"Feed algorithms are widely suspected to influence political attitudes." This study shows that some do - to significant effect.

Germain Gauthier, Roland Hodler, Philine Widmer and Ekaterina Zhuravskaya in Nature]

This is a very significant finding. Users who moved from a reverse-chronological social media algorithm to X’s:

“[…] were 4.7 percentage points more likely to prioritize policy issues considered important by Republicans, such as inflation, immigration and crime. They were also 5.5 percentage points more likely to believe that the investigations into Trump are unacceptable, describing them as contrary to the rule of law, undermining democracy, an attempt to stop the campaign and an attack on people like themselves.”

And even more surprisingly, once the algorithm was switched off, their views did not change again. The effect of the algorithm lingered, in part because it led users to follow more conservative influencers.

We intuitively knew that the algorithm mattered, but this is a key finding that puts numbers to it. If that number seems small to you, consider that 4.7% is more than enough to swing an election. It’s also interesting that findings for other algorithms were different; if this result holds up, it suggests that X’s algorithm may be particularly predisposed for political manipulation, even above Facebook and Instagram.

This should be a wakeup call for politically-engaged funders and anyone who cares about civil society. It’s not that we need to have less conservative algorithms; it’s that whoever controls the algorithms has a disproportionate say over the electorate’s view of the world.

We need more funding into open protocols that decentralize algorithmic ownership; open platforms that give users a choice of algorithm and platform provider; and algorithmic transparency across our information ecosystem.

[Link]