Sobre as publicações de hoje

Visão geral

https://nickmark.substack.com/p/a-ten-point-plan-to-fix-scientific?r=150p5i&utm_campaign=post&utm_medium=web&triedRedirect=true&open=false

Scientific publishing is among the strangest markets in the world. Researchers produce a scientific manuscript, submit it to a scientific journal, and pay to publish it. Other researchers review it for free. Universities then pay billions of dollars to access it. The companies in the middle (Elsevier, Springer, Wiley) make margins above 38%, rivaling even the most lucrative of tech companies. In return, these middle companies provide little more than a submission portal, some typesetting, and a brand name. The system was designed for a world of printing presses and physical journals. It has no business surviving in 2026.

The dysfunction runs deeper than cost. The structure of scientific publishing actively distorts the science itself: incentivizing sensational findings over careful ones, punishing negative results, creating years of unnecessary delay between discovery and dissemination, and concentrating gatekeeping power in a small number of editors whose decisions are almost entirely unaccountable. A medical treatment that could save lives may sit in a submission queue for eighteen months while its authors navigate sequential rejections.

Reform efforts have been halting and incomplete. Open access mandates transferred cost from universities to authors without reducing it. Preprint servers like arXiv and medRxiv helped, but remain disconnected from the formal review and publication process. SciHub has made accessing articles easier - by illegally bypassing paywalls - however, the authors and reviewers remain uncompensated.

No one has redesigned the system from scratch. This is proposal for how we could.

Thanks for reading! Subscribe for free to receive new posts and support my work.

Subscribe

The Problems

I. Sequential Submission Wastes Years

Authors submit to one journal at a time. Rejection triggers a new submission cycle, with new reviewers who have never seen the paper. A paper rejected by three journals before acceptance may have consumed years and a dozen reviewer-weeks, producing nothing but delay. A 2021 systematic review of biomedical publications found the mean time from submission to publication ranged from 91 to 639 days at different journals.

We need to speed up the submission/review process.

II. Publishers Extract Value Without Creating It

The labor (writing, reviewing, editing) is donated gratis by the research community. A 2020 study estimated that a typical researcher spends 100 hours per year on uncompensated peer review. In aggregate researchers contribute 100 million hours per year, totaling billions of dollars in uncompensated labor. Publishers charge institutions for access to this donated labor. APC (article processing charge) models added a second revenue stream: now authors also must pay to publish open access. Science charges authors $5,450 for open access, NEJM $11,690, Nature $12,850.

We need to compensate the people doing the work: authors and reviewers.

III. Reproducibility Is Optional

Most published papers do not include the underlying data or analysis code. Scientific data described as “available on request” is, in practice, frequently inaccessible, with studies indicating00141-X/abstract) that up to 93% of authors who use this phrase either do not respond or decline to share their data. Independent replication, where it is attempted, frequently fails. The incentive structure of journals rewards novelty over rigor; negative and null results have no viable path to publication.

We need to ensure that scientific data is available for re-analysis and that negative studies are published.

IV. Reviewer and Editorial Power Is Unaccountable

A desk rejection requires no justification and generates no record. Editors can suppress inconvenient findings, protect preferred paradigms, or simply neglect submissions indefinitely. None of this is visible. The latter, a “pocket veto” can delay resubmission for months. Reviewer accountability is similarly absent; a reviewer can accept an assignment and silently ignore it for months.

Fair review is also important; a 1991 article analyzed reviewers who peer reviewed more than 25 times and found that 1.5% of were assassins (consistently negative) whereas 1.2% were zealots (consistently positive). Neither assassins nor zealots make good reviewers.

Conflicts of interest are often hidden; a 2021 study30200-9/fulltext) found that 45% of authors do not disclose conflicts, and author COI are only typically elicited after acceptance. Editors and reviews may not be required to disclose COI.

We need to transparently hold editors and reviewers accountable.

