The use of Bayesian techniques in Delphi forecasts

Procedures are developed which allow an estimation to be made of the uncertainty associated with Delphi forecasts and which allow the judgment and intuition of the decisionmaker using the forecast to be incorporated into and impact upon the forecast results.

Abstract

Abstract Procedures are developed which allow an estimation to be made of the uncertainty associated with Delphi forecasts and which allow the judgment and intuition of the decisionmaker using the forecast (and/or statistical data on the accuracy of past Delphi experiments) to be incorporated into and impact upon the forecast results. The methodology utilized relates the confidence probability of the forecast of an event to the dispersion of individual estimates of the elapsed time for the event to occur. The confidence probabilities of individual estimates are then modified by the outcome of the group estimate and the decisionmaker's subjective judgment through the use of Bayesian techniques. The final forecast is obtained as the sum of the individual estimates weighted by their modified confidence probabilities.