Composing and constructing value focused influence diagrams: a specification for decision model formulation

Over the past three decades, significant improvements in the computer and computational sciences have enabled automated support for increasingly complex decision situations. One example of this progress is the influence diagram, which is simultaneously a graphical and mathematical model of a decision situation. Influence diagrams are a proven asset in the tool kit of decision analysts and they are making the power of decision analytic modelling more accessible to professionals in other disciplines. To develop influence diagrams, decision analysts guide decisionmakers and subject matter experts through a discovery process that necessarily migrates from the unstructured to the structured. These highly complex decision situations often require inputs from many different people with diverse expertise. One shortcoming of influence diagrams (and other methods) is that they do not specify a cohesive and comprehensive process for structuring the interactions of options, values, and uncertainties during the initial development of a decision model. This complicates (and sometimes precludes) the process of developing a comprehensive model because the model reaches a level of complexity that is very difficult for individual domain experts to think about. This paper introduces a formally specified method for eliciting influence diagram structure. We extend influence diagram notation to include special categories of value nodes that more explicitly define fundamental objectives hierarchy components. We then provide provably correct methods for decomposing the model, eliciting additional detail, and reassembling the model into a single graph. These methods make two important contributions to the modelling science: first, it is a necessary step toward providing automated model elicitation tools; second, it is an example of technology transfer from Bayesian Belief Networks to Influence Diagrams.