Heuristics for Decision Making

Published on Apr 12, 2012 by INETeconomics Gerd Gigerenzer, Director, Max Planck Institute for Human Development speaks on panel entitled "What Can Economists Know: Rethinking the Foundations of Economic Understanding at the Institute for New Economic Thinking's (INET) Paradigm Lost Conference in Berlin. April 12, 2012. #inetberlin

I like this presentation. I will nonetheless quibble.

First, let me suggest that we divide the world of decision making into three rather than two categories"

• risk, where you know alternatives, probabilities (and values);
• uncertainty, where you know alternatives (and values), but not all probabilities
• ignorance, where you don't know all the alternatives and values, much less all the probabilities.

Incidentally, assume that you are trying to make a policy decision that will influence outcomes for a large number of people, say in a health situation. Say there are several treatments, and for each treatment you know the distributions of outcomes over time -- how much suffering, how much incapacity, how many deaths. And say you also have statistics on costs. How about the distribution of values assigned by the patients to different outcomes? Some patients may prefer an intervention that has a higher risk of death but a lower likelihood of long term disability, while others would have the opposite preference. Do we have decision models that take the distribution of preferences into consideration? I think not.

The Fly Ball Example

The first example that Dr. Gigerenzer gives may not be well conceived. He compares two calculations:

1. The calculation by a fielder of where a fly ball will land and the simultaneous calculation of how to run to get there, with
2. The calculation of the pursuit path to get the fielder and the ball to meet.

The pursuit path is apparently the one chosen by actual fielders. The good doctor suggests that there is a simple heuristic for the pursuit path -- run in the direction and at the speed that keeps the angle between the ball and the player constant. That would seem to be a result that could be predicted from the equations of motion. It could be an evolved trait, since individuals that successfully followed pursuit paths in hunting or in catching falling fruit would be likely to have higher survival rates and thus more progeny. (There is an episode of the TV program, Numb3rs, that deals with pursuit paths,)

The point is, both calculations would presumably get the same result -- catching the ball. On the other hand, assuming that Dr. Gigerenzer is right, the pursuit model gives a very simple instruction to the fielder to control his speed and direction, and thus is preferable. If one is choosing between different analytic approaches that yield the same result, presumably one wishes to choose the simpler one to implement. Thus accuracy is not the only criterion to use in modelling.

I suspect that the fielder who waits until the ball is at the peak of its trajectory before starting to run will miss a lot of fly balls. There is probably a calculation that is made as to what direction to run when the ball is first hit and the first information is available on its trajectory. There is probably another heuristic on when to switch to the angle of descent procedure.

I am not sure how complex the brain's calculations are. They would seem to me to be almost certainly more complex than our conscious mind realizes.

Final Comments

I do like the idea that sometimes it is better to toss out information rather than try to deal with too much information.

I also like the idea of decision making in which areas of ignorance can not be reduced even to uncertainty -- a common real world situation.