470 – Statistical Methods for Advancing Defense and National Security
Temporal Perceptions and Heuristic Adjustments in Short-Term Forecasts
John Irvine
Charles Stark Draper Laboratory
John R. Regan
Draper Laboratory
Decision makers in many fields rely on the predictions, including forecasts from subject matter experts. New research is exploring methods for eliciting and combining judgments from multiple experts to arrive at a better overall decision. This study explores the temporal nature of expert forecasts and proposes methods for improving forecast accuracy through temporal adjustments. Consider the forecasting problem of the form: "Will event X occur before the date T?" As the date T approaches, rational forecasters should adjust their predictions, but evidence indicates that most people do not accurately account for these temporal changes. We present a heuristic method called automated updating to adjust individual forecasts. Two related versions of the method are presented. Comparing the performance of the proposed methods to the standard unweighted linear average from the pool of subjects demonstrates the benefits of this approach. Once the outcome of the forecasting problem is known, the Brier score provides an objective measure of performance. Based on the Brier score, these methods outperform the unweighted linear average across a number of forecasting problems.