The proposed procedure is intended to solve the following problem. Given are \(L\) possible decisions (options), each described by a large number of criteria. Experts compare the options by different groups of criteria and are unable to allow for all the criteria simultaneously. Use partial criteria to estimate the overall (by all criteria) utility of the available options and rank them in the order of increasing utility. Typical applications that lead to this problem include evaluation of alternative socio-economic programs, alternative plans for solving complex political problems, or alternative technical designs. For instance, plans for socio-economic development must be evaluated allowing for economic, technical, socio-political, psychological, and other groups of criteria, each of which has to be analyzed by experts with a certain level of knowledge and qualifications. The need for aggregating diverse expert judgments is the main difficulty when solving such problems. Available aggregation methods are based on hierarchical representation of the system of criteria as a tree structure, for instance the AHP method and Promethee/Electre. However, the experience with their use for practical problems, while confirming their obvious advantages, has revealed substantial weaknesses. First, the methods suffer from an insufficient theoretical foundation, especially the AHP method. The absence of a rigorous theoretical foundation creates serious difficulties when we try to explain certain properties of the AHP method, such as the rank reversal property, and complicates the interpretation of decisions produced by this method. Second, these methods assume that all the criteria are applied to estimate each option. Yet in practical problems we may use alternative groups of criteria to evaluate different decisions, e.g., under incomplete information or when different systems of statistical indicators are available. More complex logical interdependences between groups of criteria are also possible. We make an attempt to eliminate these shortcomings. The theoretical foundation of our study is provided by the theory of fuzzy sets. From the point of view of this theory, the original problem is interpreted as determining the degree of membership of each of \(L\) options in the fuzzy set of ``good'' decisions. This set is constructed by set-theoretical operations from the set of options that are passed by different groups or subgroups of criteria. The procedure includes two components: 1) hierarchical representation of criteria and the corresponding fuzzy decision sets; 2) construction of membership functions of the original sets. These components are independent of each other in the sense that a given hierarchical structure of criteria may operate with arbitrarily constructed membership functions of the original sets, while the procedure constructing these membership functions may be applied to solve other problems, too.