A Viral Marketing-Based Model For Opinion Dynamics in Online Social Networks

TL;DR

This paper introduces a new model combining opinion dynamics and viral content spread to analyze how viral marketing and polarizing content influence opinions and polarization in online social networks.

Contribution

The paper presents a novel integrated model for opinion change and information cascades, along with algorithms for simulation and optimization, validated on real-world data.

Findings

Viral content can significantly increase network polarization.

Polarizing content can raise polarization by up to 59%.

Marketing campaigns have limited impact on polarization.

Abstract

Online social networks provide a medium for citizens to form opinions on different societal issues, and a forum for public discussion. They also expose users to viral content, such as breaking news articles. In this paper, we study the interplay between these two aspects: opinion formation and information cascades in online social networks. We present a new model that allows us to quantify how users change their opinion as they are exposed to viral content. Our model is a combination of the popular Friedkin--Johnsen model for opinion dynamics and the independent cascade model for information propagation. We present algorithms for simulating our model, and we provide approximation algorithms for optimizing certain network indices, such as the sum of user opinions or the disagreement--controversy index; our approach can be used to obtain insights into how much viral content can increase these indices in online social networks. Finally, we evaluate our model on real-world datasets. We show experimentally that marketing campaigns and polarizing contents have vastly different effects on the network: while the former have only limited effect on the polarization in the network, the latter can increase the polarization up to 59% even when only 0.5% of the users start sharing a polarizing content. We believe that this finding sheds some light into the growing segregation in today's online media.