[Paper Review] Quantifying social organization and political polarization in online platforms

Isaac Waller, Ashton Anderson|arXiv (Cornell University)|Oct 1, 2020

Social Media and Politics34 references137 citations

TL;DR

This paper introduces a neural embedding methodology to quantify social organization and political polarization in online platforms using behavioral data, applying it to 5.1 billion Reddit comments across 10,000 communities over 14 years. It reveals that political polarization on Reddit surged in 2016 primarily due to new right-leaning users, not individual radicalization, and that polarization is asymmetric, with no significant shift in left-leaning communities.

ABSTRACT

Optimism about the Internet's potential to bring the world together has been tempered by concerns about its role in inflaming the 'culture wars'. Via mass selection into like-minded groups, online society may be becoming more fragmented and polarized, particularly with respect to partisan differences. However, our ability to measure the social makeup of online communities, and in turn understand the social organization of online platforms, is limited by the pseudonymous, unstructured, and large-scale nature of digital discussion. We develop a neural embedding methodology to quantify the positioning of online communities along social dimensions by leveraging large-scale patterns of aggregate behaviour. Applying our methodology to 5.1B Reddit comments made in 10K communities over 14 years, we measure how the macroscale community structure is organized with respect to age, gender, and U.S. political partisanship. Examining political content, we find Reddit underwent a significant polarization event around the 2016 U.S. presidential election, and remained highly polarized for years afterward. Contrary to conventional wisdom, however, individual-level polarization is rare; the system-level shift in 2016 was disproportionately driven by the arrival of new and newly political users. Political polarization on Reddit is unrelated to previous activity on the platform, and is instead temporally aligned with external events. We also observe a stark ideological asymmetry, with the sharp increase in 2016 being entirely attributable to changes in right-wing activity. Our methodology is broadly applicable to the study of online interaction, and our findings have implications for the design of online platforms, understanding the social contexts of online behaviour, and quantifying the dynamics and mechanisms of online polarization.

Motivation & Objective

• To address the challenge of measuring social organization in large-scale, pseudonymous online communities where self-reported data is unreliable.
• To understand how political polarization evolves on online platforms, particularly whether it results from individual radicalization or community-level shifts.
• To investigate whether polarization dynamics are symmetric across ideological lines or exhibit directional asymmetry.
• To develop a methodology that captures social dimensions—such as partisanship, age, and gender—based purely on aggregate user behavior, avoiding biases from self-reporting or expert labeling.

Proposed method

• Uses neural community embeddings to represent communities as vectors in a high-dimensional space based on user membership patterns.
• Applies a two-step algorithm: first, identifies seed community pairs differing along a target social dimension (e.g., left vs. right-wing politics); second, computes a consensus vector from similar directional pairs to define that dimension.
• Employs UMAP for visualization of community embeddings and measures partisan positioning via z-scores relative to the mean.
• Validates gender and political orientation embeddings using external data (e.g., U.S. Census occupation data and Reddit user self-identification) to ensure alignment with real-world demographics.
• Analyzes temporal shifts in community embeddings to detect changes in polarization over time, particularly around the 2016 U.S. election.
• Uses word usage patterns in political communities to interpret the semantic content associated with each end of the partisan dimension.

Experimental results

Research questions

• RQ1 To what extent does platform-level political polarization change over time?
• RQ2 Do individual users become more polarized in their political activity over time, and if so, do these changes drive platform-level polarization?
• RQ3 Are the dynamics of polarization ideologically symmetric?
• RQ4 How do new users influence the macroscale social structure of online communities?

Key findings

• Political polarization on Reddit increased significantly around the 2016 U.S. presidential election, with the platform remaining highly polarized for years afterward.
• The 2016 polarization surge was driven disproportionately by the arrival of new users, particularly on the right wing, rather than by individual-level radicalization.
• Individual-level political polarization is rare; the system-level shift was not due to users becoming more extreme over time.
• Political polarization on Reddit is temporally aligned with external events, not with users’ prior platform activity or algorithmic curation.
• There is a stark ideological asymmetry: the 2016 increase in polarization was entirely attributable to changes in right-wing user behavior, with no corresponding shift observed on the left.
• The methodology successfully captures social dimensions such as partisanship, age, and gender using only behavioral data, with validation against external demographic benchmarks.