Effects Of Heterogeneous Individuals On The Dynamics Of Social Contagions

The development of online social media has radically changed the way that individuals consume information and form opinions.The public has moved from the era of mass media communication to that of interpersonal communication where they can create contents,share information and interact with others.In the social contagions of interpersonal communication,each individual can be an information source and affect others' opinions and decisions.It is worth noting that the influence of individuals on others are different.How to identify the heterogeneous social influence and quantify its roles in information diffusion,sentiment contagion and behavior spreading has attracted much interest of the scholars from economics,sociology,physics,finance and management.Answers to these questions can help develop relevant intervention strategies(such as containing rumors,protests,etc.)or make better product marketing.This dissertation will study three parts:empirical data analysis,dynamical model construction and theoretical analysis,to study the effect of heterogeneous influence of individuals on the social contagions and aims to better reveal the evolution mechanisms of real social contagion phenomena.The first part of this dissertation studies the diffusion process of news online and identifies the influential users from all participants.Moreover,we find different spreading patterns among heterogeneous individuals and analyze the relevant diffusion network structures.Firstly,we divide all users into different communities according to their follower counts and use symbolic transfer entropy to compute the information flow between different communities(i.e.,the influence of users in a community on that in another community).The result shows that the information flow is asymmetric between different communities and the users whose follower counts are not less than 105 drive those diffusion network neighbors who have lower follower counts.Thus,the users with follower counts greater than or equal to 105 are regarded as influential users(High follower count users,H-users),and the others are ordinary users(Low follower count users,L-users)in this dissertation.After dividing the users into two categories,we find that L-users play dominant roles in some cases which is different from the common situations that the H-users are better at spreading information.Next,this dissertation gets the structural characteristics of diffusion networks and the distribution of distances among all connected nodes in different situations dominated by H-or L-users.We find that the diffusion network is central aggregate when H-users dominate and the distance distribution is Gaussian.When L-users dominate,the irregular long-range paths are observed in the diffusion network and the distance distribution is non-Gaussian which arises mainly from the long paths.The second part of this dissertation constructs a social contagion model based on the online news diffusion.We reproduce the spreading patterns and diffusion network structures of online news and reveal the generation mechanism of different diffusion network structures.We develop a social contagion model with heterogeneous influence strength.By simulating the model,we get the spreading patterns,diffusion network structures and the distance distribution of real news.We find that,when the influence of L-users is low,the news diffusion is dominated by simple contagion and most news are forwarded from the H-users.The diffusion network is central aggregate and the distance distribution is Gaussian.Nevertheless,when the message is controversial or popular,the opinions of ordinary users are valued and their influence increases.In this case,the news diffusion is complex and the diffusion network structure presents long-range irregular paths dominated by ordinary users' interaction.Thus,the distance distribution of diffusion network is non-Gaussian.In addition,we propose an index,i.e.,the average number of forwards from ordinary users,and the two dominant patterns can be distinguished by the index.Moreover,by reproducing the two dominant patterns,we reveal when the L-users can dominate the news diffusion.The third part of this dissertation analyzes the social contagions with heterogeneous influence strength theoretically and distinguishes the roles of heterogeneous individuals.Moreover,we explore how the heterogeneous distribution of influence affects the outbreak threshold and phase transition of spreading size.We analyze the spreading process with heterogeneous influence strength by developing an edged-based compartmental theory.Meanwhile,based on the fact that the ways to become adopted for a susceptible individual can be divided into three categories:only through influential individuals,only through ordinary users and through both together,we utilize the permutation and combination theory to examine how heterogeneous individuals exert their influence.We find first-order,second-order and hybrid transitions in the system and reveal how heterogeneous individuals lead to different phase transitions.In addition,crossover phase transitions exist when varying the proportion of influential individuals:it changes from being first-order to second-order phase transition when the influence of ordinary individuals is high;interestingly,when the influence of ordinary individuals is relatively low,the changes from first-order to hybrid,and then to second-order phase transition have been observed.The similar crossover phase transition phenomena exist in the system when increasing the difference of influence strength between ordinary individuals and influential individuals.In conclusion,this dissertation systematically analyzes how heterogeneous influence affects spreading patterns,diffusion network structures and phase transitions,and lays the foundation for studying how heterogeneous individuals affect social contagions.