Utility-Driven Minimum Cost Consensus Model With Preference Learning and Fairness Concerns

Nowadays the emergence of digital technology and artificial intelligence brings a lot of extra information and convenience for group decision activities, such as rich historical decision data, mobile decision platform, artificial interactive device and others. However, it also results in high complexity and uncertainty of the decision environments, modeling and analysis, and the preference of decision maker is crucial to the design of group decision and consensus mechanism. Meanwhile, the information transparency results in more attention being paid to fairness concerns in group decision making. Based on the historical group decision data, this study proposes a novel minimum cost consensus model with preference learning and fairness concerns (MCCM-PLFC), which characterizes the modification motivation of decision makers in a utility-driven way. A preference learning framework is first proposed to determine the parameters of utility functions of decision makers, where fairness concerns are modeled based on the return bias among the decision makers. Then, in a Stackelberg game consensus framework, the MCCM-PLFC is constructed to minimize the consensus cost of the moderator with consideration of the utility-driven behavior of decision makers. In addition, a best-response update algorithm is designed to solve the equilibrium solution among the decision makers and an adaptive differential evolution algorithm is used to solve the MCCM-PLFC. Finally, a throughout of experimental study is performed to quantify the validity of the proposed consensus model.