决策要素动态变化的群体决策偏好演化过程

研究交互式动态群体决策偏好收敛自适应过程的内在机制和规律，发现和控制影响群体一致度和偏好收敛时间的关键因素，以改善群体决策时间绩效. 应用计算实验方法，基于Deffuant有界自信假设的妥协策略，在微观层面上构建偏好演化动力 学模型，模型符合群体决策共识过程原则要求；在模型基础上观察分析个体属性特征对一致性绩效、收敛时间绩效的作用. 论文在决策成员社会惰性、空间位置、决策偏好动态变化的决策情景中，研究决策者社会惰性对群体决策偏好收敛时间的影 响. 建立了动态群体决策的概念框架，分别针对惰性不变的偏好演化模型和惰性变化的偏好演化模型设计了两个计算实验. 实验1：比较不同社会惰性水平p^*和不同选择半径r组合下，决策偏好时间T_c的变化. 实验2：比较不同惰性 递增速率v和不同惰性递增上限ps组合下，决策偏好时间T_c的变化. 研究发现：1）社会惰性越大，对 偏好收敛的阻滞效应越显著；2） r越小，T_c随着p*递增的趋势更强；3）当扩大决策成员交互对象的 选择范围，T_c的波动减小，群体达成共识的时间更稳定；4）当社会惰性值随休眠持续期在一定范围内线性 递增时，T_c递增存在一个分界值.

The preference evolution process of group decision-making with dynamic decision factors

In this paper, we study the intrinsic mechanisms of the adaptive process of preference convergence in interactive dynamic group decision-making, and improve group decision making time performance by identifying and controlling key factors that influence group consistency and preference convergence time. With computational experiments approach and compromise strategies based on Deffuant bounded confidence assumption, a preference evolution dynamics model has been built on the micro-level, which conforms to the principles of group decision consensus process. The model facilitates the analysis of how individual characteristics affect consistency performance and convergence time performance. We consider the impact of decision members' social loafing behavior on the preference convergence time, in the context of spatial positioning and decision preference evolution. With a conceptual framework of dynamic group decision making, two computational experiments are designed for constant and varying social loafing, respectively. Experiment 1: Comparing decision preference time T_c under different combinations of social loafing level p* and choice radius r. Experiment 2: Comparing decision preference time T_c under different combinations of incremental rate ν and increment upper bound ps of social loafing level. Research findings: 1) The higher the social loafing level the more significant the delaying effect on preference convergence; 2) The smaller the r, the stronger the increasing trend of T_c with p^*L ; 3) As the scope of interactive objects of decision members expands, the fluctuation of T_c reduces, and the time to reach consensus stabilizes; 4) If social loafing level increases linearly with dormant latency within a certain domain, there exists a utmost value for the increment of T_c.