TYPE OF EVIDENCE AS A BASIS FOR COMBINING SUBJECTIVE PROBABILITIES

Abstract

Two models for aggregating subjective probabilities are presented. One employs a multiplicative rule and the other a weighted average. The choice of a model is based on the type of evidence upon which the subjective probabilities were estimated. An experiment was developed to determine if people are sensitive to this difference in the type of evidence when combining subjective probabilities. Two other variables tested were the tense of the event and the experience of the subject with the use of probabilities. The type of evidence presented had an effect on the combination rule employed, whereas tense of the event did not. The naive and expert subjects approached the problems differently. An order effect due to the presentation order of the evidence within a problem was found. A momentum tendency, which may explain the order effect, was present in the expert subjects. Further research on combining subjective probabilities is indicated.