Pour mieux répondre à la préoccupation de robustesse en aide à la décision: quatre propositions suite à un double constat

Bernard Roy (1)

1 LAMSADE - Laboratoire d'analyse et modélisation de systèmes pour l'aide à la décision

After reviewing what the adjective "robust" means in decision aiding (DA) and explaining why it is important to be concerned about robustness in DA, I present a twofold observation (section 2) which leads me to make four proposals in order to better respond to robustness concern in decision aiding. With the first two proposal (sections 3 and 4), I show that, in many cases, the vague approximations and the zones of ignorance against which robustness helps to prevent, must be considered in terms of substituting the concept of version for the usual concept of scenario and focusing on the diverse processing procedures that must be used to study the decision aiding problem as it was formulated. Next, I show (section 5) that the traditional responses formulated in terms of "robust solutions" limit the meaning of this concept (section 5). I briefly describe a certain number of avenues for research that could be explored further, not only in order to conceive the solutions that could be qualified as robust in another way, but also to better interact with decision makers to make them aware that the adjective "robust" can be subjective. Finally, the fourth proposal is related to forms of responses which lead to state "robust conclusions", which do not necessarily refer to solutions characterized as robust. After defining what I mean by robust conclusions and giving some examples, I mention the rare approaches that have been proposed for obtaining such conclusions.

Domaines: Recherche opérationnelle [math.OC]

Bernard Roy. Pour mieux répondre à la préoccupation de robustesse en aide à la décision: quatre propositions suite à un double constat. 2009. ⟨hal-00877042⟩