An Active Opinion Dynamics Model

Opinion dynamics models are a class of models that utilize the tools of physics,mathematics,and computer science to study the evolution of group opinion.Because of their advantages in modelling interaction and development of opinions,opinion dynamics models have been successfully applied to group decision-making,pricing,marketing,and new product launch.However,with the development of information technology,the conditions and methods of interpersonal communication have gradually changed,reducing the applicability of traditional opinion dynamics models that rely on passive interactions initiated by locational or social relationships.Now,people can interact with anyone at any time and places,and are often driven by their specific inner causes.Therefore,how to model active interactions becomes the key to understanding group opinion evolution in the modern social environment.Based on concluding the development of opinion dynamics research,this thesis proposes an active opinion dynamics model from the perspectives of the activeness and dynamics of interaction and the experience and difference of interaction participants.In this model,according to the similarity effect,people tend to communicate with someone who is more likely to be an ally;using the evolutionary game,people assess possible gains and losses before interactions and decide whether to interact referring to historical experience;on the basis of Bayesian rule,people dynamically update their opinions with the information they get from interactions;back and forth,the evolution of opinions is modelled.Multi-agent experiments of the model simulate the process at both the micro and macro levels and show the ubiquitous gaps between people's voting result and their collective opinion,indicating that the voting result does not necessarily represent group opinion.Then by analyzing the model and simulation results,this thesis points out that the evolution of group opinion is impacted by the group's initial collective opinion and inclusiveness of different opinions,where the former predetermines whether the group opinion will reverse its first stance and the latter affects the consistency and intensity of group opinion evolution.Further,the thesis finds and attests to that the extent of group opinion change obeys a bimodal distribution whose parameters are related to initial group opinion and the inclusiveness of different opinions,which suggests two mutually restricted tendencies of stabilization and destabilization and contributes to predicting how much the collective opinion of a group will change after full discussion.The active opinion dynamics model proposed in the thesis takes full account of the realistic foundations and fills the gap of active interaction research in the current field of opinion dynamics.At the same time,through application cases,the thesis illustrates how the new model can help to understand,predict,and adjust group opinion evolution under active interactions,which has practical significance for group decision-making and public opinion control.