Abstract
Giving judgemental probability to a non-recurrent event, which is frequently demanded in the field of technological forecasting, is not an easy task. To lessen this difficulty, the present paper proposes to give judgemental probability in two stages: representing the likelihood of occurrence of an event in a linguistic form, i.e. a probabilistic statement, and then converting it to numerical probability with the help of the a priori constructed relationship between probabilistic statements and objective probabilities.
Major problems in constructing this relationship are how to select suitable probabilistic statements and how to scale these. The ranking method based on paired comparison data developed by the authors is adaptable to these problems. For selecting probabilistic statements rank order of statements is obtained for each subject. Those statements for which fuzziness and variance among subjects are large are discarded from consideration. For scaling the remaining probabilistic statements rank order of these statements and objective probabilities is obtained for a group of subjects.
The previous selection procedure applied to 14 probabilistic statements discards 6 statements as being unsuitable based on the paired comparison data of 8 subjects. This method also clarifies the number of distinguishable statements; it varies from 3 to 8 and averages 5 for probability greater than 0.5. Based on the rank order of the remaining 8 statements and 8 objective probabilities for a group of subjects, 5 probabilistic statements are finally selected as possessing desirable characteristics. Applications of the present approach to various case studies in technological forecasting have gained a favorable support of the experts.
