Multiple attribute scenarios, bounded probabilities, and threats of nuclear theft

Abstract

A method is presented for developing descriptions of future scenarios and using expert judgment to assess bounds on the probabilities of these scenarios. Multiple attributes are used to describe the important features of the scenarios, and the scenarios are defined as collections of different possible levels of the attributes. Experts assess either numerical values or bounds on various unconditional and conditional probabilities for different attribute levels. These are used to establish constraints for a series of linear programs which are solved to determine the highest and lowest possible probabilities for each scenario. An application is presented to the assessment of potential threats against nuclear material safeguards systems.