Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 61 of the 63 references it cites were found in a registry or an index; 13 of the 61 sources its claims cite were checked for support.

Dynamics of cross-platform attention to retracted papers

Open access
Peng, Hao, Romero, Daniel M., Horvát, Emőke-Ágnes · 2022 · Proceedings of the National Academy of Sciences

Abstract

Retracted papers often circulate widely on social media, digital news, and other websites before their official retraction. The spread of potentially inaccurate or misleading results from retracted papers can harm the scientific community and the public. Here, we quantify the amount and type of attention 3,851 retracted papers received over time in different online platforms. Comparing with a set of nonretracted control papers from the same journals with similar publication year, number of coauthors, and author impact, we show that retracted papers receive more attention after publication not only on social media but also, on heavily curated platforms, such as news outlets and knowledge repositories, amplifying the negative impact on the public. At the same time, we find that posts on Twitter tend to express more criticism about retracted than about control papers, suggesting that criticism-expressing tweets could contain factual information about problematic papers. Most importantly, around the time they are retracted, papers generate discussions that are primarily about the retraction incident rather than about research findings, showing that by this point, papers have exhausted attention to their results and highlighting the limited effect of retractions. Our findings reveal the extent to which retracted papers are discussed on different online platforms and identify at scale audience criticism toward them. In this context, we show that retraction is not an effective tool to reduce online attention to problematic papers.

Overview

CoChat Summary

TL;DR

Retracted papers attract more online attention than matched nonretracted papers across social media, news, blogs, and knowledge repositories, but most have already exhausted their attention by the time they are retracted. Retraction therefore does little to curb the online spread of problematic findings, although Twitter discussions before retraction contain more criticism of subsequently retracted papers.

Key findings

  • Across 2,830 retracted and 13,599 control papers with at least a 6-mo tracking window, retracted papers received more mentions after publication on all four platform types and in top news outlets; the surplus was higher on curated platforms such as news outlets and knowledge repositories than on social media and blogs.
  • By the time of retraction, attention was largely exhausted: 80.2% of retracted papers received no mentions in the 2 mo before retraction, and 93.6% received no mentions in the last month before retraction.
  • After excluding critical tweets, retracted papers still received more Twitter mentions than control papers, indicating uncritical sharing of potentially flawed findings.
  • Retracted papers received a higher fraction of critical tweets than control papers before retraction, especially over longer windows such as 5–6 mo, with regression coefficients for retraction status positive and statistically significant (P < 0.001 for 5–6 mo).

Methods

  • Used Altmetric (October 8, 2019 version) to track mentions of papers across four platform types—social media, news media, blogs, and knowledge repositories—and Retraction Watch for retracted papers; the analysis covered 3,985 retracted papers on 14 online platforms.
  • Matched five control papers to each retracted paper using coarsened exact matching on publication journal, publication year within ±2 y, author count within ±2, and most-highly-cited author’s citation rank percentile within 20%; this yielded 19,255 control papers matched to 3,851 retracted papers.
  • Compared attention using negative binomial regression, Mann–Whitney U tests, and interrupted time series analysis; labeled critical tweets using a 76-keyword heuristic with recall 0.9 followed by manual annotation by three to four annotators, achieving weighted average Cohen’s kappa of 0.72.

Limitations

  • The paper states that not all problematic papers are retracted, and findings are based only on the subset recorded in Altmetric, which may not represent all problematic papers.
  • The paper states that Altmetric may not cover all mentions, and that the matching scheme leaves factors such as prestige of affiliation uncharted.
  • The paper states that critical-mention detection focuses on Twitter only, and that the study analyzes public attention without distinguishing different online audiences; the text also notes that after-retraction filtering was not applied to all platforms, so mention volume after excluding retraction-related content is an overestimation.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1073/pnas.2119086119

