Creating Algorithms as an Aid to Judgment

36 pagesA critical task often performed by decision makers is to make estimates of important points of fact. Previous research has suggested that decomposition of numerical estimation problems can result in improved estimation performance, particularly when the problem is structured as an algorithm. However, algorithms used in past studies have been provided by the experimenters, rather than created by the estimator. This study reports on the usefulness of algorithmic decomposition when people are trained to create their own algorithms in the context of a task requiring them to evaluate an answer given to them for an estimation problem. The results suggest that people can be trained to construct complete and useful algorithms to verify numerical estimates, but that misinformation about factual knowledge used in an algorithm can seriously bias estimation performance