Development of software for decision analysis

Abstract

To be useful in reality, decision analytical tools must be able to handle imprecise information. This paper presents the algorithmic software design against a background of an evaluation method for analysing decision situations under semi-strong uncertainty. The design is built on a relaxation of the requirement for precise utilities, probabilities, and weights. To handle this, the calculations involved become computationally intensive to match an interactive work flow and the approach required implementation of new algorithms. We describe some particularly interesting implementation aspects of these and show how these computations can be tractable.