Opinion formation in social networks: a simulation study

• Main Author: Liu, Yituo
• Other Authors: Xiao Gaoxi
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Social network; Opinion formation; Extremism; Deffuant model
• Online Access: https://hdl.handle.net/10356/174226
• Institution: Nanyang Technological University

Extremism commonly exists in social networks. Studying the dynamics and evolution of extremism and non-extremism opinions in social networks is of significant importance. Extremists are paranoid. It is typically not easy for them to accept other people's opinions. In this paper, we extend the well-known Deffuant model to resemble the real-life observations that extremists tend to be less tolerant of different opinions. We also evaluate the impacts such differences in tolerance may make on system opinion evolution. Further, to better understand the effects of different people's different willingness to compromise in order to achieve consensus-making, we introduce two communication logics, named unidirectional consensus-making and bi-directionality consensus-making, where opinion change for consensus-making can or cannot happen on single-side, respectively. It is found that bi-directionality consensus-making leads to a more stable formation of the opinion communities in the final state, while unidirectional consensus-making, typically sourced from good willingness to achieve consensus as it is, may actually lead to less stable results and give extremism opinions a better chance to prevail. Simulations have been carried out on different cases with different tolerance ranges, and some interesting observations have been made. We also conducted simulations for scenarios with both discrete and continuous tolerance ranges, proposing the Continuous Distributed Randomly Tolerance (CDRT) and Discrete Distributed Randomly Tolerance (DDRT) models. These models assess the impact of different tolerance range distributions on network evolution. Additionally, the concept of extremism was incorporated into both models to further analyze the instability brought about by extremists. Our research found that moderates in the DDRT model play a more dominant role than those in the CDRT model. Furthermore, upon the introduction of extremists into the network, the system may exhibit a phenomenon where the entire group reaches a single consensus.