The video above is from the Institute for New Economic Thinking (INET) and features a provocative talk by Gerd Gigerenzer, Director of the Max Planck Institute for Human Development. I was not familiar with his work before this video, but I enjoyed his talk very much.
Gigerenzer places himself in opposition to much of economics, including the new fields of neuroeconomics and behavioral economics. He claims they “look down” upon certain types of human behavior as irrational, but Gigerenzer sees things differently. He praises the use of heuristics—simple experienced-based approaches to solving common problems.
He uses as an example an outfielder in baseball. The standard behavioral model posits that baseball players act as-if they are solving a complex set of differential equations to determine the path of the ball. Economists know that in practice there is no conscious (or even subconscious) effort to do so. But, alas, it doesn’t matter. The player acts as-if they have solved such problems and behaves accordingly. Gigerenzer disagrees. He introduces a simple tool, what he terms the “gaze heuristic,” and claims that it models behavior more accurately. The gaze heuristic is simple—a player runs toward the ball as it is in flight, adjusting his path toward the ball so that the ball is at a constant angle in his view. This may mean running toward the ball or running away from it.
In practice, Gigerenzer claims this heuristic better models the behavior of baseball players. The as-if model, for example, would predict that, having already “solved” the equations for the trajectory of the ball, the player would quickly run to where the ball was going to land and simply wait for it to arrive. In practice, however, players run toward the ball more in line with what the gaze heuristic would predict. Here, I am not sure if Gigerenzer is correct. I don’t watch much baseball, but I have a feeling that if an outfielder saw a ball hit and then quickly closed his eyes he could run to a spot on the field very close to where the ball would actually end up landing. Any baseball fans care to comment?
Gigerenzer also makes a key distinction about decisions under risk and decisions under uncertainty and between curve-fitting and prediction. For instance, he claims that a 1/N model for stock investment (where one simply distributes his investment equally between N number of total stocks) works better under uncertainty than the “Mean-Variance Model,” a complex statistical technique. The latter model is very good at curve-fitting, but needs a lot of data to work well as a prediction model. Thus, in most cases the 1/N model is better (yields higher returns) under uncertainty. Gigerenzer claims it would take nearly 500 years of data for the Mean-Variance Model to consistently predict a stock distribution that outperforms the 1/N model.
I’m very much looking forward to reading some of Gigerenzer papers and journal articles.
(HT: Coordination Problem)