A Bayesian Framework For Modelling The Trust Relationships To Group Decision Making Problems

In traditional social network analysis-based group decision-making problems, decision makers (DMs) commonly establish trust relationships with unfamiliar DMs based on the transitivity of trust but always ignore the effectiveness of such indirect trust and the potential risks that may arise from excessive trust/distrust. To overcome this limitation, this paper explores a Bayesian framework to construct trust relationships among a group which can update the direct trusts between DMs by considering the trusts to the trustee from other DMs trusted by the original trustor. The Bayesian trust relationship can not only supplement the incomplete trust relationships based on known trust information, but also make reasonable adjustments to existing trusts before the decision process. And it has been proven to be effective in mitigating inaccuracies in decision outcomes resulting from either excessive trust or distrust, and in reducing information loss caused by excessively long paths during trust transition. Relatedly, the algorithm of the Bayesian trust relationship is first proposed inspired by Bayesian analysis. Second, to measure and constrain the trust decision indicator (TDI) in the Bayesian framework, the rank discrepancy between DMs is defined in belief distribution environment where the uncertainties present in distributed evaluations may lead to diversity in ranking orders. Third, constrained by the expected rank discrepancy, the stochastic TDI analysis is introduced to derive the confidence weight intervals of DMs. Finally, an optimization model is provided to maximize the group consensus within the constraints of the generated interval weights. To demonstrate the practical application of the proposed model, a case study on the selection of new energy vehicles is presented. Additionally, both comparative study and sensitivity analysis are provided.