Great debate attempt #4: Some behavior really is random

  A few years ago, me and a friend scribbled down something we called “A pure theory of relationships,” based on a simple game-theory-based hostage model that we developed for this book. It was a joke, not meant for prime time or to be taken seriously.

And now I find that people are in fact using their time to seriously develop such models.  Me, I agree with James Kwak.  People are complicated, often driven by desires they don't understand, and small events can lead to behavior that is effectively random.  There are human situations in which people with the exact same set of outwardly observable characteristics will behave completely differently, and human situations in which a small unobservable change will cause the exact same person to behave entirely out of character.  Then add in the fact that some behavior (like, say, smoking) is addictive for some people but not for others, and exposure to the addictive trigger by the susceptible is a stochastic process. Trying to come up with a reduced-form model to predict such behavior strikes me as nuts.  Sure, you can come up with some sort of probalistic model for large populations, which is useful ... but not what the new modellers are trying to accomplish.

But what do I know? 

Comments

[Bernard Guerrero]

Ok, so what's the Great debate supposed to be about?

If the point is that you think some situations can be expressed, at best, as grossly simplified toy models, I can't see that I disagree. That said, toy models are meaningful if they highlight major effects, however much additional complexity there is in the real world.

If the point is that the definition of rational used for homo economicus is too simple, I'd probably agree.

If the point is that some models will only work for large populations, well, professional experience leads me to agree. I can give you probabilities for individual likelihood of default, but they're only meaningful for big groups. A 70% chance of FPD does not mean a certainty. But that's a given.

If the point is that some phenomena are just too complex to try and model, I dunno. There's a lot of computing power out there. It might not be a pretty model, but....