Bibliografie

(eng): Traditional decision theory dealing with uncertainty is usually considering criteria based on expected values, or, variantly, on selected quantiles of objective function. Both, however, take into account just rather small part of available information, in particular not counting with possible variability of involved random variables. That is why the criteria based simultaneously on a set of reasonable characteristics should be preferred. This leads to a multiobjective problem and solution based on an appropriate utility function. In the present paper we propose quantitative characteristics measuring information loss caused by reduction of information used in decision. Such measures can help us to find a trade-off between the decision problem complexity and its reasonable simplified re-formulation. This concept is illustrated on examples.