This article proposes a bidirectional feedback mechanism for consensus in group decision making (GDM) driven by the behavior of decision makers (DMs), which is discriminated with a flexible harmony degree as one of three possible states: (1) ‘tolerance behavior’; (2) ‘rationalist behavior’; and (3) ‘conflict behavior’. The first two states are possible to be resolved in the consensus reaching process with one round of feedback recommendations to the discordant DMs. However, in the conflict state, which implies the lack of harmony between the group aim of ‘consensus’ and the individual benefit, it is unreasonable to be resolved with only discordant DMs’ feedback recommendations, and concordant DMs are also expected to make concessions at some degree. To address this not so unusual research problem, a theoretical bidirectional feedback mechanism framework for consensus is developed. Firstly, a maximum consensus driven feedback model is proposed to resolve ‘conflict behavior’ between the concordant and discordant DMs. Secondly, a maximum harmony driven feedback model is activated to support the discordant DMs to reach the threshold values of group consensus. A numerical example is provided to illustrate and verify the proposed mechanism usefulness and how it compares against other existent feedback mechanisms in terms of the extent up to which DMs’ preferences are changed for reaching consensus.

本文提出了一种由决策者（DMs）行为驱动的群体决策（GDM）共识双向反馈机制，以灵活的和谐度将其区分为三种可能状态之一：（1）“容忍行为”；(2)'理性主义行为'；(3) '冲突行为'。前两种状态有可能在达成共识的过程中解决，并向不一致的 DM 提供一轮反馈建议。然而，在冲突状态下，意味着“共识”的群体目标与个人利益之间缺乏协调，仅用不一致的DMs反馈建议来解决是不合理的，而且一致的DMs也会在以下情况下做出让步一定程度上。为了解决这个不寻常的研究问题，开发了一个理论双向反馈机制框架来达成共识。首先，提出了一种最大共识驱动的反馈模型来解决“冲突行为”。' 在一致和不一致的 DM 之间。其次，激活最大和谐驱动反馈模型，以支持不一致的 DM 达到群体共识的阈值。提供了一个数值示例来说明和验证所提出的机制的有效性，以及它如何与其他现有反馈机制在 DM 的偏好改变程度以达成共识方面进行比较。