished by the co-operator on the schoolyard."

I never thought about that! I think that is the issue with studying human behaviour you can only model it so far in my opinion. I'm sure the experts will disagree but I can claim ignorance on that for now:)

Great song! Never heard it before.

[nxtblg]

Thanks! Funny: I didn't clue in until I read Sylvia Nasar's bio of John Nash, the original A Beautiful Mind She included two anecdotes from his grad-school days that sealed my hunch shut:

• When Nash was ticked off at Stephen Smale, he solved ice down the back of Smale's shirt.
• Nash was an excellent Go player, but he lost one - and he walked away from the game crying.

[thecryptofiend]

I really must look into Go - have heard so much about it lately but I know nothing about it! I think an AI defeated the no.1 player not too long ago.

[nxtblg]

You're right.

"In a Huge Breakthrough, Google’s AI Beats a Top Player at the Game of Go"

IN A MAJOR breakthrough for artificial intelligence, a computing system developed by Google researchers in Great Britain has beaten a top human player at the game of Go, the ancient Eastern contest of strategy and intuition that has bedeviled AI experts for decades.

[thecryptofiend]

Yes thought so:)