As an important extension of decision-theoretic rough sets, three-way decision theory provides a new perspective for people to deal with uncertain problems. However, the traditional multi-granularity decision-theoretic rough sets model has limited ability in describing the risk preferences of decision-makers and the processing of intuitionistic fuzzy information. In addition, as far as we know, most of the risk loss functions in existing studies are based on utility theory. However, the complete compensability between attributes is not always true, and this fact may lead to inconsistencies between the final calculated results and the actual situation. We propose a multi-granular intuitionistic fuzzy three-way decision model based on the risk preference outranking relation. In this scenario, we first define the outranking relation on the intuitionistic fuzzy set and fuse it for the purpose of risk preference calculation. Next, starting from the single granularity, the relations between the membership outranking relation class, the nonmembership outranking relation class, and the rough approximation are analyzed, and the related properties are proven. Then, the single granularity is extended to construct the multi-granular intuitionistic fuzzy decision-theoretic rough sets and their corresponding three-way decision model. Furthermore, by systematically studying the decision loss costs of optimistic and pessimistic states, three-way decision rules are induced. The rationality and effectiveness of our proposed model are verified through a case study analysis and comparisons with existing methods. The results show that our proposed model can quantitatively analyze and calculate the uncertainty of decision-makers’ cognitive risk preferences, achieve global control of the decision-making process, and reduce the loss of decision-making costs.

三路决策理论作为决策理论粗糙集的重要延伸，为人们处理不确定性问题提供了新的视角。然而，传统的多粒度决策理论粗糙集模型在描述决策者的风险偏好和处理直觉模糊信息方面的能力有限。此外，据我们所知，现有研究中的大部分风险损失函数都是基于效用理论的。然而，属性之间的完全可补偿性并不总是正确的，这一事实可能导致最终计算结果与实际情况不一致。我们提出了一种基于风险偏好优先关系的多粒度直觉模糊三向决策模型。在这种情况下，我们首先在直觉模糊集上定义优级关系并将其融合以用于风险偏好计算。接下来，从单粒度出发，分析隶属优等关系类、非隶属优等关系类与粗略近似之间的关系，并证明相关性质。然后将单粒度扩展为多粒度直觉模糊决策理论粗糙集及其对应的三向决策模型。此外，通过系统研究乐观和悲观状态的决策损失成本，推导出三向决策规则。通过案例研究分析和与现有方法的比较，验证了我们提出的模型的合理性和有效性。