V. Incentives Are Misaligned at Every Level

Impact factor rewards journals for publishing citable work, not important work. Researchers are evaluated on publication count and journal prestige, not on contribution to knowledge. Prolific reviewers receive no formal recognition. The system optimizes for metrics that are gameable proxies of quality (e.g. citation count) rather than quality itself.

We need a robust system that recognizes the quality of work done by both authors and reviewers.

VI. Format Balkanization Is a Hidden Tax

Each journal has its own reference format, word limits, figure specifications, and submission portal. Reformatting a rejected manuscript for resubmission is a significant time cost with zero scientific value added. A 2019 study estimated the average researcher spend 14 hours per manuscript, or 52 hours per year reformatting.

We need a modern system for efficient layout and formatting.

A Ten Point Plan to A Fix Scientific Journals

1. Simultaneous Submission to Multiple Journals

The overall acceptance rate for scientific articles is about 30%. This means that the typical manuscript is reviewed by 2-3 reviewers 3-4 times prior to publication, generating 6-12 reviews of the same work. The reviews are sequential, duplicative, and often subjective.

Instead imagine if authors submitted a manuscript once and designated up to ten journals in ranked order of preference. Peer review occurs once, by a single larger assembled panel (4-5 reviewers), with results shared across all the designated journals. Journals compete to accept the work; authors are not forced to guess which journal will find their work acceptable before a single reviewer has read it. This avoids “wasting” repeat peer reviews with each resubmission. It provides a less subjective assessment of the work because more reviews are obtained, so an outlier review is more obvious. It also enhances the overall efficiency of peer review because the typical paper gets just 4-5 reviews instead of 6-12.

Submitting to multiple journals at once creates a bias to action by editors. The author’s pre-specified rank list governs by default, but authors retain the right to override it if their top-ranked journal has not responded within six weeks of all other reviews being complete. This creates a structural incentive for editorial speed without mandating it.

2. Automatic Preprint Upload on Submission

Every submission is automatically deposited to a discipline-appropriate preprint server at the time of submission (e.g. arXiv, bioRxiv, etc). Science enters the public record the moment it is ready for scrutiny, not the moment a journal decides it is worthy. The preprint is versioned; the accepted manuscript supersedes it, but the record of submission date is permanent and citable.

3. Mandatory Data, Code, and Reproducibility Standards

Every submitted manuscript must include all data and analysis code necessary to reproduce its results. Every figure and table must link directly to the code that generated it. Limited exceptions apply (patient privacy, national security, proprietary clinical assays) but these must be explicitly declared and approved, not assumed.

This is the single most important structural change for the quality of the scientific record. Without it, peer review is evaluation of a summary rather than evaluation of the actual science.

4. Standardized Submission Formats

A small set of submission templates covers the major article types: randomized trial, observational study, methods paper, systematic review, case series, brief communication. All journals on the platform accept all standard formats. Layout is rendered from a single source file using LaTex. Authors format once for their article type, not once per journal submission. Discipline-specific consortia could define additional format variants within the standard framework but the overall goal is reduce the number of competing formats.

The submission, peer review, editorial workflow, and hosting all run on a single open-source software platform. A shared platform eliminates the redundant cost of every publisher maintaining proprietary submission portals and editorial systems, money currently extracted from institutional subscription fees. It also creates a powerful migration incentive: publishers who adopt the platform gain dramatically streamlined workflows, better submission tracking, and access to the shared reviewer pool. The software is a public good; improvements made by any consortium benefit all of them. Hosting costs are pooled, providing economies of scale.

5. Online-Only Publication

There is no print edition. This reduces cost, latency, and landfill waste. There are no page limits imposed by the cost of ink. Supplementary materials are not second-class content buried in inaccessible PDFs — they are part of the article. Figures are full resolution. Video and interactive data visualization are first-class outputs. Individual articles (or complete journals) can be printed on demand, but the era of printed and mailed science is over, saving hundreds of thousands of tons of paper waste.

