Bounded Rationality: The Adaptive Toolbox

Rating:
Summary: State of the Art on Behavioral Choice Theory
Review: Suppose we wanted to predict how an expert billiards player would hit a certain shot. We would measure the angles and distances, get the coefficients of elasticity of the balls and the bumpers, and we would solve a set of differential equations. But is that how the billiards player figures out what to do? Of course not! We don't know exactly what he would do, but if the authors of this book had their way, we'd give up on the differential equations (optimization theory) and find the "fast and frugal heuristic" actually used by the billiards player.

This book is the product of a conference of experts in the field. It includes wonderful contributions by the editors and their coworkers on how decisions are actually made, and argues persuasively that fast and frugal is almost as good as full optimization, and at much lower cost.

But the volume is a lot broader than that. It includes contributions on the role of emotions in decision-making (Dan Fessler), learning in animal societies (Keven Laland) and social insects (Thomas Seeley), and a lot of material on the role of culture in human societies (Boyd, Richerson, McCabe, Smith, Henrich, and others). This is important new material, very up to date.

Gigerenzter and Selten go to great lengths to cast aspersions on the old-fashioned "optimization subject to constraints" perspective, but their arguments are not persuasive. They make a category error: they maintain that models that use optimization assume that the agents the models describe use optimization. This is just silly. Just as the billiards player does not solve differential equations, decision-makers do not do complete optimization, even though we may use such models to describe their behavior.

The editors believe that optimization subject to constraints is dead in behavioral theory, but they're dead wrong. That's in fact what they are doing, but they prefer to call it "bounded rationality."

Finally, I should note that the work of Eduardo Zambrano (look up his home page) shows that the SEU (Subjective Expected Utility model---the enemy of all bounded rationalers) actually is behaviorally universal, in the sense that one can always find a set of Bayesian priors for which an observed set of behaviors is optimal.

But don't let these petty methodological issues get you down. The book is a great collection by the authors of major work in behavioral theory.