Value of information in decision systems

Abstract

The value of information (VOI) is an important concept in decision-analytic consultation as well as in normative systems. Unfortunately, exact computation of VOIs in a general or in a dynamic decision model (DDM) is an intractable task. This dissertation first examines the properties and boundaries of VOI in DDMs, and then proposes an efficient method based on reusable sub-junction trees for the exact computation of VOI in DDMs. The dissertation also considers various kinds of approximate computations of VOI in hard POMDP problems. In general decision models, by exploiting the conditional independencies in a graphical representation, an efficient non-numerical search algorithm is developed for identifying partial orderings over chance variables in terms of their VOIs. Finally, a concluding guideline for VOI computation is composed for the decision-makers.