| Abstract: | This paper presents an analysis of Delphi from a Bayesian point of view and brings to attention several hitherto neglected, but highly relevant findings of research in areas other than Delphi proper. The aim has been to seek avenues of improvement in the Delphi technique. This has led to development of a methodology, which is based on a concept of second order probabilities as a measure of one's fuzzy thinking. The justification of this new methodology lies in explicit recognition and implementation of an optimum, determined by the trade-off between advantages and complexities of hierarchial inference. It is essential that conventional Delphi applications, at the very least, be accompanied by a Turoff-type cross impact analysis. It is indicated that the tremendous potential of Bayesianized Delphi in appropriate situations has remained utapped. |
