Co-evolution of opinions and social network: a simulation study

• Main Author: Low, John Enze
• Other Authors: Xiao Gaoxi
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering::Computer hardware, software and systems
• Online Access: https://hdl.handle.net/10356/157395
• Institution: Nanyang Technological University

Studying social network graphs has always been a fascinating topic since we can learn more about real-life human interactions with one another using social interaction simulations paired with varying parameters in different scenarios. Although there have been prior studies conducted about consensus making between nodes in a social network graph, little is known about the repulsive effect when 2 nodes disagree with each other. Therefore, the goal of this research is to examine the reactions of nodes if they are in a situation of disagreement with their neighbors and observe how the overall social network structure is like after the social network reaches a steady state. This study was carried out in Jupyter acting as an IDE, with Python as the main language of use. 1 basic model and 5 other situational models were used in this study. The steps of the study are relatively straightforward. A virtual environment is first created for the social network graph, followed by the actual creation of the graph itself, and finally the implementation of different parameters onto the individual nodes. The results of the situational models where repulsive effects are present without disruption were all similar when pushed to its steady states which points to all its nodes having values of 0 or 1 only. However, the observations between all these models before its steady state is reached is slightly different. In scenarios where the repulsive effect is met with resistance such as nodes being left alone after a certain threshold, the steady state values consist of 3 distinct communities with their opinion values at 0, 0.5, and 1. Although 5 different scenarios were implemented in replicating a real-life scenario, it is evident that many more parameters must be introduced to these models in order to achieve a higher accuracy of real-life opinion repulsive depiction.