science
Study Tests Algorithm Effects on Political Content Exposure
Researchers assigned 2,000 participants to custom social media feeds for eight weeks around the 2024 U.S. presidential election. The study compared an engagement-based algorithm, a diversified-extremity algorithm, and a reverse-chronological feed.
NA
app.buzzsumo.comAudio version
Tap play to generate a narrated version.
Developing·Limited corroboration so far. This page will refresh as more sources emerge.
A Registered Report published in Nature examined how different feed-ranking algorithms affect exposure to intergroup, moralized, and emotional content during a national election. U.S. presidential election. One algorithm ranked content by predicted engagement, another reduced the influence of extreme users, and a third presented posts in reverse-chronological order.
The engagement-based feed increased exposure to moral outrage and political content compared with the reverse-chronological feed. It also raised the amount of toxic content users encountered. The diversified-extremity algorithm lowered exposure to intergroup, moralized, emotional, and toxic material relative to the engagement-based feed.
Participants using engagement-based feeds showed reduced accuracy in perceiving prescriptive social norms and reported higher perceived partisan animosity. The diversified-extremity feed improved prescriptive norm accuracy while maintaining similar levels of platform enjoyment.
Neither algorithm produced statistically significant changes in participants' own posting or engagement behaviors. The Stage 1 protocol for the study was accepted in principle on 17 September 2024.
