Résumé
When making a choice or selecting a solution, both positive and negative goals may have to be considered. Negative goals refer to what should be rejected (because it is unacceptable). Ileir negations thus correspond to constraints which define what is acceptable and should be satisfied. Positive goals rather express what is really looked for, what is desired. Clearly, a possible choice which is not rejected is not necessarily desired, and although a choice fulfilling a positive goal makes it more satisfactory, reaching a positive goal is not compulsory. Ilis bipolar view of goals may be a matter of degree and can be represented in the possibilistic logic frainework by separate sets of prioritized formulas describing at the syntactic level, what is rejected on the one hand, and stating what is particularly desirable on the other hand. Prioritized negative goals are semantically equivalent to flexible constraints, while prioritized positive goals rather have criterion functions as sernantic counterparts. Preferred choices should satisfy integrity constraints if any, should satisfy the constraints induced by negative goals (or in case of inconsistency, only the ones with low priority may be forgotten), and as far as possible should satisfy important positive goals (these positive goals may be also inconsistent). Two types of optimization can be performed, either in terms of a disjunctive combination of the prioritized positive goals, or in terms of a cardinal ity-based evaluation. The possibilistic logic approach enables us to compute a syntactic description of the set of preferred choices in each case, in agreement with the semantics, using a local computation method.