Robustness Theory: Holy Grail or Voodoo Science?

摘要

Robust. Its dictionary meaning is simply “strong” or “vigorous,” but to some it means more,rndescribing a cutting-edge approach to infrastructure management or policy-making, a sort ofrnholy grail for modern times. To most of us, “robust” sounds like something we want our utilitiesrnto be – even if we’re not sure exactly what it means.rnDefinitions and theories vary, but the basic concept of robustness is that, in the face of anrnuncertain future, there is more value in achieving “good enough” performance with very highrnreliability than “optimizing” performance for a single calculated future that will not likely occur.rnFramed this way, robustness certainly sounds like a great idea. Indeed, many talented scientistsrnand practitioners have developed theories and methodologies for robustness. It behooves us asrnutility managers to understand robustness, its current value to us as managers, and its futurernimplications.rnA first reason why robustness is so appealing to us is that it seems to be how nature itself seemsrnto work. Many organic systems, from ant colonies to ground squirrels, seem to employ some ofrnthe basic tenets of robustness. Ant colonies continue to function even after suffering devastatingrnlosses. Squirrels, rather than pursuing a simple optimization goal such as “maximize caloricrnintake,” appear at times to follow robustness-centered rules that may maximize their chance ofrnsurvival even when things don’t go according to plan.rnA second reason why robustness is appealing is that it appears to address an ingravescent need,rnthat of coping with deep uncertainties such as those posed by climate change. Uncertainties ofrnthis scale prompt nearly everybody to ask themselves “what if” the future turns out differentlyrnthan predicted by a given model, and to desire solutions that work well for a wide range ofrnfutures even if not optimally for any specific future.rnA third reason for the appeal of robustness is its potential to solve a problem we’ve brought onrnourselves. The 1980s brought electronic computing power to the desktop. We suddenly had thernpower to “optimize” nearly everything, from the size of a pipe to the maintenance schedule for arntruck. With that power of optimization came a new risk, the risk that we optimize for a particularrnfuture that likely will not occur while leaving performance against other likely futures poor orrnunknown. In short, the risk is that the power of computing has sharpened our vision forrnindividual trees and potentially blinded us to seeing the forest.rnThis paper surveys the available tools for robustness in the context of utility managementrnchallenges. Theories discussed include Robust Decision Making theory, info-gap satisficing, andrngenetic algorithms. Applications discussed include climate change adaption.