Users Polarization on Facebook and Youtube (1604.02705v1)

Abstract: On social media algorithms for content promotion, accounting for users preferences, might limit the exposure to unsolicited contents. In this work, we study how the same contents (videos) are consumed on different platforms -- i.e. Facebook and YouTube -- over a sample of $12M$ of users. Our findings show that the same content lead to the formation of echo chambers, irrespective of the online social network and thus of the algorithm for content promotion. Finally, we show that the users' commenting patterns are accurate early predictors for the formation of echo-chambers.

• The paper reveals high polarization and echo chamber formation among users consuming scientific versus conspiracy content on Facebook and YouTube, quantified by user interaction patterns.
• The study finds that users exhibit consistent interaction patterns on both Facebook and YouTube, demonstrating significant correlation in behavior despite differences in platform algorithms.
• A predictive model effectively forecasts user polarization across platforms based on early interaction patterns, highlighting consistent engagement transcending specific algorithmic designs.

Analysis of Users Polarization on Facebook and YouTube

This paper examines the formation of echo chambers across two major social media platforms, Facebook and YouTube, focusing on the consumption of content that falls into two distinct categories: scientific and conspiracy-like narratives. Utilizing a dataset comprising 12 million users, the paper scrutinizes how algorithms designed to promote user-specific content inadvertently result in polarized communities, disregarding the specific online network or content promotion mechanism at play.

The research methodically compares the actions of users (likes, shares, comments) on videos linked through Facebook posts and hosted on YouTube. By leveraging statistical analysis techniques such as Spearman's rank correlation and the Mantel test, the paper identifies similar behavioral patterns across the platforms despite differing content recommendation algorithms.

Key Findings

1. Echo Chamber Formation: The paper reveals a high level of polarization among users interacting with scientific versus conspiracy content. This is evidenced by bimodal distributions in user actions, indicating that most individuals are polarized towards one narrative, as quantified by a substantial Bimodality Coefficient (BC).
2. Cross-Platform Behavioral Consistency: The investigation shows that users demonstrate analogous patterns in their interactions on both Facebook and YouTube. The research establishes a significant Spearman correlation in user behavior irrespective of the underlying content promotion algorithms, such as Facebook’s News Feed prioritization and YouTube’s Watch Time algorithm.
3. Predictive Modeling of Polarization: A multinomial logistic regression model effectively forecasts user polarization. The model, based on early interaction patterns, predicts whether a user will become entrenched in a specific narrative or continue consuming diverse viewpoints. This ability to generalize across platforms underscores a consistent engagement pattern that transcends the specificities of the platform's algorithms.

Implications and Future Directions

The implications of this paper are noteworthy, as the findings suggest that algorithmic content promotion facilitates the crystallization of echo chambers, an outcome consistent across diverse social networking sites. Notably, this suggests the need to reconsider algorithmic designs that overly cater to user preferences, potentially exacerbating ideological segregation.

Practically, these insights can guide platform designers in developing more effective content delivery systems that promote exposure to diverse opinions and mitigate polarization. Future research could explore intervention strategies that leverage cross-platform user behaviors to disrupt echo chamber formation. Further investigations could also delve into cognitive and psychological factors driving these patterns, paving the way for more holistic solutions aimed at enhancing online dialogue inclusivity.

In conclusion, this paper contributes significant empirical evidence on the phenomena of online polarization. By unveiling consistent patterns amid different content promotion systems, it provokes essential discussions on the broader impacts of algorithmic content curation in social media environments.