X’s algorithm leads users towards more conservative political positions

Details of the research, carried out by an international team of scientists from Italy, Switzerland and France, have been published in Nature.

For many people, social networks have become the main source of information, and that has generated concern about misinformation, polarization and the influence of algorithms (which filter, select, and order content on personalized walls to build customer loyalty).

Some previous large-scale experiments — including a collaboration with Meta — found little evidence that turning off the information-filtering algorithm and reverting to a chronological wall altered users’ political attitudes. But those studies couldn’t determine whether early exposure to the algorithm had already shaped those opinions, the authors note.

To clarify, they carried out an independent experiment with 4,965

The study participants were randomly distributed among a ‘feed’ algorithmic (a news filter) and the chronological wall (the normal version) for seven weeks.

Additionally, they surveyed participants before and after the experiment and reviewed wall content and users’ online behavior with a browser extension that allowed them to monitor wall content and online interactions.

The results revealed that users assigned to the algorithmic wall interacted more with the platform, They adopted political views closer to the right and were more likely to follow conservative political activists.

In contrast, when users were moved from the algorithmic wall to the chronological one, their opinions or behavior had little effect.

In addition, content analysis revealed that the algorithm showed more conservative and activist publications, and reduced the visibility of traditional news media, the study notes.

For the authors, these results demonstrate that social media algorithms “significantly shape political attitudes” and that this effect persists even after eliminating algorithmic choice.

Besides, algorithms influence not only what users see, but also the digital political environment they inhabitthey warn.

As limitations of the study, the authors highlight, firstly, the fact that the results must be limited to network