Abstract
The purpose of the paper was an inference process’s formalization – in terms of a rough sets theory – of an existing DOE (design of experiment) supporting expert system DAX. This was the first step in extending the system from determined decision tree and rules-based into fuzzy-logic based inference and considering uncertainty in other approaches than only probabilistic. The paper is primarily addressed to the developers and users of expert systems. The second addressees are statisticians preparing experimental designs.