PaperlayerView exact revision on arXiv

Source-linked AI summary

Tracing high-profile attention to questionable research as a case for funder due diligence

Federica Silvi, Leslie D. McIntosh

arXiv:2608.30613v1cs.DLcs.SI

TL;DR

Evidence on whether questionable research reaches policy, patents and clinical guidance, and whether its authors continue publishing or receiving funding after exposure, is limited. This study examines a known authorship-for-sale network and finds that its research informed these domains and its authors benefited from public grants.

  • Problem

    Evidence on whether authorship-for-sale research reaches policy, patents and clinical guidance, and whether its authors retain funding and publishing after exposure, is limited.

    Source: Introduction · p. 5, §2 · p. 5

  • Method

    The study uses a known authorship-for-sale network to examine research citations in policy, patents and clinical guidelines, and authors’ publishing and funding after exposure.

    Source: Introduction · p. 5, §2 · p. 5

  • Results

    PNN-linked research had informed policy documents, patents and clinical guidelines before exposure, and its authors had benefited from legitimate public grants.

    Source: Discussion · p. 22

  • Takeaways & Limitations

    The paper argues that funders should consider authorship and network structure in holistic due diligence alongside citation counts and attention data.

    Source: Discussion · p. 20

    The analysis examines one documented network in depth and covers only a small fraction of the activity possible at this scale, rather than a representative sample of paper-mill activity.

    Source: Discussion · p. 22

Abstract

from arXiv · show

When questionable research has entered the scientific record, it stands the same chance as other literature of shaping national and international policy, clinical guidance, and research and development activities. This study uses a known authorship-for-sale network identified in August 2022 to examine whether questionable research influences policy, patents, and clinical guidelines, and whether its authors continue to secure funding and publish after exposure. Across nearly 2,000 publications in our dataset, 57 were cited in policy documents, 12 in clinical guidelines, and 480 in patents. Nearly a quarter of these publications are linked to one or more grants: funding that could otherwise have supported rigorous, ethical science. Among the 278 authors we traced, publishing and funding continued well after the PNN's public exposure in 2022: over 90% continued publishing, and 23% were linked to a grant. As our findings show, the participants of paper mills and authorship-for-sale networks can still inform policy and clinical practice and support their authors' career progression even after the questionable practices behind them are exposed. We present a case for funders to consider authorship and network structure as part of a holistic due diligence process, alongside metrics such as citation counts and attention data.

Introduction

Questionable research can retain influence after exposure, while research metrics may miss or distort that influence. This introduction motivates funder due diligence that considers attention alongside authors, networks, and their funding and publication activity.

Introduction

  • Paper mills sell authorship on pre-written manuscripts, and their activity has outpaced legitimate science since 2016, with over 20,000 positions advertised [7].Source: Introduction · pp. 2–3
  • 1 review in 2015 versus 105 in 2023 cited retracted paper-mill articles; the trend is likely understated because retraction lags publication and citation.Source: Introduction · p. 2
  • A paper’s first organic citation can take years, influential work may remain uncited, journal impact factors are often incorrectly applied to individual articles, and citation metrics can be gamed [12, 13, 14].Source: Introduction · p. 3
  • Policy, patent, or guideline mentions can strengthen confidence in authors’ work, yet due diligence rarely tests whether attention reflects genuine influence rather than volume or coordination.The introduction proposes looking beyond cited documents to authors’ bodies of work, collaboration networks, and funding and publication activity over time.Source: Introduction · pp. 4–5
  • The study uses a known authorship-for-sale network to ask whether questionable research receives attention in policy documents and patents.This is framed as RQ1; the introduction also identifies policy documents, clinical guidelines, and patents as sources of research attention.Source: Introduction · pp. 4–5

2. RQ2: Are its authors receiving legitimate grant funding?

The supplied passages introduce RQ3: whether authors continue publishing and receiving funding after their network or questionable research practices are exposed.

2. RQ2: Are its authors receiving legitimate grant funding?

  • The text frames RQ3 as asking whether authors continue publishing and receiving funding after exposure.The passages provide the research question but no findings answering it.Source: §2 · p. 5

Methods

The study uses a forensic scientometric case-study approach to examine the PNN through linked publication, attention, grant, and author-history data, with distinct datasets for its research questions. It assesses mentions, grants, post-exposure author activity, and retraction status using descriptive statistics without inferential tests.

