Håkan L. Younes, Carnegie Mellon University, USA; Love Ekenberg, Mid Sweden University, Sweden
Today there are numerous tools for decision analysis, suitable both for human and artificial decision makers. Most of these tools require the decision maker to provide precise numerical estimates of probabilities and utilities. Furthermore, they lack the capability to handle inconsistency in the decision models, and will fail to deliver an answer unless the formulation of the decision problem is consistent. In this paper we present an algorithm for evaluating imprecise decision problems expressed using belief distributions, that also can handle inconsistency in the model. The same algorithm can be applied to decision models where probabilities and utilities are given as intervals or point values, which gives us a general method for evaluating inconsistent decision models with varying degree of expressiveness.