Towards a More General Theoretical and Mathematical Model of Probability for Policy Analysis

Abstract

Uncertainty pervades policy analysis in ways that transcend classical concepts of probability. To benefit policy analysis, the concept of probability must be considerably broadened. It is argued that probability can be conceptualized with respect to the characteristics of policy problems that produce inherent uncertainty. Problems that encompass uncertainty can be characterized according to their: (1) fundamental requirements, for example forecasting, knowledge creation, fact establishment; (2) system properties such as disorderly versus orderly systems; (3) problem-solution strategy, for example subjective judgement, model-based analysis, data analysis; (4) problem-solution data requirements—from numerous and hard-to-measure variables to few and easy-to-measure variables, and (5) problem-solution frame—ranging from unbounded solution spaces to small and discrete solution spaces. The theory of lower probability is presented as a generalization of classical additive probability that can handle this generalized conceptualization of probability. Information-theoretic methods for integrating the two generalizations of probability are considered.