Tuesday, May 19, 2009

A Knew Kind of Science

Recently I've watched a few videos from the Perimeter Institute for Theoretical Physics recent conference on The Economic Crisis and its Implications for The Science of Economics. Some of the stuff is super complicated, as you might expect from a bunch of theoretical physicists. So complicated that even their fucking website is too complicated to use unless you have a PhD, so I'll just mention the link to the various presentations that they have at the Edge. So far I have only taken a look at those from Taleb, Leigh Tesfatsion and Eric Weinstein. Taleb says the same thing he always says, though in a slightly more rigorous fashion -- we don't know shit about shit. While certainly true, and bizarrely contrary to what most folks in economics think, I fail to see how this is especially enlightening. Still, the presentations is fun and data rich and worth on hour if listening to the guy repeat himself doesn't already make you nauseous.

Leigh Tesfatsion made some interesting remarks about how to start using agent based models in economics. Some of this stuff definitely makes you say "duh" if you have not had your brain poisoned by modern economics, but the idea of simulating how an agent based system can reach equilibrium, rather than simply assuming that it does, turns out to be pretty revolutionary thinking for the discipline. The best part of her presentation was the abstract way she was using the concept of agent. She wasn't simply trying to perturb the classical image of homo economicus with a little bit of irrationality, a little imperfect information, and then see what the simulation produced. She wanted to be able to have higher level agents, agents composed of agents (by the way, why isn't this idea in the matrix?), that wold represent things like corporations, governments, unions, etc ... This came out naturally in her presentation because she simply thinks of all agents as just objects in her programming.

Weinstein's skit was more intriguing to me. I don't know how useful gauge theory is going to be in economics anytime soon (hell, there are people who wonder if these equations are worthwhile even in current physics), but I don't think that your opinion one way or the other is crucial to appreciating the talk. The gist is that markets are an animal behavior pattern that you can study with very precise numbers. This makes economics a sort of marriage of biology and physics, one dealing with adaptation and selection and the other with how systems with interacting particles tend towards (or away from) meta-stable equilibrium. The guy is clearly very clever, and their are a lot of great quotes are packed in there -- "humans should be subclassed from primates in economic models", "there is only one truly free market" (followed by a picture of the sibicidal behavior of the masked booby), and my favorite, "economics is too sentimental" because it assumes that the self-interest of humans has an almost Victorian invisible fence that somehow magically prevents it from spilling over into grisly violence. Instead it assumes some sort of benevolent invisible hand that forces it to settle down into some self-correcting equilibrium, coincidentally, one that fits with our current moral values. WWII was a great program for full employment from an economic perspective.

Real food for thought, even if computational economics is going to be pretty tough to get off the ground.

UPDATE: I have hacked the overwrought organizational underpinnings of the Perimeter Institute website, and all the videos can also be found here. No word yet regarding my honorary PhD for this feat.

No comments: