“In that case you’re in an equilibrium where changing to deny leaves you worse off, so in a single iteration you should always confess, not deny. I guess the confusion arises from the slightly non-conventional wording of the PD given at the top.”
Yup, the wording got me. Didn’t look closely, it kinda looked like the usual set up, which it isn’t. I think we can both agree that this is the usual treatment is this: http://en.wikipedia.org/wiki/Prisoner%27s_dilemma#Strategy_for_the_classical_prisoner.27s_dilemma

The one above, now that I look closely, is even more weird: The way I read it, using the colour coding, if one player confesses the crime and the other denies it, the one denying gets 10 years, the one confessing gets one?! Better stay with the standard one, that is less confusing…

Then again, this does happen in real life. How many profitable business deals are thrown out because one party thinks they are being taking advantage of – even if the amount on offer is more than enough to make it worthwhile? It’s the whole notion of “fair”.

Also worth considering the classic research (which I can remember the exact details of), but basically proved that people need nothing more than the idea there is “another”, to feel hostile towards it.

Min

Yeah I’m familiar with it. I do question what you say about it though.

For one thing, the strategy has to be the same whether you’re A or B so I don’t know why you say “if you’re prisoner A”. For another, the optimal strategy is widely recognised to be both confessing. In that case you’re in an equilibrium where changing to deny leaves you worse off, so in a single iteration you should always confess, not deny. I guess the confusion arises from the slightly non-conventional wording of the PD given at the top.

In the multiple iteration case, the dominant strategy remains ‘always confess’ but in practice ‘tit-for-tat’ does indeed work better – I think its fair to call it a near-optimal solution to the problem in these cases.

2. @edd:
“While it is rational to take a small offer if you only consider the one game, it may well be rational to refuse it if you expect ever to have dealings with that person again”. Yup. But this is not the game. The game is about one iteration only. The point of the one iteration is, imho, to show that people make too many assumptions and therefore do not act rationally. Because in the context of that one game, any decision other than offering a very small amount (sounding smart: A very small epsilon) and, for the recipient, to take even that small amount, is irrational.
Several iterations will change the game, of course. As will other, arbitrarily introduced rules. But that is not the point.
The game pictured above, the prisoner’s dilemma, might be interesting to you, because the solution to the one-iteration-version (always deny if you’re prisoner A) is different from the result of the multiple-iteration-game (where tit-for-tat usually wins)

3. @Mrs G:
“eheheh .. what dum game is that? It’s too simple stated here …”. It is that simple. That is the point of it. There is nothing except these rules. That is what makes it so easy to analyse. All the reasons you quote, from morals to needs, are not part of the game. They are just parts of the rationalisation process (read: making up excuses) for choosing a non-optimal (read: wrong) solution.

In a one off theoretical situation as above then the outcome is not that unexpected – the question the subject is being asked is ‘do you mind if someone takes more theoretical money than you’

Put the cash on the table and explicitly tell the participants that they get to take their share of the £100 home with them and the situation would become far more interesting. Would it tend towards big offers so that the chance of acceptance of the offer would rise, or small offers as it’s still something for nothing?