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

TL;DR

The paper analyzes how X's algorithmic recommendations shape political content exposure during the 2024 U.S. presidential election. It uses a six-week sock-puppet audit with 120 accounts in neutral, left, right, and balanced groups to measure out-of-network exposure via a weighted occurrence metric, Gini coefficient, and amplification ratio. Key findings show exposure concentrated on a small set of high-popularity accounts, with right-leaning users experiencing the greatest inequality; both left and right timelines amplify ideologically aligned content and mute opposing viewpoints; neutral accounts display a default right-leaning bias. The work underscores the need for transparency-aware algorithms to safeguard election integrity and promote a more informed digital public sphere.

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 what implications does this have for democratic discourse online? In this paper, we present a six-week audit of X's algorithmic content recommendations during the 2024 U.S. Presidential Election by deploying 120 sock-puppet monitoring accounts to capture tweets from their personalized "For You" timelines. Our objective is to quantify out-of-network content exposure for right- and left-leaning user profiles and assess any potential inequalities and 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 that new accounts experience a right-leaning bias in exposure within their default timelines. Our work contributes to understanding how content recommendation systems may induce and reinforce biases while exacerbating vulnerabilities among politically polarized user groups. We underscore the importance of transparency-aware algorithms in addressing critical issues such as safeguarding election integrity and fostering a more informed digital public sphere.

Figures (8)

• Figure 1: Distribution of Gini coefficient across different groups of accounts. Significant disparities are found in all pairwise comparisons (Mann-Whitney U test: $p$ < 0.001), with right-leaning users experiencing the highest out-of-network exposure inequality.
• Figure 2: Lorenz curves for different groups of accounts. Each curve represents the average Lorenz curve for all accounts in the group, with error bars indicating the standard deviations at each cumulative point.
• Figure 3: Top 20 recommended users for neutral accounts, ranked by their average weighted occurrence per 1,000 tweets. Each box in the boxplot shows the distribution of exposure across all neutral accounts, with red and blue colors indicating right- and left-leaning users, respectively. The figure suggests that right-leaning users are more frequently recommended than left-leaning users in the algorithm's out-of-network recommendations for neutral accounts.
• Figure 4: Amplification ratio of the top 50 recommended users in left-leaning (top) and right-leaning (bottom) accounts, compared to the baseline of balanced accounts. Colored bars indicate a significant difference in exposure metrics (weighted occurrence per 1,000 tweets) between the groups at the 0.05 significance level (using the Mann-Whitney U test), while gray bars indicate no significant difference. Usernames are displayed in blue (left-leaning) or red (right-leaning) based on their political stance, according to publicly available data.
• Figure 5: Overview of data collection: (a) Number of active accounts per day, and (b) Number of tweets collected per day.