News release

From: Springer Nature

Social sciences: X’s algorithm may influence political attitudes

Turning on the ‘For You’ algorithm on X (formerly Twitter) may shift users’ political opinions towards more conservative views, suggests research involving nearly 5,000 X users. These effects are shown to persist even after users return to a chronological feed, according to a paper published in Nature.

Social media has become a central source of political news for many people, prompting concerns about misinformation, polarization and the influence of algorithms (which select and order content in personalized feeds). Previous large scale experiments, including a collaboration with Meta, found little evidence that switching off the algorithm and reverting to a chronological feed altered users’ political attitudes. However, those studies could not determine whether early algorithmic exposure had already shaped political views.

Ekaterina Zhuravskaya and colleagues conducted an independent field experiment with 4,965 active X users in 2023 in the US. Participants completed surveys before and after the experiments and were randomly assigned to use either the algorithmic or the chronological feed for approximately 7 weeks. The authors also collected data on the participants' feed content via a custom web browser extension and monitored participants’ online behaviour. The results show that users assigned to the algorithmic feed engaged more with the platform, adopted more conservative policy priorities and were more likely to follow conservative political activist accounts. By contrast, switching users from the algorithmic to the chronological feed had little effect on their views or following behaviour. Analysis of feed content revealed that the algorithm showed more conservative and activist posts while demoting traditional news outlets.

The study offers new evidence that social media algorithms can meaningfully shape political attitudes and that these effects endure even after algorithmic curation is removed. The authors conclude that algorithms influence not only what users see but also the online political environment they inhabit.