A theorem on aggregating classifications


Mathematical Social Sciences

janvier 2016, vol. 79, pp.6-10

Départements : Economie et Sciences de la décision, GREGHEC (CNRS)

Mots clés : Aggregation of classifications, Group identification problem, Task assignment problem, Nonbinary evaluations

Suppose that a group of individuals must classify objects into three or more categories, and does so by aggregating the individual classifications. We show that if the classifications, both individual and collective, are required to put at least one object in each category, then no aggregation rule can satisfy a unanimity and an independence condition without being dictatorial. This impossibility theorem extends a result that Kasher and Rubinstein (1997) proved for two categories and complements another that Dokow and Holzman (2010) obtained for three or more categories under the condition that classifications put at most one object in each category. The paper discusses an interpretation of its result both in terms of Kasher and Rubinstein’s group identification problem and in terms of Dokow and Holzman’s task assignment problem.