群体共识快速收敛：建模、实验及仿真分析研究

Consensus convergence is the prerequisite for effective group decision-making (GDM). Under the constraints of limited time (e.g., online community, haze pollution reduction, etc.), the rapid convergence of the group consensus depends on the restrictions of limited cost, bounded rationality and bounded confidence, characterizing by the convergence evolution in system dynamics and the interactive convergence of optimization system. Based on intuitionistic preferences in GDM, the characteristics of consensus convergence are simulated and extracted, with the help of behavioral experiments and simulation technology; the optimization primal cost and dual compensation models are respectively constructed under the behavioral constraints of decision-makers (DMs), and the disturbance law of the consensus opinion triggered by the changes in the threshold of DMs’ opinions is analyzed; the dynamic GDM process is interpreted by the combination of the multi-level programming model and the traditional dynamic model, where the dynamic model is used to simulate the consensus interpretation and to show the features of the initiative convergence, while the multi-level programming model is applied to characterize the process details of the cooperative game carried by multiple decision-making groups and to reveal the traits of the balanced convergence. This project will design relevant algorithms and experimental programs to characterize and simulate the dynamic game relationship between the subsystem and the large group system, and build the optimization models of consensus convergence on the basis of the dynamic and multi-level programming models, so as to dynamically simulate the evolution of a large group consensus. This research provides the theoretical support for the consensus convergence in large complex GDM problems, develops new paradigm for the experimental simulation on the features of the dynamic consensus convergence, and provides available references for modeling the complex relationships among socio-economic systems based on system dynamic theory and optimization theory.

群体共识收敛是实现有效群体决策的前提。有限时间（网络社群、雾霾治理等）环境下的群体共识意见的快速收敛取决于有限成本、有限理性及有界信心约束，表征为系统动力学收敛演化、系统交互式收敛优化。基于直觉型群决策意见，采用行为实验、仿真技术模拟并抽取共识收敛特征；构建有决策者行为约束的原始成本-对偶补偿优化模型，预变分析决策者意见参数域值变动触发共识意见扰动规律；分别采用多层规划模型结合传统的动力学模型演绎动态群体共识决策，动力学模型模拟共识演绎及主动性收敛特征，多层规划模型表征决策群的合作博弈过程及均衡收敛特征。形成表征、模拟子系统与大群体系统动态博弈关系的算法设计及行为实验方案，构建动力学-多层优化共识收敛模型演绎体系，动态模拟大群体共识演化过程。研究为复杂大群体共识收敛决策提供理论支持，为实验仿真模拟动态共识收敛特征开拓新研究范式，为动力学-系统优化融合建模演绎社会经济系统复杂关系提供借鉴。

研究背景：在社会经济学领域，群体共识收敛是实现有效群体决策的前提。现有的多智体动力学、舆情动力学模型及共识收敛算法很难系统模拟与仿真有约束条件的群体共识决策问题。有限时间环境下的群体共识意见的快速收敛取决于有限成本、有限理性及有界信心约束，表征为系统交互式收敛优化。.研究内容：研究系统构建了系列理论模型，包括不确定偏好关系的一致性与一致性建模、不完全信息不确定偏好关系的加、乘一致性建模及其权重获取、不确定约束直觉模糊偏好关系最优排序建模；基于随机意见的最小成本共识建模、考虑带有效用函数的直觉模糊偏好关系最优排序建模及基于多阶段波动效用约束的群体共识测度建模。探讨了共识收敛模型在不同环境中的应用，包括不确定性系统的社交网络信任度测量；灾害链的风险阈值及信任度的解析以及微博大数据的城市暴雨内涝灾情的实时快速判别研究等。.重要结果：研究分别从行为科学、社会物理学、优化与控制等多学科交叉角度，采用系统优化方法结合随机、不确定系统建模，演绎群体决策动态共识过程；考虑行为变量如效用函数拟合共识收敛特征：构建不同偏好、不同运动、不同通讯模式下群体共识收敛规则；用多层优化模型表征决策群体之间的合作博弈过程及均衡收敛特征；探析群体意见偏好的知识表达模式、群体意见系统的运动方式、群体通讯模式、协调者的沟通能力及决策个体的有界信心及视野等因素变动触发群体共识演化的路径、模式及结果。.关键数据：共发表学术论文30篇，在科学出版社出版专著一部；获得4项省厅级奖励；主办或者参加各类学术会议近20次，其中举办国际学术会议两次。共培养博士后1人，博士研究生4人，硕士研究生14人。有一名学生获得2019江苏省优秀硕士论文。.科学意义：通过抽象、简化模型、重构算法、行为实验等模式，初步重现和预测群体共识意见演进过程，进一步揭示群体共识快速演化的本质特征，为系统优化融合建模演绎社会经济系统复杂关系提供了借鉴。