AbstractMany algorithms have been developed for multiple-criteria decision-making problems. Goal programming (GP) is one of these algorithms. This model is a special extension of linear programming. Usually, it is not easy for the decision-maker to choose his aspiration levelsa priori. Moreover, the incommensurability of the measurement units of the various objectives creates an aggregation problem. However, in the standard GP formulation, the decision-maker is not required to arbitrate among conflicting objectives. To deal with these difficulties, we explicitly introduce the structure of the decision-maker's preferences into the GP model in order to evaluate the impact of deviations from the decision-maker's aspirations levels. Easily and naturally, the idea of a generalized criterion, as introduced in the Promethee outranking method, will be used to build this structure of preferences.