When can we have confidence in distillation results of ELECTRE III? or: Has France the wrong president

Abstract

The family of ELECTRE methods have been well established in multi-criteria decision aiding since long. However, in multi-criteria rankings the derivation of a final result is sometimes tricky and even relatively simple requirements may not be met, as e.g. Roy and Bouyssou (1993: 356ff, 362) emphasize. Therefore, to gain confidence in any proposed method it seems a necessary condition that it (re)produces results, that have been established otherwise and that are very plausible if not generally accepted, in a transparent and equally plausible way. Here we make this test by applying the ELECTRE III distillation process to a fictional election campaign in order to derive a ranking of different candidates, and we obtain counter-intuitive results.