TikTok's recommendations skewed towards Republican content during the 2024 U.S. presidential race

TL;DR

Significant asymmetries in content distribution exist across all three states and persist when accounting for video- and channel-level engagement metrics, and are driven primarily by negative partisanship content.

Abstract

TikTok is a major force among social media platforms with over a billion monthly active users worldwide and 170 million in the United States. The platform's status as a key news source, particularly among younger demographics, raises concerns about its potential influence on politics in the U.S. and globally. Despite these concerns, there is scant research investigating TikTok's recommendation algorithm for political biases. We fill this gap by conducting 323 independent algorithmic audit experiments testing partisan content recommendations in the lead-up to the 2024 U.S. presidential elections. Specifically, we create hundreds of "sock puppet" TikTok accounts in Texas, New York, and Georgia, seeding them with varying partisan content and collecting algorithmic content recommendations for each of them. Collectively, these accounts viewed ~394,000 videos from April 30th to November 11th, 2024, which we label for political and partisan content. Our analysis reveals significant asymmetries in content distribution: Republican-seeded accounts received ~11.8% more party-aligned recommendations compared to their Democratic-seeded counterparts, and Democratic-seeded accounts were exposed to ~7.5% more opposite-party recommendations on average. These asymmetries exist across all three states and persist when accounting for video- and channel-level engagement metrics such as likes, views, shares, comments, and followers, and are driven primarily by negative partisanship content. Our findings provide insights into the inner workings of TikTok's recommendation algorithm during a critical election period, raising fundamental questions about platform neutrality.

Figures (14)

• Figure 1: Experimental Setup. 21 experiment runs are performed each week between April 30th and November 4th 2024. Bots are assigned an experimental condition based on two attributes, namely, their VPN and GPS location (New York, Texas, or Georgia), as well as the political leaning (Democratic or Republican) of the videos they watch during the conditioning stage. Each bot initially passes through a conditioning stage where it views up to 400 videos of a given political leaning, and then transitions to the recommendation stage, where it views up to 1200 videos on their TikTok "For You" page.
• Figure 1: A device's timeline during a weekly experimental run
• Figure 2: Differences in recommendation rates over time. (A-C) The mean ideological alignment for bots of a given conditioning-leaning in New York, Texas and Georgia, measured as the difference between the proportion of recommended videos which are Republican aligned minus the proportion of Democratic aligned videos. OLS regression lines are plotted with 95% confidence intervals. (D-F) The observed ideological skew for each state over time; positive values indicate a Republican skew. (G) The observed ideological skew, and the expected ideological skew based on counterfactual models built on different engagement metrics. (H) Log odds ratios computed through logistic regression on the likelihood of a video being a mismatch/cross-partisan. Statistically significant coefficients are highlighted in navy.
• Figure 2: Rolling average of political content viewed by bots of different conditioning in 10 video windows.
• Figure 3: Mismatch proportions of top Democratic and Republican channels. (A) The mismatch proportion, or the proportion of video watches by bots of an opposite conditioning-leaning for top Democratic and Republican TikTok channels by follower count. (B) The mean mismatch proportions of the TikTok accounts for the main political figures in the 2024 U.S. elections (Trump, JD Vance, Kamala Harris, and Tim Walz) (C) Logistic regression estimates on the log-odds ratio of recommended video mismatch for top Republican and Democratic channels (Statistically significant coefficients are highlighted in navy).