Fuzzy specification in software engineering

Abstract

Judging quality of any decision making procedure is a key problem whenever there is no possibility of developing a sequence of experiments allowing some kind of ratio relative to good results. It may be the case that we have only chances for a unique experiment or no similar experiences are available, but it may be also the case that no standard experiment allows the observation of such a good behavior, simply because such good behavior can not be properly defined. This situation is quite often associated to complex decision making problems. Then the only support we can find for our decision is the decision process itself, the consistency of the arguments leading to such a decision. Checking the quality of such a procedure is therefore a key issue. In this paper we postulate that the design and formal specification of algorithms and processes require a fuzzy approach, since quite often specification is being poorly dened.