Max Henrion, Ph.D
Founder and CEO of Lumina Decision Systems, Inc.
Coping with Uncertainty and Surprise in Science and Forecasting
December 3, 2008
Pacific Forum – 3:00 p.m.
Historical measurements of the speed of light and other fundamental physical constants reveal a systematic underestimation of uncertainty about the error. Retrospective comparison of model forecasts of energy usage and energy prices with what actually happened show similar - if larger - rates of surprise. These results are consistent with cognitive psychologists' experimental results on subjective probability intervals for almanac quantities that show consistent "surprises". It's important to be explicit about uncertainties when reporting measurements or making forecasts - and to recognize that estimated uncertainties are usually a lower bound on actual uncertainties. The National Renewable Energy Lab and DOE is developing Monte Carlo computer model to explore how rapidly the US could reduce greenhouse gas emissions. It uses expert estimates of the uncertain effects of R&D on the performance and cost of renewable energy technologies. If we want to use such models to guide decisions on energy and environment, it's helpful to see how and why we've been surprised in the past.