The choice of weights vectors in multiple attribute decision making (MADM) problems has generated an important literature, and a large number of methods have been proposed for this task. In some situations the decision maker (DM) may not be willing or able to provide exact values of the weights, but this difficulty can be avoided by allowing the DM to give some variability in the weights. In this paper we propose a model where the weights are not fixed, but can take any value from certain intervals, so the score of each alternative is the maximum value that the weighted mean can reach when the weights belong to those intervals. We provide a closed-form expression for the scores achieved by the alternatives so that they can be ranked them without solving the proposed model, and apply this new method to an MADM problem taken from the literature.
