Prisoners Dilemma

The Prisoner's Dilemma

Two prisoners are each independently given an offer: Snitch on the other prisoner to reduce your sentence. However, if they both snitch, they both get a longer sentence.

	Prisoner B stays silent (cooperates)	Prisoner B testifies (defects)
Prisoner A stays silent (cooperates)	Each serve 1 year	Prisoner A: 3 years Prisoner B: goes free
Prisoner A testifies (defects)	Prisoner A: goes free Prisoner B: 3 years	Each serve 2 years

According to game theory, the optimal move is to snitch. Loyalty is, in this game, irrational. Mutual cooperation would yield a greater net reward, but rational agents don’t care about net reward, they care about my reward.

What’s interesting/important about the Prisoner’s Dilemma is how inaccurate it is to real life. PD suggests altruism is illogical, but cooperative behavior is overwhelmingly the dominant strategy - it’s the norm.

So, the critical question is why the Prisoner’s Dilemma is a bad model for looking at cooperation in general.

Reasons It Might Be a Bad Model

It assumes no communication between A & B: They cannot strategize, beg, or haggle.

It assumes that both Agents have Rational Self Interest and will logically optimize for their score, when it could instead be closer to The Scorpion And The Frog.

Most critically, it assumes that the knowledge of their actions won’t leave this game.

It assumes a 1-step game: A & B will never play this game with each other again. Prisoner A won’t “get a reputation” as an aggressor.

The Prisoners Dilemma assumes Prisoner A won’t get shanked by Prisoner B’s gang once they get out. Snitches Get Stitches, don’t they?

Life seems to be closer to an Iterated Prisoners Dilemma. It models human behavior more accurately in most scenarios.