Cross-impact analysis experimentation using two techniques to revise marginal probabilities of interdependent events

Abstract

Cross-impact analysis relies on decision makers to provide marginal probability estimates of interdependent events. Generally, these have to be revised in order to ensure overall system coherency. This paper describes cross-impact analysis experimentation in which a Monte Carlo based approach and a dierence equation approach, respectively, were used to revise these marginal probabilities. The objective of the study was to determine the consequences of such revisions on the expected impact rankings of these events. A cross-impact analysis system was developed and used to conduct the experiments. The experiments show that the impact ranking of interdependent events may indeed depend on the technique used for revising event marginal probabilities. Moreover, the Monte Carlo technique generates a world view closer to the one of the decision makers, while the world view generated by the dierence equation technique diers from that of the decision makers.

Key words: Cross-impact analysis, probability revision, monte carlo simulations, dierence equations.