Group decision-making processes can be supported by group recommender systems that help groups of users obtain satisfying decision outcomes. These systems integrate a consensus-achieving process, allowing group members to discuss with each other on the potential items, adapt their opinions accordingly, and achieve an agreement on a selected item. Such a process, therefore, helps to generate group recommendations with a high satisfaction level of group members. Our article provides a rigorous review of the existing consensus approaches to group decision-making. These approaches are classified depending on the applied consensus models such as reference domain where a set of group members or items is selected for calculating consensus measures, coincidence method that calculates the consensus degree between group members depending on the coincidence concept, operators that aggregate user preferences, guidance measures where the consensus-achieving process is guided by different consensus measures, and recommendation generation and individual centrality that enhance the role of a moderator or a leader in the consensus-achieving process. Further consensus techniques for group decision-making in heterogeneous and large-scale groups are also discussed in this article. Besides, to provide an overall landscape of consensus approaches, we also discuss new consensus models in group recommender systems. These models attempt to improve basic aggregation strategies, further consider social relationship interactions, and provide group members with intuitive descriptions regarding the current consensus state of the group. Finally, we point out challenges and discuss open topics for future work.

群体推荐系统可以支持群体决策过程，帮助用户群体获得满意的决策结果。这些系统集成了一个达成共识的过程，允许小组成员就潜在的项目相互讨论，相应地调整他们的意见，并就选定的项目达成一致。因此，这样的过程有助于生成群组成员高满意度的群组推荐。我们的文章对现有的群体决策共识方法进行了严格的审查。这些方法根据所应用的共识模型进行分类，例如选择一组组成员或项目来计算共识度量的参考域、重合方法根据重合概念计算群体成员之间的共识程度，聚合用户偏好的算子，由不同共识度量指导达成共识过程的指导度量，以及推荐生成和个人中心性加强调解人或领导者在达成共识的过程中的作用。本文还讨论了异构和大规模群体中群体决策的进一步共识技术。此外，为了提供共识方法的总体情况，我们还讨论了群体推荐系统中的新共识模型。这些模型试图改进基本聚合策略，进一步考虑社会关系交互，并为群体成员提供关于群体当前共识状态的直观描述。最后，我们指出挑战并讨论未来工作的开放主题。