A team at three U.S. universities developed an AI-powered browser extension that modifies how content appears on X and then tested it on 1,200 participants in the run up to the 2024 U.S. presidential election. The tool scans posts for anti-democratic and negative partisan views (such as calling for violence towards supporters of an opposing party) and then re-orders the posts so polarizing content is shown either more or less to participants. The scientists found that attitudes towards the opposing party changed on average by two points, and that the participants who saw less hostile content reported feeling less angry and sad while using X. The extension did not need the participation of X and, the team behind it noted, could be a way of reranking social media posts "without platform collaboration." Here's more from @euronews.