A study published in the scientific journal Nature has found that TikTok's algorithm repeatedly prioritized pro-Republican content in three US states during the 2024 presidential election campaign.
The researchers created hundreds of test accounts and “trained” them to behave like real users, watching videos related to the Democratic or Republican Party. They then analyzed the content that TikTok recommended on the For You page, the platform’s main feed.
According to the study authors, the results showed a consistent disparity. Accounts initially exposed to pro-Republican content saw about 11.5 percent more videos that aligned with their beliefs, compared to pro-Democrat accounts. Meanwhile, accounts trained with pro-Democrat content were about 7.5 percent more likely to see pro-Republican content on their For You page.
The study was conducted with 323 test accounts, virtually located in New York, Texas, and Georgia. Over 27 weeks of the 2024 presidential campaign, the researchers analyzed more than 280 recommended videos, using a combination of human rating and artificial intelligence.
Talal Rahwan, one of the study's authors from the New York University Abu Dhabi campus, said the algorithm wasn't just showing users more of what they wanted to see. He said pro-Democrat accounts were exposed significantly more to anti-Democratic content than pro-Republican accounts were to anti-Republican content.
The researchers also noticed differences in the topics that appeared. Pro-Democrat accounts received more content from the other side on issues like immigration and crime, while pro-Republican accounts saw more opposing content on abortion.
TikTok disputed the study, saying the experiment with fake accounts did not reflect how people actually use the platform. The company said users discover and view a wide range of content and have multiple tools to control what is shown to them.
The study's authors acknowledge that the findings should not be generalized beyond the three countries analyzed. They also emphasize that the study does not measure the impact of these videos on users' political beliefs or behavior, nor does it explain why this disparity exists.
Another limitation is that the test accounts only reflected the early stages of a user's experience on the platform. Also, the analysis was based primarily on English-language transcripts, which may not capture political messages conveyed through visuals, tone, or other languages.
However, researchers say the findings are important for the debate over platform transparency and algorithm accountability, especially when it comes to political content and young voters.
According to them, in elections where the margins can be very close, the way platforms recommend political information to millions of users deserves serious attention. /GazetaExpress/
t7 live
