Wei_Dai comments on Eight Short Studies On Excuses - Less Wrong

210 Post author: Yvain 20 April 2010 11:01PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (224)

Sort By: Leading

You are viewing a single comment's thread. Show more comments above.

Comment author: Wei_Dai 22 April 2010 03:16:47AM 9 points [-]

Suppose you are committed to tit-for-tat, and so is your partner, then if his finger truly slipped, he will defect after you defect, because his commitment doesn't include an exception for being punished for his own finger slipping (just like your commitment doesn't include an exception for his finger slipping).

The outcome to the story given by Yvain does seem wrong to me, and two actual human players would probably produce the outcome you (and Dagon) give: you defect once, and your partner keeps cooperating, to "make up" for the slipped finger. But this seems to have more to do with fairness than with commitment.

Comment author: Jonii 23 April 2010 05:00:32AM *  11 points [-]

But this seems to have more to do with fairness than with commitment.

Fairness? More like easy, common sense. You play tit for tat, opponent defects, you defect. If opponent keeps on defecting, you defect too. Simple winning strategy for you, to which you should absolutely precommit to.

Your opponent, knowing your precommitment, can't really do much. Strategically speaking, he's the one to act, as your precommitment to tit-for-tat works kinda like removing your steering wheel in a game of chicken, making you indifferent and unable to stop. He can crash you just as many times as he wants to, but the choice is always left to him.

And why it is his? Because you're precommitted to a strategy, but he, by defecting, he either strayed from his precommitted one, or revealed that he didn't have any to begin with. Now that he doesn't play pure tit-for-tat, his second best choice would be co-operating twice and continuing with tit-for-tat. To him, strategy "play tit-for-tat, defect once, continue with tit-for-tat" is quite far from optimal.