Researchers tone down polarisation on X with tweaks to algorithm
Researchers at Stanford University say they have developed a tool that can noticeably reduce partisan hostility in X feeds by reordering posts rather than blocking them.
The study, published in the journal Science, suggests it may one day be possible to let users control their own social media algorithms, not only on X, formerly Twitter, but on other platforms.
After acquiring Twitter in 2022, tech billionaire Elon Musk removed many restrictions intended to protect users from hate speech and misinformation. Users who share Musk’s right-leaning political views saw their voices gain more weight on the service.
The Stanford team, working without collaboration from X, built a browser extension that resorted participants’ feeds.
Posts expressing anti-democratic attitudes and partisan hostility were pushed lower down, including content that endorsed violence or called for jailing supporters of the opposing party.
An AI language model analysed the feed in real time. Unlike an ad blocker that hides items, no content was deleted or blocked in the experiment.
The field trial was conducted ahead of the 2024 US presidential election with 1,256 participants on X. Subjects were randomly assigned to two parallel experiments in which their feeds were dynamically resorted for one week.
One group saw polarising posts ranked higher, the other saw them ranked lower. Among those whose anti-democratic content was downgraded, attitudes towards the opposing party became more positive.
The effect appeared across party lines and held for people identifying as liberal and those identifying as conservative.
“Social media algorithms directly impact our lives, but until now, only the platforms had the ability to understand and shape them,” said Michael Bernstein, a professor of computer science at the Stanford School of Engineering and the study’s lead author.
“We have demonstrated an approach that lets researchers and end users have that power.”
The tool could also open up ways to design measures that not only reduce partisan hostility but foster greater social trust and a healthier democratic discourse across party lines, Bernstein said.
Josephine Schmitt, scientific coordinator at the Center for Advanced Internet Studies (CAIS) in Germany, commented that the study showed robust, in some cases strong, effects on emotional tensions between political camps.
“The study makes it clear that even small algorithmic interventions measurably shift feelings towards the political ‘other’. That supports the basic statement: sorting a feed is not neutral, since it impacts on emotions and thus on affective polarisation,” she said.
Schmitt also pointed out that X plays a much smaller role in the day-to-day media landscape of some other countries than in the United States. Schmitt and Lorenz-Spreen were not involved in the study.
DPA Trends
Subscribers 0
Fans 0
Followers 0
Followers