# Paul_Crowley2 comments on The True Prisoner’s Dilemma

• In­ter­est­ing. There’s a para­dox in­volv­ing a game in which play­ers suc­ces­sively take a sin­gle coin from a large pile of coins. At any time a player may choose in­stead to take two coins, at which point the game ends and all fur­ther coins are lost. You can prove by in­duc­tion that if both play­ers are perfectly self­ish, they will take two coins on their first move, no mat­ter how large the pile is. Peo­ple find this para­dox im­pos­si­ble to swal­low be­cause they model perfect self­ish­ness on the most self­ish per­son they can imag­ine, not on a math­e­mat­i­cally perfect self­ish­ness ma­chine. It’s nice to have an “in­tu­ition pump” that illus­trates what gen­uine self­ish­ness looks like.

• Hmm. We could also put that one in terms of a hu­man or FAI com­pet­ing against a pa­per­clip max­i­mizer, right? The two play­ers would suc­ces­sively save one hu­man life or cre­ate one pa­per­clip (re­spec­tively), up to some finite limit on the sum of both quan­tities.

If both were TDT agents (and each knows that the other is a TDT agent), then would they suc­cess­fully co­op­er­ate for the most part?

In the origi­nal ver­sion of this game, is it turn-based or are both play­ers con­sid­ered to be act­ing si­mul­ta­neously in each round? If it is si­mul­ta­neous, then it seems to me that the pa­per­clip-max­i­miz­ing TDT and the hu­man[e] TDT would just cre­ate one pa­per­clip at a time and save one life at a time un­til the “pile” is ex­hausted. Not quite sure about what would hap­pen if the game is turn-based, but if the pile is even, I’d ex­pect about the same thing to hap­pen, and if the pile is odd, they’d prob­a­bly be able to suc­cess­fully co­or­di­nate (with­out nec­es­sar­ily com­mu­ni­cat­ing), maybe by flip­ping a coin when two pile-units re­main and then act­ing in such a way to en­sure that the ex­pected dis­tri­bu­tion is equal.