A signed network analysis-based Consensus Reaching Process in Group Decision Making

In the previous Consensus Reaching Process (CRP) models, although it has been a common understanding that positive social relationships between experts can prompt cooperation, meanwhile the existence of negative relations and the consequent antagonistic behaviors have been largely overlooked. In an attempt to fill in that gap, in this paper a signed network analysis-based CRP is proposed, where both the positive and negative relations can be taken into full account. In our proposed model, two types of centrality are taken advantage of to derive the experts' importance degrees from their signed social network. Such importance degrees are then incorporated into the aggregation process via an Importance-Induced Ordered Weighting Average operator, where experts with higher importance degrees are assigned more weights to reflect their credibility. Consensus indices are defined to measure each expert's consensus level to the group. And to help those experts with insufficient consensus indices, a feedback mechanism is developed where personalized preference revision advices are generated based on two rules: first, experts are attracted to the opinions of their positive social connections; and second, experts are repelled by the opinions of their negative social connections. An illustrative example is implemented to demonstrate the practicality of our model, and to investigate possible factors that may affect the CRP outcome. Finally, comparative analyses to several existing social network-based CRP methods are conducted. (c) 2020 Elsevier B.V. All rights reserved.