Steven Popper and his colleagues at the RAND Corporation have a thoughtful perspective on computer models and their uses in decision making, which he describes in a letter in this week’s Economist:

SIR – Your excellent report on economic models raises troubling questions for both the builders and the consumers of such models (“Big questions and big numbers”, July 15th). The root of many problems lies not in the models themselves but in the way in which they are used. Too often we ask “What will happen?”, trapping us into the mug’s game of prediction, when the real question should be: “Given that we cannot predict, what is our best move today?” This subtle shift in emphasis from forecasting to informing resolves many of the conundrums you raised.

Instead of determining the “best” model that solves optimal strategies we should instead seek the most “robust” model that achieves a given level of “goodness” across myriad models and uses assumptions consistent with known facts. My colleagues and I use such methods to address intractable policy issues fraught with arguments over which model is “right”, what assumptions are valid and what is the nature of the good? This method makes the decision to be informed part of the analysis itself and the results are more readily accepted by policymakers.

Steven Popper
Senior economist
RAND Corporation
Santa Monica, California