【作者】 刘硕；

【导师】 夏昊翔；

【作者基本信息】 大连理工大学 ， 管理科学与工程， 2023， 博士

【摘要】 近年来,观点双极化现象在世界范围内日益严重,已经成为威胁到社会和谐的重大因素之一。随着信息技术的快速发展,社交网络拓展了人们获取信息的渠道,但不少研究表明社交网络的使用与观点双极化现象密切相关。虽然大众更容易接触到多样的信息和观点,但事实上在社交媒体环境下反而更易于形成极端对立的观点集群,难以达成共识。为了揭示观点双极化现象产生的原因,本研究通过主体建模方法构建了双阈值观点动力学模型,从个体层面,人际关系层面,社区关系层面以及共演化层面,探究同时考虑同化性和排斥性社会影响条件下,群体观点双极化的产生条件及其影响因素。首先,本文聚焦于基于同化性和排斥性社会影响的观点动力学,提出了一个新的双阈值观点动力学模型。该模型通过构建的双阈值,即同化阈值和排斥阈值,来调节观点受到同化性和排斥性社会影响的范围。本文通过仿真实验对比分析了双阈值模型与经典HK模型的演化差异,发现本模型更能够准确地体现群体双极化的形成过程以及形成条件。同时,采用双极化程度指标来评估不同阈值条件下观点的分布和对立情况。研究结果表明,较强的同化性社会影响反而会促进观点双极化的形成。此外,本研究还探讨了在均匀混合群体中群体规模对观点演化的影响。其次,本文在双阈值观点动力学模型的基础上,进一步研究了个体层面的观点交互机制对复杂网络中观点演化的影响。在个体层面的观点交互机制中,个体的观点更新仅由其自身已有观点与邻居观点的差异所决定,不受其它因素的直接影响。因此,基于个体层面的观点交互机制,本文详细分析了代表性的WS小世界复杂网络结构中,当个体观点同时受到同化性和排斥性社会影响时,网络观点演化的规律和特征。同时,研究了网络结构和初始观点分布情况对网络观点演化达到稳态时的双极化程度的影响。研究发现,在WS小世界网络结构中,增加观点相近个体之间的接触可能导致网络观点双极化程度增强。相反,减少网络中个体之间的连边机会反而会抑制网络观点的趋同,从而促使网络观点呈现多元化。再次,为了探究个体之间连边关系异质性对网络观点双极化的影响,本研究应用符号网络并扩展了双阈值观点动力学模型,进一步提出了人际关系层面的观点交互机制,分析了同化性和排斥性社会影响对符号网络中观点演化的作用。在人际关系层面的观点交互机制中,个体观点更容易受到通过正连边连接的邻居产生的同化性社会影响,相反,更容易受到通过负连边连接的邻居产生的排斥性社会影响。研究发现,个体提高对正连边邻居的信任度,即扩大对自身的同化性社会影响范围,将会强化符号网络中的观点双极化程度。此外,符号网络的正连边比例与观点双极化之间存在着非线性关联。随着正连边比例的增加,观点双极化程度逐渐增强,但在正连边比例达到一定水平后,双极化现象会受到抑制。这一研究结果从人际关系层面为舆情防控领域提供了有益的见解。之后,基于前述研究,本文更深入地探讨了社区结构层面的观点交互机制对网络观点演化的影响。在社区结构层面的观点交互机制中,本研究基于异质性连边关系构建了符号社区网络结构,并结合了双阈值条件来判定同化性和排斥性社会影响的作用范围。本文分析了不同类型的符号社区网络模式中观点演化特征的差异,进一步研究了正负连边在网络中的位置对网络观点演化造成的影响。研究结果发现,当子社区内部仅包含正连边时,通过减少社区之间的负连边概率,可以有效地减少网络观点双极化的程度。而当存在负连边时,通过减少子社区内部的连接紧密度,同样可以降低网络观点双极化的水平。此外,子社区内部的负连边会增加网络分裂的风险,而初始观点的多样性可以有效降低观点双极化程度。这些发现强调了网络结构和连接模式对观点演化的重要影响,有助于更好地理解和管理社交网络中的观点传播。最后,考虑到实际情境中存在的观点与网络结构共同演化现象,本文基于符号网络构建了观点与网络结构共演化的动力学模型,并研究其演化规律,重点关注了动态符号网络中观点双极化程度的影响因素。在共演化层面的交互机制中,基于双阈值构建了观点交互模式,并结合结构平衡性理论提出了个体之间连边属性转换的条件。本研究进一步深入探究了在动态符号网络中考虑同化性和排斥性社会影响对观点演化的作用,以及观点演化对网络结构演化的影响。研究结果发现,在动态符号网络中,依然存在个体观点受到较强同化性社会影响会加剧网络观点双极化程度的现象。同时,相比于静态网络结构,观点共识所需的排斥阈值明显减小。此外,观点受同化性社会影响的范围越大,网络结构演化到稳态时存在的正连边比例越高,但在演化过程中受到观点演化的影响,正负连边比例会出现波动的情况。总体上,本文研究工作围绕观点双极化问题提出新的观点动力学模型,并重点结合符号网络情景开展研究,对观点双极化的形成和演化机理形成了一定新的认识;同时,本文研究结果为舆情的预测和控制具有一定的参考意义。更多还原

【Abstract】 In recent years,the phenomenon of opinion bi-polarization has become increasingly severe worldwide and it has become one of the major factors threatening social harmony.With the rapid development of information technology,social networks have expanded people’s access to information,but many studies have shown that it is closely related to the phenomenon of opinion bi-polarization.Although it is easier for the public to access diverse information and opinions,clusters of extremely opposing opinions are easily formed in the social media environment,making it difficult to reach a consensus.In order to reveal the causes of bipolarization,this study constructed a dual-threshold opinion dynamics model through the agentbased modeling approach to investigate the conditions and factors affecting the generation of bi-polarization of group opinions under the conditions of assimilative and repulsive social influences at the individual level,interpersonal relationship level,community relationship level,and co-evolutionary level.First,this study focuses on opinion dynamics based on assimilative and repulsive social influences.This study proposes a new dual-threshold model of opinion dynamics that moderates the extent to which opinions are subject to assimilative and repulsive social influences by means of constructed dual thresholds,namely,an assimilative threshold and a repulsive threshold.In order to simulate the scenarios of the influence of social media and Internet information on individual opinion updating,as well as to study in depth the combined effects of assimilative