A huge Facebook experiment changed the feed. It didn’t change people’s politics.

The echo-chamber argument sounds simple: show people fewer like-minded posts and they should become less polarized. A large real-world experiment found a more complicated answer.

During the 2020 U.S. presidential election, researchers changed what appeared in some Facebook feeds. Among 23,377 adult participants, the intervention cut exposure to posts from politically like-minded sources by about one-third.

People saw more posts from sources with different political leanings and less uncivil content. But the experiment found no measurable change in eight preregistered political attitudes, including affective polarization and ideological extremity.

That does not mean feeds have no influence: the intervention clearly changed what people saw. It means this particular change, on Facebook, with U.S. adults over a few months in 2020, did not measurably shift those attitudes. The researchers also found that people were more likely to engage with like-minded posts they still encountered.

When an argument erupts: two people can see different posts without either screenshot proving the other side is lying. Check the original source and the exact claim before deciding what the feed left out.

The study concerns the United States, Facebook and the 2020 election; it does not establish what every platform’s current algorithm does.

Source: A preregistered field experiment on Facebook, published in Nature.