A probabilistic cross-impact methodology for explorative scenario analysis

Abstract

As one of the approaches to scenario analysis, cross-impact methods provide a structured approach to building scenarios as combinations of outcomes for selected uncertainty factors. Although they vary in their details, cross-impact methods are similar in that they synthesize expert judgments about probabilistic or causal dependencies between pairs of uncertainty factors and seek to focus attention on scenarios that can be deemed consistent. Still, most cross-impact methods do not associate probabilities with scenarios, which limits the possibilities of integrating them in risk and decision analysis. Motivated by this recognition, we develop a cross-impact method that derives a joint probability distribution over all possible scenarios from probabilistically interpreted cross-impact statements. More specifically, our method (i) admits a broad range of probabilistic statements about the realizations of uncertainty factors, (ii) supports the process of eliciting such statements, (iii) synthesizes these judgments by solving a series of optimization models from which the corresponding scenario probabilities are derived. The resulting scenario probabilities can be used to construct Bayesian networks, which expands the range of analyses that can be carried out. We illustrate our method with a real case study on the impacts of three-dimensional (3D)-printing on the Finnish Defense Forces. The scenarios, their probabilities, and the associated Bayesian network resulting from this case study helped explore alternative futures and gave insights into how the Defence Forces could benefit from 3D-printing.