- Co-evolution of Opinions and Signed Network dynamics in Real World Scenarios
- Project Year:
2021
- REU Student (s):
Soham Palande | Rutgers University-New Brunswick NJ
- Student 1 Institution:
Rutgers University-New Brunswick
- Project Mentor:
Jie Gao
- Project Mentor Area:
Computer Science
- Project Abstract:
With the pervasion of social media in modern society, evolution of opinions and network dynamics in social networks have gained special interest from industry, academia, and governments. The study of opinion and network dynamics has several applications- social sciences, behavioral economics, game theory- with consequential implications. Adversarial attacks through fake news and troll bots are increasingly common and there is a need for a rigorous understanding of their impact on individuals and the social networks. We study opinion and network dynamics and characterize social networks at their limit states extending recent work in co- evolution of signed networks and opinions to better represent the real world. We build upon intuition of the real world, in which individuals hold multiple opinions on a range of issues for example- healthcare, gun control and seek to better represent complex relationships/social ties between individuals in a network. We formulate these intuitions mathematically and incorporate them in a co-evolution model which combines signed network dynamics based on structural theory and opinion dynamics. We conduct a comprehensive study through numerous simulations and validate the model on the benchmark Zachary's Karate Club dataset to achieve 100% accuracy. We mathematically address divergence of social ties to infinity, which in the real world would translate to limitless polarization, by introducing normalization which keeps the relative magnitude of the social ties intact but prevents divergence.
