Opinion Dynamics On Networks:Opin-Ion Diversity,Individual Activity And Group Selection

Opinion dynamics on networks is one of the dynamic processes on complex networks.It focuses on how opinions are formed,how opinions evolve among the population,how to reach consensus,and the time for evolutionary convergence.Research on opinion dynamics is based on tools from statistical physics,mathematical models,computer simulations and related analyt-ical methods.Opinion dynamics provide theoretical support for understanding human behavior,especially cultural dissemination,media competition,risk perception,etc.At the same time,it plays a significant role in building more secure information systems and understanding rumor cascades or science of fake news.Current works on opinion dynamics are mainly based on mod-eling abstract mechanisms.This work studies opinion dynamics on three aspects:opinion diversity,individual activity and group strategy selection.It broadens the explanatory ability of opinion modeling in spe-cific scenarios such as political opinion evolution,online social networks,as well as group com-petition.I assess the diversity of opinion distribution and opinion types,persistence of social networks,individual activity on opinion spreading,group rewiring strategy and other influences on the formation,dissemination,and competition of opinions.This work explores the theoreti-cal implications of opinion dynamics on practical issues such as business decision-making and community operations.The specific work of this paper is as follows:1.Propose a framework for analyzing political opinion evolutions.We consider the effects of initial opinion distribution and heterogeneous individual tolerance in to an opinion model;we catch the critical phases of fragmentation,polarization,and consensus during the evolution.Further,we propose a new measure of "average ratio of the nearest neighbors" as an index for evaluating results on stable stage.Our results show that the initial distribution of opinions leads to different evolutionary results even on similar random networks.The existence of stubborn nodes plays a special role:only extremely stubborn nodes can lead to changes in the final share of opinions;if the number of stubborn nodes is small,their influence is limited within a small range.2.Propose a novel way to characterize the types of opinions and the diversity in it.The model takes opinions covering single type or two types into consideration.The opinion type is linked to the spreader's interest,forming a differentiated forwarding.For single-type opinions,high-popular nodes generally have higher propagation capabilities than low-popular nodes.In the forwarding process,the peaks of forwarding shift to deeper propagation depths as the spreader's popularity decreases;When spreaders forward opinions that differ greatly from their own inter-ests,opinion dissemination capabilities are significantly reduced.Further,for two-type opinions,we consider the case where one opinion is consistent/inconsistent with the spreader's interest,and we consider the interaction between the types for two-type opinions.Our results show the non-linear interaction between the multiple types of opinions,as well as some special phenomena.3.Simulate individual activity,and introduce the active/inactive individuals into opinion models.We compare the changes of individual activities under three network growth modes:nodes join from high popularity to low popularity;nodes join from high similarity to low similar-ity,or uniformly random join.The results show that when nodes with high spontaneous activity,network generated under popularity-first growth mode could get the highest persistent activity;otherwise,under low spontaneous activity,similarity-first mode performs better.Besides,com-pound mode,by adding similar nodes with a few high popularity nodes within short times can achieve the highest network persistence.For further comparison and analysis,we consider the difference between individual activity and opinion propagation with/without considering indi-vidual activity,and we find that it cannot be ignored that impact of the difference in individual activity on opinion dynamics.Besides,the essence of difference in individual activity is the link types of the network,and the link types show the consistency on the role of network persistence and role of opinion dissemination.4.Propose a co-evolutionary model in which multiple rewiring mechanisms compete on the same layer of network topology.We introduce three typical rewiring rules:random rewiring(Rule ?),opinion-preferred rewiring(Rule ?),and degree-preferred rewiring(Rule ?)as op-tional strategies,embedded into an adaptive voter model.We compare the binary opinion com-petition under balanced/unbalanced initial proportion and find the superior rewiring rules under determined/indeterministic initial share or rewiring rate.We find all the evolution of opinion dynamics based on rewiring rules can be classified to two categories:the first category of evo-lutions satisfy the correlation between the density of opinions and density in distorted edges,in which the entire evolutionary process(including the final competition result)can be described by simulation and numerical approximation.Another kind of evolution does not possess the above-mentioned correlation,while the special types of open triads with opinions are good indicators,which indicate the final results well,in the early stage of evolution.The results show that the binary opinion competition under group rewiring rules can be effectively perceived in advance,and the homophily based rewiring rules have comparative advantages under most conditions.