Researchers created a web extension paired with a large language model that reads posts in an X feed. The tool identifies posts with antidemocratic words or strong partisan anger and then moves those posts up or down in the feed.
About volunteers used the tool during the 2024 election in short trials. In some tests people saw the negative content downranked and in other tests they saw it upranked. No posts were removed; only their order changed. People who saw negative content downranked reported warmer feelings toward the opposing party.
Difficult words
- extension — small software added to a web browserweb extension
- large language model — computer program that creates text like humans
- feed — a list of posts shown on a website
- antidemocratic — against democracy and democratic principles
- partisan — showing strong support for one political group
- downrank — to move something lower in a list or feeddownranked
- uprank — to move something higher in a list or feedupranked
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- How would you feel if negative posts were moved lower in your feed?
- Do you think changing the order of posts can change people's views? Why or why not?
Related articles
OpenSpeaks Archives helps document Van Gujjari
OpenSpeaks Archives, launched in 2024, helps Wikimedians cite Indigenous oral knowledge and now hosts nearly 20 languages from India, Nepal and Sri Lanka. An interview with Taukeer Alam explains why audio and video matter and led to a 2024 project.
Study finds flaws in cloud password managers
Researchers at ETH Zurich tested three cloud-based password managers and found multiple attacks that could expose or change users' passwords. They followed responsible disclosure, gave companies time to fix the issues, and recommended stronger encryption and audits.