Reranking partisan animosity in algorithmic social media feeds alters affective polarization

Experimental research using novel large language model-based methods demonstrates how, when it comes to the feed-ranking algorithms of social media platforms, reranking the visibility of hostile political content can reduce political polarization.