Love Ekenberg, Johan Thorbiörnson
We present a theory for evaluating decisions under risk when the available information is indeterminate. The probability and utility estimates involved in such a situation is expressed as sets of distributions, representing beliefs in various vectors in the decision space. We also demonstrates some consistency requirements between beliefs in local values, i.e. vectors representing values for a singleton probability or utility variable, and beliefs in global values, i.e. beliefs in vectors in the decision space. The evaluation of the different strategies is performed with respect to a generalisation of the principle of maximising the expected utility. We show that, despite the possible complexity of the various inputs, the computational efforts for evaluating strategies are tractable.