A Progressive Search for a Group Ranking with Robust Conclusions on Prudent Orders

We consider the problem where rankings, provided for instance by a group of evaluators, have to be combined into a common group ranking. In such a context, Arrow and Raynaud suggested that the compromise ranking should be a prudent order. In general, a prudent order is not unique. That is why, we propose to manage this possible multiplicity of compromise solutions by computing robust conclusions. This allows for a progressive refinement of the decision model and supports the group to eventually select one group ranking. The approach is illustrated on a problem where a group of junior researchers has to agree on a ranking of research domains.