and repulsive social influences,this study employs a many-to-one interaction model with a simultaneous updating strategy for the study.The evolutionary differences between the dual-threshold model and the classical HK model are compared and analyzed through simulation experiments,and it is found that the present model more accurately reflects the formation process as well as the formation conditions of group bi-polarization.Meanwhile,this study also adopts the bi-polarization degree index to assess the distribution and opposition of opinions under different threshold conditions.The findings suggest that stronger assimilative social influence instead promotes the formation of opinion bi-polarization.In addition,this study explores the effect of group size on opinion evolution in homogeneous mixed groups.Second,based on the dual-threshold opinion dynamics model,this paper further investigates the influence of the individual-level opinion interaction mechanism on the evolution of opinions in complex networks.In the individual-level opinion interaction mechanism,an individual’s opinion update is only determined by the difference between his/her own existing opinions and the opinions of his/her neighbors,and is not directly affected by other factors.Therefore,based on the individual-level opinion interaction mechanism,this paper analyzes in detail the laws and characteristics of network opinion evolution in a representative WS small-world complex network structure when individual opinions are simultaneously affected by assimilative and repulsive social influences.Meanwhile,the effects of network structure and initial opinion distribution on the degree of bi-polarization when network opinion evolution reaches a steady state are investigated.It was found that in the WS small-world complex network structure,increasing the contact between individuals with similar opinions may lead to an enhanced degree of bi-polarization of network opinions.On the contrary,reducing the chance of connecting edges between individuals in the network inhibits the convergence of network opinion,but decreases the degree of bi-polarization,thus contributing to the diversification of network opinion.Third,in order to investigate the effect of heterogeneity of connecting edge relationships among individuals on the bi-polarization of network opinions,this study applies the signed network and extends the dual-threshold opinions dynamics model,and further proposes an interpersonal level opinion interaction mechanism,which analyzes the roles of assimilative and repulsive social influences on the evolution of opinions in the signed network.In the interpersonal level opinion interaction mechanism,individual opinions are more susceptible to assimilative social influences generated by neighbors connected through positive connecting edges,and conversely,more susceptible to repulsive social influences generated by neighbors connected through negative connecting edges.It was found that individuals who increase their trust in their positive edge neighbors,i.e.,expand the scope of assimilative social influence on themselves,will strengthen the degree of opinion bi-polarization in the signed network.In addition,there is a nonlinear association between the proportion of positive connecting edges and opinion bi-polarization in signed networks.As the proportion of positive edges increases,opinion bi-polarization gradually increases,but after the proportion of positive edges reaches a certain level,the bi-polarization phenomenon will be suppressed.The results of this study provide useful insights into the field of opinion prevention