Eliciting parameters in decision analysis and fuzzy rationality

A general scheme of parameter elicitation in quantitative decision analysis is presented. It is demonstrated in examples for elicitation of utilities, probabilities, quantiles, and scaling constants. Real decision makers disobey axioms of rational preferences and elicit uncertainty intervals instead of point estimates. That leads to partially non-transitive preferences, and to the introduction of fuzzy rationality. Its influence over the elicitation process is described both within the general scheme and in examples. The interval estimates of utilities and probabilities need to be approximated by point estimates of some sort. As demonstrated, approximation of utilities in fuzzy rational decision analysis does not depend on the probabilities, but the approximation of probabilities is dependent on preferences.