Measuring Polarisation Through Network Graphs

Core Concepts

Polarisation is measured by the decrease in dimensionality of network graphs, reflecting increasing correlation in social networks.

Abstract

Abstract

Current approaches to detecting polarisation rely on bimodality in social networks or voter surveys.
Defining polarisation as increasing correlation between ideological positions reduces political pluralism.

Methods

Using Random Dot Product Graphs to embed social networks in metric spaces.
Truncated Singular Value Decomposition indicates increasing polarisation in networks.

Data

Analyzing climate change discussions on New Zealand Twitter and COP conferences.
Synthetic data simulations to explore polarisation dynamics.

Results

Dimensionality decreased over time in climate change discussions, indicating increased polarisation.
Unexpected results in COP discussions, with dimensionality not linearly decreasing.
Simulated data showed the impact of group engagement and size on network dimensionality.

Conclusions

Novel method using RDPGs to measure polarisation efficiently and reliably.
Von Neumann entropy did not closely relate to dimensionality, suggesting dimensionality as a better measure.
Limitations in capturing affective polarisation and potential improvements in future research.

Stats

A decrease in the optimal dimensionality for the embedding of the network graph is indicative of increasing polarisation.
The adjacency matrix was constructed based on user interactions in COP discussions.

Quotes

"Our method captured the presence of polarisation in all scenarios where it was expected and found by other researchers."
"Our approach is highly interpretable, focusing on the number of dimensions rather than specific ideological positions."

Key Insights Distilled From

The process of polarisation as a loss of dimensionality

by Sage Anastas... at arxiv.org 03-28-2024

https://arxiv.org/pdf/2403.18191.pdf

Deeper Inquiries

How can affective polarisation be better integrated into the measurement method?

To better integrate affective polarisation into the measurement method, we can incorporate sentiment analysis techniques to determine the nature of interactions within the social network. By classifying interactions as positive, negative, or neutral, we can distinguish between positive social bonds and antagonistic reactions. This approach would allow us to capture the emotional dynamics of the network and differentiate between genuine connections and hostile interactions. By incorporating sentiment analysis, we can refine the measurement of polarisation to account for affective dimensions, providing a more comprehensive understanding of how emotions influence network dynamics.

What are the implications of unexpected results in COP discussions for future research on polarisation?

The unexpected results in COP discussions, where the dimensionality of the network did not align with the von Neumann entropy as expected, have significant implications for future research on polarisation. These findings suggest that traditional assumptions about polarisation dynamics may not always hold true and that the relationship between network complexity and polarisation is more nuanced than previously thought. Researchers should consider alternative metrics and approaches to measure polarisation, such as the embedded dimensionality method used in this study, to capture the multifaceted nature of polarisation in social networks accurately. Additionally, future studies should explore the interplay between network complexity, entropy, and polarisation to gain a more comprehensive understanding of how these factors interact in different contexts.

How can the method be extended to capture polarisation dynamics in more complex social networks?

To capture polarisation dynamics in more complex social networks, the method can be extended in several ways:

Multi-dimensional Embeddings: Instead of focusing on a single dimension, the method can be expanded to incorporate multi-dimensional embeddings to account for the diverse ideological positions and social determinants within the network.
Dynamic Analysis: Implementing a dynamic analysis approach to track changes in dimensionality over time can provide insights into the evolving nature of polarisation within the network.
Incorporating Network Structures: Considering the network structures, such as community detection algorithms, can help identify subgroups and their interactions, offering a more nuanced understanding of polarisation dynamics.
Sentiment Analysis: Integrating sentiment analysis to capture affective polarisation can enhance the method's ability to differentiate between positive and negative interactions, providing a more comprehensive view of polarisation dynamics in complex social networks.