Story perspectives

Engagement Algorithm Amplifies Toxic Outrage; Diversified Feed Reduces It

5/29/2026

Summary

Story summary

Researchers ran an eight-week randomized experiment with 2,000 U.S. participants during the 2024 presidential election.
The engagement-based algorithm amplified intergroup and emotional (IME) toxic content and boosted moral outrage and political messaging versus a reverse-chronological feed.
That feed lowered prescriptive norm accuracy, raised perceived partisan animosity, and left users’ own engagement unchanged.
The diversified-extremity algorithm cut IME toxic exposure, improved norm accuracy, and kept platform enjoyment similar to traditional feeds.

1 of 1

Story summary

Researchers ran an eight-week randomized experiment with 2,000 U.S. participants during the 2024 presidential election.
The engagement-based algorithm amplified intergroup and emotional (IME) toxic content and boosted moral outrage and political messaging versus a reverse-chronological feed.
That feed lowered prescriptive norm accuracy, raised perceived partisan animosity, and left users’ own engagement unchanged.
The diversified-extremity algorithm cut IME toxic exposure, improved norm accuracy, and kept platform enjoyment similar to traditional feeds.