A method for determining maximin OWA operator weights

Abstract

In this paper, we present a mathematical programming-based approach to determine the ordered weighted averaging (OWA) operator weights by maximizing the smallest difference between adjacent weights, referred to as the maximin OWA operator weights. Behavioral evidence suggests that decision-makers’ implicit weights, which influence their choices, are often quite steep. Thus, they tend to intuitively prefer alternatives that excel in several important criteria. If one alternative’s score is comparable to others, they might consider secondary and tertiary important criteria. The proposed maximin approach aligns more closely with this philosophy compared to previous methods that aim to evenly distribute operator weights.

Furthermore, we derive a closed-form solution for the maximin OWA operator weights using results from convex analysis. We also revisit the minimax disparity model, which is similar to our maximin approach, to emphasize the similarities and differences between the two methods.