A team of researchers built a web extension that uses a large language model to scan an X feed for antidemocratic attitudes and partisan animosity, like calls for violence or jailing opponents. The tool reorders posts so that targeted content appears lower or higher in the feed; it works in a matter of seconds and does not remove any posts.
In experiments about 1,200 volunteers used the extension over ten days during the 2024 election. Separate seven-day trials compared groups that saw content downranked, upranked, or left unchanged as a control.
Participants who saw antidemocratic content downranked reported warmer feelings toward people in the opposing party. The effect appeared for both people who identified as liberals and those who identified as conservatives. Researchers measured attitudes on a 1 to 100 scale and found an average improvement of two points for the downranked group.
Difficult words
- antidemocratic — against democratic rules and institutions
- partisan — strongly supporting one political party
- animosity — strong anger or dislike between groups
- reorder — change the order of items in a listreorders
- downrank — move an item lower in a listdownranked
- uprank — move an item higher in a listupranked
- scale — a set of numbers for measuring something
Tip: hover, focus or tap highlighted words in the article to see quick definitions while you read or listen.
Discussion questions
- Do you think reordering posts is better than removing them? Why or why not?
- Would you use an extension like this on your social media? Explain your reasons.
- The study found an average improvement of two points on a 1 to 100 scale. Do you think this change matters? Why?
Related articles
New AI tools for tuberculosis shown at lung health conference
Researchers presented four new AI approaches for detecting and monitoring TB at the Union World Conference on Lung Health in Copenhagen (18–21 November). The tools include breath analysis, cough screening, vulnerability mapping and a chest X‑ray tool for children.
Small pause to slow misinformation on social media
Researchers at the University of Copenhagen propose a small pause before sharing on platforms like X, Bluesky and Mastodon. A computer model shows that a short delay plus a brief learning step can reduce reshares and improve shared content quality.