A0654
Title: Structured expert judgment for sequential decision-making
Authors: Tina Nane - TU Delft (Netherlands) [presenting]
Abstract: Expert opinion is typically employed to support decision-making when data are not (or no longer) representative or simply unavailable, in high uncertainty contexts. The classical model for structured expert judgment (SEJ) provides a structured and validated methodology to elicit and mathematically aggregate experts' uncertainty assessments. Probability distributions are constructed from the (typically three) elicited percentiles. Uncertainty assessments of calibration variables, for which realizations are known to the analyst but not to the experts, enable a validation step, where objective measures indicate how statistically accurate and informative the assessments are. These measures yield performance-based weights, which are used to aggregate experts' assessments into distributions for variables of interest. The classical model for SEJ has been applied in more than 250 professional studies across numerous domains, spanning from epidemiology, natural hazards, and climate change, to energy (transition). Most of the studies are a one-time application of SEJ. The aim of this research is to extend the classical model for SEJ to a sequential context. Through the dynamic nature of the data, when realizations become quickly available, the framework enables the sequential validation of individual and aggregated assessments. Performance metrics are derived to support sequential decision-making. Results from two studies are presented.