and control from the interpersonal level.Fourth,based on the previous study,this study explores in greater depth the impact of opinion interaction mechanisms at the level of community structure on the evolution of network opinions.Among the opinion interaction mechanisms at the community structure level,this study constructs a signed community network structure based on heterogeneous contiguous relations and incorporates the dual-threshold condition to determine the scope of the role of assimilative and repulsive social influences.This study analyzes the differences in opinion evolution characteristics in different types of signed community network models,and further investigates the effects caused by the position of positive and negative connecting edges in the network on the evolution of network opinions.It is found that when the sub-communities contain only positive connecting edges,the degree of network opinion bi-polarization can be effectively reduced by reducing the probability of negative connecting edges between the communities.And when there are negative connecting edges,the level of network opinion bipolarization can be similarly reduced by reducing the tightness of connections within subcommunities.In addition,negative connectivity edges within subcommunities increase the risk of network fragmentation,while initial opinion diversity can effectively reduce the level of opinion bi-polarization.These findings emphasize the important influence of network structure and connectivity patterns on opinion evolution,and contribute to a better understanding and management of opinion propagation in social networks.Finally,considering the phenomenon of co-evolution of opinion and network structure that exists in practical situations,this study constructs a dynamics model of co-evolution of opinion and network structure based on signed networks and studies its evolution law,focusing on the influencing factors of the degree of bi-polarization of the opinion in the dynamics of signed networks.In the interaction mechanism at the co-evolutionary level,the opinion interaction model is constructed based on the dual thresholds,and the conditions for the conversion of connecting edge attributes between individuals are proposed in conjunction with the theory of structural equilibrium.This study further delves into the role of considering assimilative and repulsive social influences on opinion evolution in dynamic signed networks,as well as the impact of opinion evolution on the evolution of network structure.The results of the study found that the phenomenon that stronger assimilative social influences on individual opinion exacerbate the degree of bi-polarization of network opinions still exists in dynamic signed networks.At the same time,the repulsion threshold required for opinion consensus is significantly reduced compared to the static network structure.In addition,the greater the extent of opinion exposure to assimilative social influences,the higher the proportion of positive contiguous edges that exist when the network structure evolves to a steady state,but the proportion of positive and negative contiguous edges fluctuates during the evolutionary process under the influence of opinion evolutions.Overall,the research work in this study proposes new opinion dynamics models around the issue of opinion bi-polarization and focuses on the signed network scenarios,forming a new understanding of the formation and evolution mechanism of opinion bi-polarization;at the same time,the results of this paper have certain reference significance for the prediction and control of public opinion.更多还原