Methods

  • The study is a forensic scientometric case study of the Pharmakon Neuroscience Network (PNN), publicly identified in August 2022 [26, 27] and previously described as over 300 authors across 40 countries [28].Source: Methods · p. 5

Data sources

  • The study combined Dimensions publication metadata, researcher profiles, and grant records [29], Altmetric attention data principally from policy documents and clinical guidelines [30], Policy Commons classifications [31], and retraction status from Retraction Watch [32] and PubMed's E-utilities API as fallback [33].Source: Data sources · pp. 5–6

Dataset construction

  • The study used three data collections for its three research questions: RQ1 drew on publication records of 16 authors linked to the PNN, RQ2 used RQ1’s combined population, and RQ3 used 278 unique authors.For RQ1, publications were restricted to those with 25 or fewer authors.Source: Dataset construction · pp. 6–7

Policy, clinical guideline and patent mention curation

  • Policy and guideline mentions were checked against citing-document text where accessible, deduplicated, and manually assessed for sentiment; policy organisations and guideline issuers were classified.The sentiment criteria adapted a framework developed for social-media mentions [34].Source: Policy, clinical guideline and patent… · pp. 8–9
  • All clinical guideline mentions were retained, while 28 policy mentions were excluded for inaccessibility, duplication, or false positives; patent mentions were not manually verified.Patent counts were treated as a quantitative measure of tracked attention, and sentiment was assessed by a single reviewer using criteria adapted from social-media classification.Source: Policy, clinical guideline and patent… · pp. 8–9

Grant linkage

  • Grant links came from Dimensions’ supporting_grant_ids field, which cannot reliably distinguish grants authors held from grants linked through resulting publications.Accordingly, the study describes grants as authors “held or were linked to”; grants starting in 2022 could not be placed definitively before or after the 24 August cutoff.Source: Grant linkage · p. 10

Post-exposure activity per author

  • For each author in the deposited dataset, the researchers queried Dimensions for complete publication and grant histories, including outputs beyond PNN-named or Altmetric-tracked records.Source: Post-exposure activity per author · p. 10
  • For logged post-exposure records, the study tracked grant start years, URLs and counts per author, plus publication counts and Dimensions IDs.The records were dated after the exposure cutoff.Source: Post-exposure activity per author · p. 10

Retraction status determination

  • Retraction status was checked first by DOI lookup in Retraction Watch, then by a PubMed E-utilities query for a “Retracted Publication” record when no Retraction Watch record was found.Statuses were categorized as “retracted,” “not retracted,” or “no retraction on record.”Source: Retraction status determination · pp. 10–11
  • “No retraction on record” indicates no evidence in the two sources, not confirmed absence of retraction; Retraction Watch records only retracted publications.The passage also states that “not retracted” statuses were confirmed only via a source, but the supplied text cuts off before identifying it.Source: Retraction status determination · p. 11

Statistical approach

  • The study reports descriptive counts and proportions, with no inferential statistical tests or claims of statistical significance.Source: Statistical approach · p. 11

Results

The Results examine policy, clinical-guideline, and patent attention to network-linked publications, their links to grants, and case studies of policy and clinical attention. They also trace authors’ publishing and grant activity after the PNN became public and report the retraction status of papers cited in policy or guidelines.

Questionable research and real-world attention

  • Within this set, 57 publications received at least one policy mention (76 mentions); 12 received at least one clinical-guideline mention (22 mentions); and 480 were mentioned in patents (2,623 patent mentions).Source: Questionable research and real-world… · p. 11
  • Policy citations were concentrated among intergovernmental and government organisations, while guideline citations spanned societies across 13 clinical areas.The guidelines included internationally recognised medical societies and national and specialty societies; a guideline citation marks entry into a practice recommendation.Source: Questionable research and real-world… · pp. 12–13

Attention in practice: three researcher case studies

  • Researcher 244 accounted for 18 of 22 clinical-guideline mentions; 12 referenced a single 2013 publication, itself a clinical guideline, cited by major medical associations as recently as 2024.In 2024, two US government documents by CMS and HHS mentioned a different grant-funded paper that ORI found involved in a research-misconduct case.Source: Attention in practice: three… · p. 14
  • A CADTH policy document cited one paper co-authored by three examined researchers, illustrating how one publication can add to multiple network members’ profiles [37].Two co-authors had funding activity investigated in 2025; the third was the author whose publication activity first prompted the PNN’s public exposure.Source: Attention in practice: three… · pp. 15–16

Grant funding linked to questionable research

  • 450 publications (22.8%) had at least one grant linked in Dimensions, including 129 publications linked to grants in the 2025 investigation.Those 129 represented 28.7% of grant-funded publications; all policy-cited and 41 of 43 patent-cited publications predated the PNN’s exposure.Source: Grant funding linked to questionable… · p. 16
  • Between the paper’s 2013 publication and the 2024 misconduct report, Researcher 244 was linked to six grants; funding continued after the Pharmakon network’s 2022 exposure and throughout the period the paper sat uncorrected.Source: Grant funding linked to questionable… · p. 17

Publishing and funding after exposure

Discussion

PNN-linked research informed policy, patents, and clinical guidelines, and its authors received public grants; neither pattern was disrupted after exposure. The discussion argues that funders should look beyond attention and citation metrics to research integrity and authorship networks.

