Low-dimensional Visualization of Experts’ Preferences in Urgent Group Decision Making under Uncertainty

Abstract

Urgent or critical situations, such as natural disasters, often require that stakeholders make crucial decisions, by analyzing the information provided by a group of experts about the different elements of interest in these contexts, with the aid of decision support tools. This framework can be modeled as a Group Decision Making problem defined under uncertain environments, which are characterized by the imprecision and vagueness of information about the problem tackled. One of the main aspects to consider when solving Group Decision Making problems in urgent situations, is the fact that decisions should be taken under the highest level of possible agreement amongst all participating experts, in order to avoid serious mistakes or undesired responsibilities by some experts. Due to the complexity of urgent situations and the great amount of information about experts’ preferences that must be managed by stakeholders, it would be difficult to reach consensus in the group within a reasonable time period, since it requires an adequate analysis of experts’ preferences. Such an analysis might become a difficult task in these contexts, due to the inherent complexity, time pressure, etc. In order to help making consensual decisions in urgent situations and urgent computing environments, we propose a visual decision support tool for Group Decision Making problems defined under uncertainty. Such a tool provides stakeholders with a two-dimensional visual representation of experts’ preferences based on the similarities between them, and it enables the analysis of easily interpretable information about the state of the decision problem, as well as the detection of agreement/disagreement positions between experts. The tool is based on Self-Organizing Maps, an unsupervised learning technique aimed at the visual projection of information related to preferences of experts into a low-dimensional space.