Fund Project:Project supported by the National Natural Science Foundation of China (Grant Nos. 71731002, 61573065) and the National Key Research and Development Plan, China (Grant No. 2017YFC0804000).
Received Date:23 December 2018
Accepted Date:20 March 2019
Available Online:01 June 2019
Published Online:05 June 2019
Abstract:Entering the information era, the formation of public opinion is largely associated with the complex system constructed by the Internet, thereby possessing new characteristics. The formation of public opinion is the result of the interaction of individual behavior with social environment. In reality, the environmental factor and the individual behavior are usually related to each other and co-evolve with time. Based on the Ising model, in this paper established is an opinion formation model that includes the process of the accumulation and digestion of the social tension. In the model, a parameter named effective dissolving factor c is designed to represent the extent of the interaction between the system and the social environment. A two-dimensional dynamical system is involved in the model to describe the dynamics of individual behavior and social tension. The co-evolution behavior of the system is studied. Based on the Landau mean field theory, the stationary states of the dynamical system under different parameter values, i.e. the value of effective dissolving factor c, their stability and bifurcation of the system, are analyzed. Finally, the computer simulation method is used to verify the results. The research shows that with the co-evolution mechanism of the system, our model exhibits certain self-organization characteristics. When the effective dissolving factor c is smaller than the threshold value, the system will reach final consensus opinion, resulting in a macroscopically ordered state. Otherwise, when the dissolving factor c exceeds a threshold value, the system is stable in the disordered state. It is interesting to find that there is such a critical value of the parameter that it leads the system to be self-organized into a critical state from any initial state. The future detailed investigation on the criticality of the co-evolving system is also suggested, such as testing whether the system has evolved into the critical state according to the finite-sized scaling theory and calculating the critical exponent of the system. In addition, in this paper provided is a new perspective to tackle practical problems in public opinion. Based on the mechanism of the formation of public opinion revealed by our model, researchers are encouraged to conduct studies on how to monitor the state of public opinion more precisely and to predict the tipping point of the system evolution. Keywords:opinion formation/ Ising model/ co-evolution/ self-organized criticality
全文HTML
--> --> --> -->
2.1.耦合演化的舆论生成模型
在模型建立和讨论部分, 为了方便起见, 仍然选取温度T作为模型参数和变量. 由朗道1937年提出的平均场理论, Ising模型在不同温度下的热力学势如图1所示. 图 1 不同温度下的热力学势 Figure1. The Landau potential under different temperatures.
1) c = 2.5 如图4所示, 系统温度随时间演化不断上升, 最终达到定态; 系统绝对平均磁矩逐渐减小到0, 并一直保持在0附近. 图 4c = 2.5时系统状态演化行为 Figure4. Evolution of the system state given c = 2.5.
取系统演化到定态时最后500个Metropolis步的平均磁矩, 并在不同初始条件下系统演化重复6次, 得到3000个平均磁矩值进行系综平均, 得到的统计分布如图5. 图 5c = 2.5时系统定态时磁矩M的统计分布 Figure5. Distribution of the magnetic moments (M) after system evolved to the stationary state, given c = 2.5.
由图5看出, 系统磁矩分布于区间(–0.04, 0.04), 经计算, 该分布的峰态系数K ≈ 3, 偏态系数S ≈ 0, 故该分布接近均值为0的正态分布, 说明此时系统演化至了无序态, 与平均场理论中的结果一致. 2) c = 1.3 如图6所示, 系统温度和绝对平均磁矩随着时间演化都始终在一定区间内波动. 图 6c = 1.3时系统状态演化行为 Figure6. The evolution of system state given c = 1.3.
同样地, 取系统演化到定态时最后500个Metropolis步的平均磁矩, 并使系统在定态条件下重复演化16次, 得到8000个平均磁矩值进行系综平均, 得到统计分布如图7所示. 图 7c = 1.3时系统定态时磁矩M的统计分布 Figure7. Distribution of the magnetic moments (M) after system evolved to the stationary state, given c = 1.3.
此时, 系统的平均磁矩在(–0.4, 0.6)的区间内, 分布范围远远超过c = 2.5时的情形, 且峰态系数K ≈ 2.5, 小于正态分布的情形, 表明系统在均值0附近有较大的涨落. 由于随机性的存在, 仅仅通过对给定参数下系统演化及稳态分布的观察, 很难准确找到平均场理论中得到的临界点参数. 在后续的研究中, 我们将利用有限尺度标度理论及统计方法, 确定耦合演化系统的临界有效疏解系数c, 并定性地确定系统是否具有临界性质, 进而定量刻画系统的临界行为, 包括标度律和临界指数. 3) c = 0.8 如图8所示, 温度随时间演化不断降低, 最终达到定态; 系统平均磁矩从0上升, 表现出一定程度的极化行为. 图 8c = 0.8时系统状态演化行为 Figure8. The evolution of system state given c = 0.8.
为验证系统降温过程中出现的与前文所述两种情形所对应的暂态, 我们观察了系统磁矩分布随时间的变化. 我们使系统重复演化9次, 在每一次演化中记录每一次变温演化时500个Metropolis步的平均磁矩, 每一个时间步得到4500个平均磁矩进行系综平均, 绘制出系统的平均磁矩分布随时间的演化图像, 如图9所示. 图 9c = 0.8时系统磁矩M的统计分布随时间的变化 (a) t = 100; (b) t = 300; (c) t = 500; (d) t = 600—700 Figure9. Distribution of the magnetic moments (M) when the system is evolving to the stationary state, given c = 0.8: (a) t = 100; (b) t = 300; (c) t = 500; (d) t = 600?700.