Using qualitative information elicited from a panel to obtain robust conclusions: a protocol and an application to improve integrated pest management systems

Multicriteria aggregation methods typically require inputs from decision makers concerning the relative importance of the criteria. This work presents an approach to use qualitative information elicited from a panel, which can be applied to compensatory and non-compensatory multicriteria aggregation methods. In particular, it considers the additive multiattribute value function and ELECTRE, two classical methods with well-known differences in the meaning of the criteria weights. Moreover, the proposed protocol makes a distinction between the importance of improving the current situation and the importance of not worsening the current situation. The inputs from the panel are aggregated to define constraints on the importance-related parameters, which can then be used for robustness and stochastic analyses. As a real-world application, a comparison of Integrated Pest Management (IPM) is performed, considering the case of carrots cultivation in a French region. The comparisons are based on a sustainability assessment of the current practices and alternative IPM systems, using data from field trials, laboratory experiments and preferences from stakeholders. Results are robust to weighting choices, thus identifying which changes are recommended.