Discussion

  • Paper mill activity is estimated to account for more than 1.5% of research literature, representing tens of thousands of papers at current global publication volumes [38], while detection is fragmented [2, 40].Source: Discussion · p. 20
  • Funders have a distinct integrity role because impact assessments rely on policy, clinical, and patent attention without sufficient checks on research or authorship legitimacy, while finite grants to unreliable work displace other research.A funder's acknowledgement also ties its name to work that may later prove unreliable.Source: Discussion · pp. 20–22
  • Grant acknowledgements can link funders to outputs and reputational risk, but without consistently recorded, auditable grant and ethics identifiers, funders and institutions cannot verify which outputs warrant scrutiny [27].The paper argues that identifiers should be recorded in structured, auditable form.Source: Discussion · p. 21
  • Funders are encouraged to require integrity commitments, include integrity clauses in grant contracts, and ensure breaches are promptly reported rather than resolved without their knowledge [25].The discussion presents these steps as existing guidance for closing the due-diligence gap.Source: Discussion · p. 21
  • The PNN case is a single documented network and an in-depth analysis of only a small fraction of what an enterprise at this scale can produce, not a representative sample of paper-mill activity.The discussion also notes limitations including Altmetric data coverage.Source: Discussion · p. 22
  • Only one publisher was directly notified by LDM about one author's problematic practices, and exposure through reporting avenues did not ensure publishers knew of the reporting or authors' questionable practices.This limits what publishers' awareness can be inferred from public exposure alone.Source: Discussion · p. 22
  • PNN-linked research informed policy, patents, and clinical guidelines before exposure, while its authors received legitimate public grants and neither pattern was disrupted afterward.Exposure through retractions and misconduct findings did not reliably restrict funding or attention.Source: Discussion · p. 22
  • Citation indicators can accrue slowly, be misapplied, or be manipulated, while citation volume and prestige cannot establish that attention is warranted or authorship legitimate.The discussion recommends examining who appears in an author's network and how that network behaved under scrutiny; attention metrics alone cannot resolve the issue.Source: Discussion · pp. 22–23

Declaration of generative AI use

The authors disclosed using Claude for coding, data exploration, and language editing, and said the resulting code was independently validated and AI-assisted content verified.

Declaration of generative AI use

Data availability statement

The study’s curated data and supplementary methodological details, along with the analysis code, are available through Figshare. The deposited data use anonymised author labels, while the full named dataset is available from the corresponding author on reasonable request.

Data availability statement

  • The study dataset, including curated Altmetric mentions, Dimensions publication and grant metadata, retraction-status determinations, and supplementary methodological detail, is available at Figshare [44].The deposited author names are replaced with anonymised labels, but Dimensions IDs are retained; those IDs resolve to public researcher profiles, so researchers remain identifiable through cross-referencing.Source: Data availability statement · p. 24
  • The full dataset with author names, used to develop and test the accompanying code, is available from the corresponding author upon reasonable request.Source: Data availability statement · p. 24
  • The Python code used to construct the datasets and produce the reported results is available at Figshare [45].Source: Data availability statement · p. 24
  • The Retraction Watch Database snapshot, downloaded 14 April 2026, is available from retractiondatabase.org subject to Retraction Watch’s terms of use [32].Source: Data availability statement · pp. 24–25

Data usage statement

The study’s data came from five named sources and was accessed and used in accordance with each platform’s terms and data usage policies.

  • The study data came from Dimensions, Altmetric, Policy Commons, the Retraction Watch Database, and PubMed.The Methods and Supplementary Information provide further details.Source: Data usage statement · p. 25
  • Further details about the data sources are provided in the Methods and Supplementary Information.Source: Data usage statement · p. 25
  • The authors state that data from each source was accessed and used in accordance with each platform’s terms of use and data usage policies.Source: Data usage statement · p. 25

Funding

The research reported in this article received no funding.

  • No funding was obtained for the research reported in this article.Source: Funding · p. 25

Declaration of interest statement

Two authors are Digital Science employees, and the study used its commercial Altmetric and Dimensions platforms through employee licenses; the company had no role in the study, which was conducted independently. LDM also serves as VeriMe’s volunteer, unpaid Pre-Launch Advisor.

Figure captions

The figure captions map the datasets used to trace attention and funding and describe how policy mentions and grant linkage are presented.

Figure captions

  • The diagram distinguishes 16 “2025 investigation authors” with 1,975 Altmetric-tracked publications from 278 co-authors with 12,508 post-exposure Dimensions publications.The two datasets support Research Questions 1 and 2, and Research Question 3, respectively.Source: Figure 1 · p. 32
  • 76 policy mentions were tracked across 27 organisations, ranked by mention count, with source-category breakdown shown in Figure 1.Source: Figure captions · p. 33
  • The funnel chart traces grant-linkage identification from 1,975 publications through grant-linked publications, then breaks them down by attention type and timing relative to 24 August 2022.The caption specifies patents and policy documents as attention types and refers to grants connected to a prior investigation in 2025.Source: Figure 4
Loading 2608.30613v1…