Revealing Uncertainties in Computer Models

摘要

As computers have become faster and smarter, scientists have used them to build models of complex phenomena—every- thing from wildfires on rugged terrain to traffic snarls on urban streets. These simulations can yield precise answers to problems a person with a pencil ana paper would need millennia to solve. But just how reliable are these virtual solutions? Although the precise numbers and realistic pictures produced by computer simulations give an illusion of accuracy, a ravening swarm of assumptions, simplifications, and outright errors lurk beneath. New tools are needed, scientists say, to quantify the uncertainties inherent in calculations and to evaluate the validity of the models. But, as several recent uncertainty workshops attest, the quest for such tools is itself an uncertain and challenging process.