Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Kaj_Sotala comments on Robust Cooperation in the Prisoner's Dilemma - Less Wrong

69 Post author: orthonormal 07 June 2013 08:30AM

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

Comments (145)

You are viewing a single comment's thread.

Comment author: Kaj_Sotala 09 June 2013 05:25:48PM 8 points [-]

Stupid question: are there clear examples of any real-life events that would actually correspond to classic, one-shot Prisoner's Dilemma? I thought that I knew of plenty of examples before, but when I was challenged to name some, all the ones I could think of failed some of the conditions, like being tragedies of the commons rather than PD, or being iterated PD rather than one-shot PD, or the players being able to adapt their plans based on the way the opponent reacts instead of it being a single all-or-nothing choice (arms races are sometimes held up as an example of PD, but there both sides have access to intelligence reports on the kind of strategy that the enemy seems to be adopting, and can adjust their strategy accordingly).

Of course, one can always construct hypothetical scenarios in a way that makes them classic one-shot PD by construction, but being unable to name any clear real-life examples would seem to suggest that this particular scenario isn't that interesting to focus on.

Comment author: orthonormal 10 June 2013 02:39:07AM *  6 points [-]

The point of using the one-shot Prisoner's Dilemma (and variants like this one) isn't actually that it's a realistic approximation of real-life phenomena, so much as that it's one of the simplest decision-theory problems that standard game theory looks suboptimal for, and so it's a good proving ground for further development. (Think of it like the assumption that objects in kinetic physics are frictionless.)

Similarly, game theory started with assumptions of perfect common knowledge of payoff matrices, which didn't model real-life situations all that well, but the theory developed there was later extended to more realistic setups in politics and business.

That being said, I'm also curious to know if there are real-world examples that model one-shot Prisoner's Dilemmas reasonably well.