How Much of the Future will turn out to be Like the Past?
Image via Wikipedia
Nassim Taleb is popular for for his tart but cogent counter examples found in the investment forecasts.
So far, the critiques simply focus on what is wrong with rules that assume that the future will closely resemble the past, without providing guidance for a world in which the future doesn't closely resemble the past.
Finally with, Gerd Gigerenzer, in Gut Feelings, the Intelligence of the Unconscious, we have some serious suggestions about what sort of decision rules work in an uncertain world. An uncertain world differs from what we think of as a risky world, the latter has well known mathematical models based on, in part, on balls in urns and counting the chances of getting a particular ball after a number of drawings.
Gigerenzer writes of a different type of rules:
"In an uncertain environment good intuitions must ignore information."
The one example that I found useful was this. In discussing a rule of thumb for predicting school drop-out rates, Gigerenzer contrasts a simple rule: Step 1: check attendance rates, if about the same move to Step 2. Otherwise pick the school with the highest attendance rate. Step 2 Pick the school with the best writing score.
Ok, that is an interesting rule. But how well does it stack up against complex weighting and optimization of a number of factors?
According to Gigerenzer, the simple rules predicts less of the past, but more of the future. It is a better predictor than the complex regression rule.
"In an uncertain world, a complex strategy can fail because it explains too much in hindsight. Only part of the information is valuable for the future, and part of the art of intuition is to focus on the part and ignore the rest."
Today, Ian Ayres has another example about a simple rule, albeit an regression rule. which beat the combined intelligence of the experts.
"The NBA draft this year provides a vivid real world test of whether very simple regressions can out-predict experts on a central business decision -- the NBA draft."
My sense is that Gigerenzer and his researches are on to an important distinction between risk worlds in which Bayesian models are the paradigm of reason and our uncertain worlds in which the future does not unfold in a calculated regression from the past. (This despite Tyler Cowen's view that "The author is a smart guy and an accomplished scholar, but despite his best efforts this book is a few years too late.")
Buy this book today.
