Thursday, September 13, 2012


To quote the now defunked Mr John Blog's 'if you want a good explanation of game theory read Richard Dawkin's "The Selfish Gene"' which isn't a quote at all but a paraphrase because I couldn't be bothered looking up the post (I think it was about Pascal's wager) on game theory type stuff.

Anyway, the prisoners dilemma, the old chestnut of game theory, let's review for those who don't play games or alternatively think playing Angry birds makes you a gamer.

Two friends are held prisoner. They are kept isolated from eachother. They are each offered the same deal - if you cooperate and testify that your friend is guilty we will reduce your sentence to a year. If you don't cooperate, you get 10 years, BUT the thing is that they have no hard evidence on you or your friend, if neither of you testify you both walk free.

Here are the various outcomes.

Both cooperate - 1 year sentence each. AKA lose-lose
One cooperates - 1 does 1 year, the other 10 years AKA win-lose, lose-win.
Neither cooperate - both walk free AKA win-win.

Accept presumably justice, these win-lose evaluations are all from the prisoners perspective.

The thing is that because neither prisoner knows how the other will react, game theory predicts that their desire not to be the 'sucker' - the one who naively cooperates while their friend is ratting them out for a 10 year sentence, they will cut their losses and cooperate, so both cooperate and both lose a year of their lives.

But what I like to believe, and law enforcement has discovered with organised crime and such - is that cooperating is rarer than you would think. People are far more trusting, or you will discover have crazy value systems like 'they would rather do 10 years in prison than betray a friend' making processes like waterboarding and 'going to the dentist' popular.

But anyway, the point is that you have all these permutations of outcomes of the various 'games' we play. Dawkins applies it to evolution, many apply it to gambling/investment (it's hard to tell the difference these days, gamblers though tend to be less superstitious) but I had like an epiphany yesterday of no practical value, just descriptive value. You see -


these are the outcomes. Believing people to be good and decent, you could say that everybodies 'golden rule' in life is to create win-win situations. Of course, some people get trained into the naive belief that somebody winning entails somebody losing, but this is not what has built all the good shit in our civilisation.

That's one except, but it occured to me that most people act on the assumption that people want to 'win' and generally in interacting with others we defend against 'win-lose' or 'lose-win' situations and aim for 'win-win'.

My housemate of yesteryear wrote a whole honours thesis on whether somebody could have agency over an act they genuinely believed to be wrong, and now I will do him a slap in the face by just asserting that somebody may actually want to lose. That is what fucks you up, throws you off your game.

Of course, I don't believe, or choose not to believe that people who play to lose, or tank, have agency over it. They simply have no concept of the winning paradoxes.

Salient in my mind, almost always are Mark Horstman's 'Avoiding failure is not the same as seeking success' (failure and success have pretty much identical preparation processes a lot of the time) and Dr Gordon Livingston's paradox 'the ultimate risk is taking no risk at all'

and I think ultimately, those that to me appear to be playing to lose, are most often actually playing 'not to lose' in their own minds, not realising that playing 'not to lose' ensures losing.

And remember those hypothetical people I asserted that say things like 'I would rather do 10 years in prison than betray a friend' they fuck up the game because to them, their friend winning is equivalent to them winning. These people live in a world where almost every game looks 'win-win' to them, and are genuinely the most wonderful people. Economists tear their hair out trying to figure out 'rational incentives' for them to play the game.

But when these people meet somebody who plays to lose - they get caught out by the lack of permutations. When you are playing with somebody who plays to lose you only have:


This causes economists AND real people to tear their hair out. It's a crazy fucking game. Your friend is sitting in prison, and you have to accept that the moment the guard tells them 'you're free to go' they are going to throw themselves down and confess, anything to keep the security of prison around them. So what the fuck do you do?

No comments: