IMP: A decision aid for multiattribute evaluation using imprecise weight estimates

• Provides a simple way to incorporate uncertainty in multicriteria evaluations.
• Allows for uninformed and point only weight estimates.
• Greater uncertainty in multiple judgements than self-assessed uncertainty.
• Monitors the effect on imprecision of aggregation of individual assessments.

In multicriteria analysis, weights are inferred from judgements. Judgements are subject to uncertainty and the weights found depend on the method used so that weights must necessarily be imprecise. Describing weights by a suitable probability distribution captures this imprecision and leads to correspondingly imprecise estimates of the multiattribute scores given to alternatives. A method is described which enables the parameters of a Dirichlet distribution to be found from three point weight estimates and for this to be used to determine whether or not alternatives may justifiably be differentiated by their scores. A simple chart shows this information and permits easy interaction. An illustration shows the effects of different sources of imprecision and of aggregating individual assessments in a group.