A stochastic multicriterion problem in a risk aversion context is defined by a finite alternative set A and an attribute set E. It is assumed that the performance probability distribution for each alternative on each attribute is known. the approach takes roots from Martel and Zaras who built crisp outranking relations based on stochastic dominances (FSD, SSD, TSD) without distinction between the prevailing types of dominance. In the suggested approach, one supposes that a decision maker’s preference between two alternatives on each attribute is determined by his perception of the probabilities. In turn, this perception depends, among other things, on the level of overlapping of the compared distributions and is expressed by degrees of preference (local) as measured from three functions connected with the prevailing type of dominance. The degrees of preference (local) are the aggregated to build an overall preference relation between each pair of alternatives. The approach could be an appropriate method, e.g. in finance, for stocks selection and in strategic planning.