Cites

61 of 61 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1. Illustration of the research process that compares the online attention received by retracted and control papers. (A) We match five control papers to each retracted paper using the Altmetric database. (B and C) We track the change in mentions over time on four different types of platforms and in top news outlets. As an example, we show here the retracted paper “Effect of a program combining transitional care and long-term self-management support on outcomes of hospitalized patients with chronic obstructive pulmonary disease: A randomized clinical trial” published in JAMA (DOI: 10.1001/jama.2018.17933) and one of its matched control papers: “Vitamin D, calcium, or combined supplementation for the primary prevention of fractures in community-dwelling adults: US Preventive Services Task Force Recommendation Statement” (DOI: 10.1001/jama.2018.3185). (D) We compute the average cumulative number of mentions across all platforms within 6 mo after publication (and before retraction) for all retracted and control papers in the dataset. (E) Similarly, we compute the average cumulative number of mentions within 6 mo after retraction. Error bars indicate 95% CIs.
Fig. 1. Illustration of the research process that compares the online attention received by retracted and control papers. (A) We match five control papers to each retracted paper using the Altmetric database. (B and C) We track the change in mentions over time on four different types of platforms and in top news outlets. As an example, we show here the retracted paper “Effect of a program combining transitional care and long-term self-management support on outcomes of hospitalized patients with chronic obstructive pulmonary disease: A randomized clinical trial” published in JAMA (DOI: 10.1001/jama.2018.17933) and one of its matched control papers: “Vitamin D, calcium, or combined supplementation for the primary prevention of fractures in community-dwelling adults: US Preventive Services Task Force Recommendation Statement” (DOI: 10.1001/jama.2018.3185). (D) We compute the average cumulative number of mentions across all platforms within 6 mo after publication (and before retraction) for all retracted and control papers in the dataset. (E) Similarly, we compute the average cumulative number of mentions within 6 mo after retraction. Error bars indicate 95% CIs.CC BY-NC-ND · PubMed Central
Fig. 2. After publication and before retraction mentions. (A–E) Average cumulative number of mentions received within 6 mo after publication on four types of platforms and in top news outlets for both retracted and control papers. (F) Average cumulative number of mentions on Twitter after excluding critical tweets. Comparisons across different types of platforms show that retracted papers receive more attention after publication than nonretracted papers. Error bars indicate 95% CIs.
Fig. 2. After publication and before retraction mentions. (A–E) Average cumulative number of mentions received within 6 mo after publication on four types of platforms and in top news outlets for both retracted and control papers. (F) Average cumulative number of mentions on Twitter after excluding critical tweets. Comparisons across different types of platforms show that retracted papers receive more attention after publication than nonretracted papers. Error bars indicate 95% CIs.CC BY-NC-ND · PubMed Central
Fig. 3. Average cumulative fraction of critical tweets within 6 mo after publication. Control papers are selected from a matching process that considers the publication year, publication venue, number of authors, and authors’ citation count. We focused on papers that have at least one tweet mention in each time window. Retracted papers receive a higher fraction of critical tweets.
Fig. 3. Average cumulative fraction of critical tweets within 6 mo after publication. Control papers are selected from a matching process that considers the publication year, publication venue, number of authors, and authors’ citation count. We focused on papers that have at least one tweet mention in each time window. Retracted papers receive a higher fraction of critical tweets.CC BY-NC-ND · PubMed Central
Fig. 4. Average number of weekly mentions within 2 mo before and after the retraction. Trends are shown on four types of platforms and in top news outlets for data that exclude posts containing the phrase retract during the whole time period for both control and retracted papers. In the case of retracted papers, we also manually excluded all after-retraction posts that discussed the retraction in some form. Comparisons across different types of platforms show that retracted papers do not receive statistically more mentions than control papers immediately before or after retraction. Error bars indicate 95% CIs.
Fig. 4. Average number of weekly mentions within 2 mo before and after the retraction. Trends are shown on four types of platforms and in top news outlets for data that exclude posts containing the phrase retract during the whole time period for both control and retracted papers. In the case of retracted papers, we also manually excluded all after-retraction posts that discussed the retraction in some form. Comparisons across different types of platforms show that retracted papers do not receive statistically more mentions than control papers immediately before or after retraction. Error bars indicate 95% CIs.CC BY-NC-ND · PubMed Central
Fig. 5. Average fraction of mentions in the last month before retraction. The x axis represents different time windows between publication and retraction. The green bar plot (left y axis) shows the number of papers in each time window, excluding those with zero total mentions between publication and retraction. The black line plot (right y axis) shows the average fraction of mentions in the last month before retraction. Error bars indicate 95% CIs. (A) Data based on mentions in all platforms. (B) Similar results for uncritical mentions on Twitter.
Fig. 5. Average fraction of mentions in the last month before retraction. The x axis represents different time windows between publication and retraction. The green bar plot (left y axis) shows the number of papers in each time window, excluding those with zero total mentions between publication and retraction. The black line plot (right y axis) shows the average fraction of mentions in the last month before retraction. Error bars indicate 95% CIs. (A) Data based on mentions in all platforms. (B) Similar results for uncritical mentions on Twitter.CC BY-NC-ND · PubMed Central

Verification

Claims the paper makes

8 Valid · 1 Not valid · 21 Not checked · 66 Own results and interpretations

36 more claims · 1 with their cited sources checked

Create a free account to see all 96 claims and to verify the 48 references still open.

See all claims — free account
  • Probably not a substantial claim. 48
  • It may state two separate claims as one. 16

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

5 retracted

Show all 63 references

Research works printed but not linked: 1

  1. 10.1038/478026a · DOI · Paper page

Cited by

See all 57

Metadata

Authors
Peng, H., Romero, D. M., & Horvát, E.-Á.
Affiliations
  • University of Michigan (US)
  • Northwestern University (US)
Corresponding author
  • Romero, Daniel M.
  • Horvát, Emőke-Ágnes
Venue
Proceedings of the National Academy of Sciences
Published
2022-06-14
Volume
119
Issue
25
Pages
e2119086119-e2119086119
ISSN
  • 0027-8424 (ISSN-L)
  • 1091-6490
OpenAlex
W4282823329
Cited by
57
References
63
Citations by year
2026: 12 · 2025: 19 · 2024: 13 · 2023: 11 · 2022: 2
FWCI
9.41
Type
journal article
Language
en
Open access
Locations
  • Proceedings of the National Academy of Sciences (publishedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • arXiv (Cornell University)
Funders
  • National Science Foundation (IIS-1943506)
  • DOD | USAF | AMC | Air Force Office of Scientific Research (FA9550-19-1-0029)
Topics
  • Social Sciences › Library and Information Sciences › Academic Publishing and Open Access
  • Social Sciences › Library and Information Sciences › scientometrics and bibliometrics research
  • Social Sciences › Education › Academic integrity and plagiarism
Keywords
retracted article, scientific retraction, online attention, social media, Twitter, news organizations, knowledge repository, misinformation, public understanding of science, audience criticism

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-04, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-04 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Dynamics of cross-platform attention to retracted papers. https://app.cochat.ai/papers/10.1073/pnas.2119086119
The paper
Peng, H., Romero, D. M., & Horvát, E.-Á. (2022). Dynamics of cross-platform attention to retracted papers. Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.2119086119

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.