Study Reveals Algorithmic Influence on Political Discourse During the 2024 U.S. Election

Researchers built two custom feed-ranking algorithms and randomly assigned 2,000 participants to use one of them for eight weeks spanning the 2024 U.S. presidential election. The “engagement-based” algorithm mimicked the ranking logic of major social-media platforms, while a “reverse-chronological” feed served as a control. A third “diversified extremity” algorithm was tested to limit the visibility of hyper-active, extreme users. Compared with the chronological baseline, the engagement-based feed amplified intergroup, moralized, and emotional (IME) content and toxic material, especially moral outrage and political messaging. It also reduced the accuracy of participants’ perceptions of prescriptive social norms and heightened perceived partisan animosity, though it did not change participants’ own engagement rates. The diversified extremity algorithm lowered IME and toxic exposure, improved norm-perception accuracy, and preserved platform enjoyment.

NatureAlgorithms Redesign to Fix Election Social Norms

Digital environments now dominate civic conversation, and algorithmic curation determines which political information users encounter. Prior literature has linked algorithm-driven amplification of emotionally charged content to increased polarization on platforms such as Twitter and Facebook. The present experiment provides controlled, empirical evidence of these dynamics during a high-stakes electoral cycle.

Algorithms Redesign to Fix Election Social Norms

• Participants: 2,000 U.S. adults recruited for an eight-week trial before and after the November 5, 2024 election.

Algorithms Redesign to Fix Election Social Norms

• Algorithms Tested:

Algorithms Redesign to Fix Election Social Norms

1. Engagement-based feed – prioritizes content with the highest interaction metrics.

Algorithms Redesign to Fix Election Social Norms

2. Reverse-chronological feed – orders posts by posting time.

Nature

3. Diversified extremity feed – down-weights content from users identified as extreme or hyper-active.

Nature

• 17 Sept 2024: Stage 1 protocol for the registered report accepted.

Nature

• Oct 2024 – Dec 2024: Eight-week data collection covering pre- and post-election periods.

Algorithms Redesign to Fix Election Social Norms

• 2026: Findings published in *Nature*.

Algorithms Redesign to Fix Election Social Norms

• Sample size: 2,000 participants.

Algorithms Redesign to Fix Election Social Norms

• Duration: 8 weeks surrounding the election.

Algorithms Redesign to Fix Election Social Norms

• Outcome patterns: Engagement-based feeds produced the largest relative increase in moral outrage and political content; diversified extremity feeds produced the largest relative decrease in IME and toxic exposure while maintaining user satisfaction.

Nature

The results suggest that algorithmic emphasis on engagement can distort users’ perception of normative political discourse and inflate partisan hostility without altering overt engagement behavior. Conversely, redesigns that limit extreme user influence can improve normative accuracy and sustain enjoyment, offering a potential pathway for platforms to mitigate polarization while preserving user experience.

Algorithms Redesign to Fix Election Social Norms

The study’s authors note that “engagement-based feeds amplified IME and toxic content relative to reverse-chronological feeds, with the largest increases in moral outrage and political content.” They also emphasize that “the diversified extremity algorithm reduced IME and toxic content exposure, improved prescriptive norm accuracy, yet maintained comparable platform enjoyment—suggesting that reducing the influence of extreme users can curb algorithmic distortions without diminishing user experience.”

NatureAlgorithms Redesign to Fix Election Social Norms

The investigation reports an “unexpected direction” in how engagement-based feeds altered prescriptive norm perception accuracy, indicating that the mechanisms linking exposure to normative judgments remain unclear. Additionally, the study acknowledges that long-term behavioral consequences of algorithmic exposure were not measured, leaving a gap for future longitudinal research.

Algorithms Redesign to Fix Election Social Norms

• “For the first time in history, civic discourse commonly occurs in digital environments in which algorithms influence exposure to social information1,2.” — Researchers, lead authors

Nature

• “We found that engagement-based feeds amplified IME and toxic content relative to reverse-chronological feeds, with the largest increases in moral outrage and political content.” — Researchers, lead authors

Nature

• “The diversified extremity algorithm reduced IME and toxic content exposure, improved prescriptive norm accuracy, yet maintained comparable platform enjoyment—suggesting that reducing the influence of extreme users can curb algorithmic distortions without diminishing user experience.” — Researchers, lead authors

Nature

• “These dynamics contribute to skewed social norm perceptions, where users inaccurately infer that extreme views represent a broader consensus.” — Researchers, lead authors

Algorithms Redesign to Fix Election Social Norms

The authors propose extending the experimental framework to assess long-term behavioral shifts, exploring additional algorithmic designs that balance content diversity with user satisfaction, and testing interventions across varied political contexts to generalize the findings.

Algorithms Redesign to Fix Election Social Norms