Last time we finished up our Prisoner’s Dilemma experiments by tossing all of our prisoners into a big arena and then having them compete in a multi-generational survival-of-the-fittest pit fight in order to see which strategies have the best long term survivability.
Now it’s time to see the final outcome, but nobody wants a blog post filled with nothing but thousands of lines of pure numbers. Instead I’ve dumped my test results into a graphing program program and built this nice little animation for you.
So what did we see happen here?
Things started out about like you’d expect. The first few generations saw the devils taking advantage of the naïve saints and quickly becoming the majority of our population. They then steamrolled their way through the madmen who just aren’t smart enough to purposely avoid being taken advantage of by the devils.
We’re now left with a population that’s 75% devil with a mere handful of judges. But judges ARE smart enough to notice when people keep cheating them so we see a sudden reversal. The judges stop feeding free points to the devils while simultaneously working together to keep their score high. A few more generations is all it takes for the devils to be completely wiped out by the superior teamwork of the judges.
Wasn’t This Supposed To Have Real World Applications?
Waaaaay back in part one I mentioned that the main goal of game theory was to use simple math games to analyze real world decisions making.
So what real world lessons can we learn from all our experiences with the prisoner’s dilemma?
First, that a group full of trustworthy people is a good thing for everybody.
Second, that a cheater can easily exploit a group full of trustworthy people for massive gain.
Third, that having too many cheaters isn’t just bad for the group, it eventually leaves the cheaters worse off too.
Fourth, that a reliable way to prevent cheaters from messing up a group is by identifying and then punishing them until they either reform or leave the group.
These four observations together explain the rise and fall of a lot of real world organizations. You start off with a group of trustworthy people who work together to accomplish something. Eventually a cheater infiltrates the group and realizes that he can exploit the good will of the group for personal gain. Other cheaters eventually notice the scam and decide they also want to join in on exploiting the organizations.
At this point one of two things can happen: Either the group notices the cheaters and manages to remove them before irreparable harm can be done or the group fails to stop the cheaters and collapses under their corrupting weight (possibly after limping along for years or decades).
But why do so many groups fail to stop the cheaters?
Sometimes it’s because the group is just too nice for their own good. They feel bad about having to punish or expel anyone, so instead they just put up with the cheaters until it’s too late.
But more often than not the problem comes from the fact that identifying cheaters is really hard.
This is, in fact, one of the big limitations of the Prisoner’s Dilemma: It assumes you can tell when people are cooperating with or betraying you. But relatively few real world situations fit that pattern. Usually it’s really really hard to tell between a good natured cooperator and a really sly betrayer. A competent cheater can get away with dozens of mini-betrayals before anybody figures out what’s really going on.
Consider a politician who helps pass a popular law, but only after adding in a few sneaky sections that will help slowly funnel money and power to his buddies.
Consider a CEO who generates record profits in the short term by knowingly sabotaging the long term health of the company that hired him.
Consider a forum member who claims to be interested in giving people constructive criticism but in reality is a troll who just likes frustrating people by picking apart everything they say.
Knowing that the tit-for-tat Judge strategy works doesn’t actually do us much good unless we happen to have a talent for judging the behavior of others. And that is sadly a feat simple game theory can’t help us with.
Don’t be a cheater. In the short run it might seem like a good idea but in the long run the odds are pretty high you’ll get stuck in a downward spiral with other cheaters and wind up worse off than if you had played it straight all along.
Don’t be naive. If there are no consequences to taking advantage of you cheaters will eventually drain you dry.
Use your judgment. Helping helpful people and avoiding cheaters seems to be the most reliable path to both individual and group success.
Acknowledge that figuring out who is helpful and who is a cheater is much harder than our simple simulations made it seem. Real life is messy.
Crack open a history book and find your favorite failed civilization, company, club or whatever. Do some research into the specifics of it’s rise and fall. Did it start out as a bunch of people cooperating towards a common goal? Can you spot a point at which cheaters started infiltrating? Was there a final tipping point where too many people were scamming the group and it could no longer hold up?
Now look up a group that seems to have survived the test of time. What did it do differently? Did it have some sort of rule or tradition that helped prevent cheaters from taking advantage of the group?
Remember, real life is messy so it’s entirely possible that whatever real life examples you find won’t quite fit into the standard pattern of an iterated prisoner dilemma. In which case you now have a great excuse to dig deeper into game theory and see if any of the more advanced scenarios do fit your historical example.