Prisoner's Dilemma

Promises as a solution?

Suppose we had anticipated being caught and promised one another not to confess.

Isn't that much more realistic? That's what real criminals do, after all. And wouldn't that solve the problem? Instead of hitting the second worst outcome (three year sentences) we could guarantee ourselves the second best outcome (one year sentences).

The same strategy

I'm still going to confess, for the same reason.

Here's the way it looks from my point of view. Let's say you keep your promise and don't confess.

Promising
Prisoner's Dilemma
Me
Keep my promise (Don't Confess) Break my promise (Confess)
You Keep your
promise
(Don't Confess)
-1 0
     

On the other hand, suppose you break your promise:

Promising
Prisoner's Dilemma
Me
Keep my promise (Don't Confess) Break my promise (Confess)
You      
Break your
promise
(Confess)
-5 -3

See a familiar pattern here? I'm better off breaking my promise no matter what you do: it's my dominant strategy.

But what about the future?

What if I run into you again later? Won't you try to punish me for breaking my word? Or what if I need you to trust me in the future? Aren't these good reasons for keeping my word?

Sure, it's just that they don't apply in this case. Remember, I said that we can each avoid one another in the future (feel free to check the scenario again, if you like).

This is an important limitation on the kind of prisoner's dilemma game that we've been playing, however: it only involves a one-shot interaction.

Quite often, we have a reasonable assurance of interacting with others on more than one occasion. At the very least, we can hope to ruin their reputations with others.

As it happens, the prisoner's dilemma can be relevant to cases in which people interact many times. Before considering that, however, take a look at the range of cases that even this simple version of the prisoner's dilemma illuminates; I'll briefly discuss these multiple or "iterated" cases in my description of the significance of this game.