Topological Conditions for Echo Chamber Formation under the FJ model: A Cluster Consensus-based Approach

TL;DR

This paper investigates the network structures that lead to echo chamber formation in social networks using the Friedkin-Johnsen model, providing topology-based conditions applicable to arbitrary directed graphs.

Contribution

It introduces necessary and sufficient topological conditions for cluster consensus under the FJ model, applicable to arbitrary digraphs, and explains the emergence of bow-tie structures in echo chambers.

Findings

Topology-based conditions for cluster consensus are derived.

Conditions are independent of edge weights and stubbornness values.

Methodology for verifying conditions computationally is developed.

Abstract

The Friedkin-Johnsen (FJ) model is a popular opinion dynamics model that explains the disagreement that can occur even among closely interacting individuals. Cluster consensus is a special type of disagreement, where agents in a network split into subgroups such that those within a subgroup agree and those in different subgroups disagree. In large-scale social networks, users often distribute into echo chambers (i.e. groups of users with aligned views) while discussing contested issues such as electoral politics, social norms, etc. Additionally, they are exposed only to opinions and news sources that align with their existing beliefs. Hence, the interaction network plays a key role in the formation of an echo chamber. Since cluster consensus can represent echo chambers in a social network, we examine the conditions for cluster consensus under the FJ model with the objective of determining the properties of the interaction network that lead to echo chamber formation. We present topology-based necessary and sufficient conditions for cluster consensus under the FJ model, regardless of the edge weights in the network and stubbornness values (which are difficult to estimate parameters in a social network). A major advantage of the proposed results is that they are applicable to arbitrary digraphs. Moreover, using the proposed conditions, we explain the emergence of bow-tie structures which are often observed in real-world echo chambers. Finally, we also develop a computationally feasible methodology to verify the proposed conditions for cluster consensus.