Back to the directoryFind in the timeline
The Meta 2020 US election studies on feeds
Guess and colleagues (2023) moved consenting Facebook and Instagram users to chronological feeds in 2020 and found no significant change in polarization over three months. Critics question Meta's role and a temporary Meta algorithm change during the study window.
- Where it sits
- Academic psychologyResearch published for other researchers and students.
- Evidence level
- ContestedThe evidence conflicts or is disputed, for example after failed replications.
- Organizations
- Facebook (Meta)An organization is listed where the sources concern it. The text says exactly how.
- Sources inspected
- 4 of 43 with the relevant text, company document or archive item checked, 1 as an abstract or reference article, 0 not verified.
What happened
In late July 2023 Science and Nature published the first four papers from the US 2020 Facebook and Instagram Election Study, a collaboration between Meta researchers and external academics, with Talia Stroud and Joshua Tucker leading the academic team. In the feed paper by Guess and colleagues, consenting users were moved from algorithmic to reverse-chronological feeds during the 2020 US election. Outcomes were measured over three months.
The psychological idea
The feed paper is an experiment on feed ordering. The other papers covered removing reshares, ideological segregation in news exposure (aggregated data on about 208 million US adult Facebook users) and reducing content from like-minded sources (in Nature). Meta drew stratified random samples for a separate account deactivation experiment, and survey invitations were placed in users' feeds.
In practice
In the feed paper, chronological feeds reduced time on the platforms, increased political and untrustworthy content on both platforms and, on Facebook, reduced uncivil content. They did not significantly change issue or affective polarization, political knowledge or other key attitudes over three months. In the like-minded sources paper, the median Facebook user got 50.4 per cent of content from like-minded sources and 14.7 per cent from cross-cutting sources, and reducing like-minded content had no measurable effect on attitudes.
Why it mattered
The papers are experiments on feed design carried out with a platform's cooperation, so they give measured evidence about one election period. Their reach is limited by the three-month window, the 2020 setting and the disputes below.
Limits and qualifications
Contested. The program says Meta could reject designs only for legal, privacy or feasibility reasons and could not restrict findings. Michael Wagner, the rapporteur, wrote that Meta set the agenda in ways that affected overall independence, while calling the studies rigorous. An eLetter co-authored by Filippo Menczer says the study ran from 24 September to 23 December 2020, while Meta used a temporary stricter news algorithm from November 2020 that cut misinformation views by at least 24 per cent. The study authors reject this, and Science will not ask for corrections.
Evidence behind the claims
Guess and colleagues, Science 381(6656): 398-404, moved consenting users to reverse-chronological feeds during the 2020 US election; time on the platforms fell, political and untrustworthy content rose, and issue polarization, affective polarization, political knowledge and other key attitudes did not significantly change over three months.
The first four papers were published on 27 July 2023 in Science and Nature; the program states that academics chose designs, Meta could reject them only for legal, privacy or logistical reasons, and Meta could not restrict findings.
In the Nature paper on like-minded sources, the median Facebook user got 50.4 per cent of content from like-minded sources and 14.7 per cent from cross-cutting sources, and reducing like-minded content had no measurable effect on attitudes.
Tech Policy Press reports that the segregation study covered 208 million US adult active Facebook users, that Meta drew stratified random samples for the deactivation experiment and placed survey invitations in users' feeds, and that Meta says neither it nor its researchers could restrict findings.
Michael Wagner, the program's rapporteur, wrote that Meta set the agenda in ways that affected the overall independence of the researchers, and also called the studies rigorous, transparent and ethical; Tech Policy Press adds that three months may be too short and users may still have seen ranked content through friends' posts.
A Science eLetter co-authored by Filippo Menczer says the study ran from 24 September to 23 December 2020 while Meta used temporary news feed changes from November 2020 that cut misinformation views by at least 24 per cent, with the standard algorithm returning in March 2021.
The original authors reject the eLetter's argument and say their findings are more limited and nuanced than often perceived; Science's editor-in-chief Holden Thorp said the journal will not ask for corrections but will make readers aware of the critiques.
People involved
- Andrew Guess. Lead author of the feed algorithm paper
- Talia Stroud. Co-led the external academic team
- Joshua Tucker. Co-led the external academic team
- Michael Wagner. Rapporteur who assessed the collaboration
- Filippo Menczer. Co-author of the Science eLetter critique
In the Psy101 course
- OpenStax Psychology 2e, section 2.3: Analyzing FindingsChapter 2, Psychological Research. Method. The section discusses this topic.The analyzing findings section covers experiments, random assignment and statistical significance, which this study relies on.
This is course relevance only. It is not evidence for the historical or corporate claims on this page.See the full course map.
Sources and inspection
4 of 4 sources were opened and read at the level shown. A source marked as not verified supports no claim on this page by itself.
- [1]How do social media feed algorithms affect attitudes and behavior in an election campaign?Guess, A; Malhotra, N; Pan, J; Barberá, P; et al.. Science 381(6656): 398-404 (repository record at University of Amsterdam), 2023. Journal article. Primary source.Abstract or publisher record onlyRead the University of Amsterdam repository record: authors, journal details (28 July 2023) and abstract. The Science paper itself was not read.
- [2]First Four Papers from US 2020 Facebook & Instagram Research Election Study Published in Science and NatureNYU Center for Social Media and Politics. NYU Center for Social Media and Politics, 2023. Report. Secondary account.Relevant text checkedRead the program's announcement as returned by the fetch tool: paper summaries, numbers and independence statements. These come from the study program, not an independent party.
- [3]A Primer on the Meta 2020 US Election Research StudiesTech Policy Press. Tech Policy Press, 2023. News report. Critique or replication.Relevant text checkedRead the article as returned by the fetch tool: Meta's role, Wagner's assessment, participant numbers and criticisms. Whether feed assignment was random was not stated.
- [4]Science eLetter: Social media algorithms can curb misinformation, but do they?Observatory on Social Media, Indiana University (post about an eLetter co-authored by F. Menczer). Observatory on Social Media, Indiana University, 2024. News report. Critique or replication.Relevant text checkedRead the post by the critics' own institution as returned by the fetch tool. It is a partisan summary; the eLetter on science.org was not opened.
Access note: Science and Nature pages were not opened. Whether the feed assignment was random was not stated in the sources read. The eLetter was known only through a post by its authors' institution.
Independent source review: Claude review agent, independent of the research agent, 2026-10-21. Verdict: pass-with-edits.