On Fri, 10 Sep 2010, Warrigal wrote: > On Thu, Sep 9, 2010 at 3:19 PM, Kerim Aydin <ke...@u.washington.edu> wrote: > > I was also thinking towards the end that it's a pretty good Prisoner's > > Dilemma situation set up. Towards the end (when chance was pretty > > near 50/50) there were a few people who could better their position by > > one by rebelling; then there were some folks who would worsen their > > position by one if they rebelled, but lose by much more if they didn't > > rebel and the rebellion worked... etc. This presupposed people were > > paying attention which not everyone was, but it was interesting to > > analyze and guess whether I'd rebel in their shoes. > > Not every game theory situation is a prisoner's dilemma. The > prisoner's dilemma is when there is one option (defecting) where, for > each player, defecting is better than not defecting, but it's better > for both players if neither player defects than if both players > defect. This, however, is pretty much a game of guess-the-most-popular > option.
No, PD is was exactly at one point. The logic: "All of us non-rebels left will move lower on the list if the rebellion wins. So we (collectively) want the rebellion to fail, so we shouldn't rebel (that's cooperation). However, individually, I want to move the least-far-down, so I (personally) should rebel before the others do (that's defecting). Even if that means we non-rebels collectively get a worse outcome, at least I beat the other non-rebels." That's pretty classic PD, except for the fact that players could try to communicate if they wanted. The fact that the outcomes could be probabilistically weighted to come up with the reward matrix (based on expected outcome) doesn't change the dynamics. -G.
