Auditing Political Exposure Bias: Algorithmic Amplification on Twitter/X Approaching the 2024 U.S. Presidential Election

Jinyi Ye,Luca Luceri,Emilio Ferrara

2024-11-12

Abstract:Approximately 50% of tweets in X's user timelines are personalized recommendations from accounts they do not follow. This raises a critical question: what political content are users exposed to beyond their established networks, and how might this influence democratic discourse online? Due to the black-box nature and constant evolution of social media algorithms, much remains unknown about this aspect of users' content exposure, particularly as it pertains to potential biases in algorithmic curation. Prior research has shown that certain political groups and media sources are amplified within users' in-network tweets. However, the extent to which this amplification affects out-of-network recommendations remains unclear. As the 2024 U.S. Election approaches, addressing this question is essential for understanding the influence of algorithms on online political content consumption and its potential impact on users' perspectives. In this paper, we conduct a three-week audit of X's algorithmic content recommendations using a set of 120 sock-puppet monitoring accounts that capture tweets in their personalized ``For You'' timelines. Our objective is to quantify out-of-network content exposure for right- and left-leaning user profiles and to assess any potential biases in political exposure. Our findings indicate that X's algorithm skews exposure toward a few high-popularity accounts across all users, with right-leaning users experiencing the highest level of exposure inequality. Both left- and right-leaning users encounter amplified exposure to accounts aligned with their own political views and reduced exposure to opposing viewpoints. Additionally, we observe a right-leaning bias in exposure for new accounts within their default timelines.

Social and Information Networks

What problem does this paper attempt to address?

### What problems does this paper attempt to solve?

This paper aims to explore and quantify the bias in the exposure of users to political content on the Twitter/X platform's algorithm - recommended system during the 2024 US presidential election. Specifically, the study focuses on the following issues:

1. **Transparency and responsibility of the algorithm - recommended system**: Since the algorithms of social media platforms (such as X) are black - box in nature and constantly evolving, users have a limited understanding of the content they are recommended. Especially when it comes to political content, this opacity may affect democratic discussion.

2. **Exposure to political content outside the user's network**: Approximately 50% of tweets come from accounts that users do not follow, which raises key questions: What kind of political content will users be exposed to outside their own social networks? How does this content affect online political discussion?

3. **Algorithm amplification effect**: Previous studies have shown that certain political groups and media sources are amplified in tweets within the user's network. However, it is unclear whether the degree of this amplification will affect the recommended content outside the network.

4. **Differences in exposure among users with different political leanings**: Will left - leaning and right - leaning users be more exposed to content that is consistent with their own views and less exposed to opposing views? Is there a bias in the default content recommendation for newly registered users or users without specific preferences?

To answer these questions, the author conducted a three - week audit experiment, using 120 "puppet accounts" to capture tweets in the personalized "Recommended for You" timeline. Through this method, researchers were able to quantify the out - of - network content exposure of users with different political leanings and evaluate potential biases.

### Main findings

- **High - popularity accounts dominate exposure**: In the timelines of all users, the exposure is concentrated on a small number of high - popularity accounts, especially right - leaning users who experience the highest exposure inequality.
- **Homogeneous content amplification**: Both left - leaning and right - leaning users are more frequently exposed to accounts that are consistent with their own political positions and less exposed to opposing views.
- **Default right - leaning bias**: For newly registered users or users who do not follow any accounts, there is a right - leaning bias in the default recommended content.

### Significance of the study

These findings are of great significance for the 2024 US presidential election. Social media platforms play an important role in shaping political discussion, and the amplification of political content by algorithms may strengthen the information cocoons effect and influence users' perceptions. Especially for new users or users without strong pre - existing preferences, the right - leaning bias in the default recommended content may subtly influence their perspectives.

In addition, the study also reveals that algorithms not only amplify the content of media organizations and politicians, but also increase the exposure rate of political commentators and influencers, which may have a profound impact on public opinion. Therefore, the study emphasizes the need for continuous monitoring of algorithm changes and their impact on content exposure dynamics to ensure transparency and accountability, especially at critical moments related to the democratic process and election fairness.