Probabilistic Knowledge Processing and Remaining Uncertainty

Information is indispensable in preparing economic decisions purposefully. In this paper knowledge is represented by a probability distribution. Knowledge acquisition is realized by the principle of maximum entropy which guarantees an unsophisticated expansion of knowledge and provides an unbiased basis for decision processes. Under a mathematical point of view entropy measures remaining uncertainty and reduction of entropy is knowledge increase (first order uncertainty). In general, an actual epistemic state might vary due to further information. Thus, it is necessary to have an instrument which reflects also the potential of distribution’s possible variation. Therefore, we present the concept of so called second order uncertainty, which allows not only a ’look’ on an epistemic state, as the entropy does, but also ’looks’ into an epistemic state. First the arguments will be complemented by a short example and then they will be illustrated by a real world model in the field of business to business transactions. All calculations will be supported by the expert system SPIRIT.