One of rapidly developing areas of Artificial Ántelligence is managing uncertainty. Át is not surprising that intelligent systems are expected to be able to exploit uncertain or vague information since also humans often have to reason and decide without having precise and certain information. One can distinguish between two basic types of uncertainty: one caused by uncertain or vague information, another by unknown, imprecise, or stochastic relations between variables that are part of a model of a reality. Diverse frameworks were proposed to tackle the challenging problem of reasoning with uncertain and vague information: Dempster-Shafer theory of evidence [3, 11], theory of imprecise probabilities [14], possibility theory [16], fuzzy set theory [15], etc. Án this paper we deal with the standard probabilistic inference. Át means that all uncertainties considered in this paper are treated in the sense of randomness and will be quantified and processed by the means and tools of classical probability. For updating we use the standard Bayes rule. The knowledge of a modeled domain is represented by a probability distribution r (V) defined for all combinations of values of variables from a set V. The initial probability model r (V) is usually built using data measured for a population of individuals, repeated events, etc. The built model can be used for individual case analysis-