6. University Subscription Fees Fund the Ecosystem

Institutions pay tiered access fees based on size and research output, replacing both journal subscription fees and APCs. This pool funds platform operations, editorial infrastructure, and reviewer compensation. Authors pay nothing to submit. Readers at subscribing institutions pay nothing to read. A public access tier makes all articles freely readable after twelve months. Reviewers can use accumulated credits to access articles.

7. A Credit System Compensates Reviewers and Authors

Subscription revenue is distributed through a structured credit system. Reviewers earn credits for completed reviews, weighted by quality ratings from editors and authors. Authors of highly impactful work earn additional credits through a revenue-sharing mechanism: for papers exceeding a field-normalized citation threshold (e.g. the top 1% within discipline, with self-citations excluded) a share of subscription revenue is allocated proportionally.

This rewards genuinely transformative work (a paper cited thousands of times by independent researchers) without rewarding incremental work in high-volume fields. An important design consequence: editors have a direct financial incentive not to miss a potentially important paper. The cost of a wrongful desk rejection is now visible and quantifiable.

8. Full Editorial and Reviewer Transparency

Every journal’s acceptance rate, desk-rejection rate, median time-to-decision, and reviewer completion rate are public, updated in real time. Every editor’s decision record is public; not the content of decisions, but the pattern: how often they desk-reject, how often accepted papers are later retracted, how often they passed on papers that subsequently became highly cited elsewhere.

Reviewers are similarly tracked: assignment acceptance rate, completion rate, time-to-review, and quality ratings (by editors). Chronic non-completers lose eligibility to earn review credits. There are no more “pocket vetoes.” A reviewer who accepts an assignment and misses the deadline is flagged; the editor must reassign within a defined window. Conversely, prolific and punctual reviewers get public recognition for the quality work that they do, with visible stats that can influence promotion.

In addition to the stats, the public profile for each editor/reviewer shows their conflicts of interest, which is updated continuously. The current publishing paradigm requires a COI statement from authors only, and only at the time of publication. In addition to ties to industry, the profile shows a timeline of grants, patents, etc providing a complete assessment of a reviewer’s potential conflicts. In this new system, everyone involved in the manuscript, including authors, reviewers, and editors, would have a profile that includes any and all COIs. This page could integrate with other databases such as NIH funding, clincaltrials.gov, and Sunshine Act, providing automatic updates and additional transparency.

9. A Structured Post-Publication Review Process

The historical failure of post-publication critique has been an incentive problem, not a cultural one.

Post-publication peer review is integrated into the platform, not bolted on. Registered researchers can submit formal methodological critiques, data re-analyses, or replication reports. These are reviewed, credited, and linked permanently to the original article. Authors have a defined right of response. Editors can escalate serious concerns to an independent panel empowered to recommend correction, retraction, or expressions of concern.

Post-publication review earns credits at rates comparable to pre-publication review, meaning that major methodological flaws that bypass peer review still get noted. With greater transparency, journals have a stronger incentive to address concerns about fraud and reproducibility.

10. Governance: A Consortium of Consortia

The platform is governed by a federation of discipline-specific consortia (e.g medicine and health sciences, physical sciences, life sciences, social sciences, etc) with subgroups for more specific topics (genetics, randomized controlled trials, topology, etc). Each consortium sets field-specific norms: format variants, citation thresholds for revenue sharing, reviewer qualification standards.

Above them sits a small central body, responsible only for platform infrastructure, cross-disciplinary standards (authorship, data sharing, retraction protocols), and antitrust protection. This central body has no editorial power. It cannot be acquired by a commercial publisher. Its charter requires supermajority approval from member consortia to amend.

This structure guarantees that the model can be flexible to changing needs, unlike the current system, while avoiding capture. There are precedents such as arXiv and ORCID.

How do we built this?

Change is difficult, and replacing an entrenched and ossified system like scientific publishing is especially hard. None of these 10 points is technically difficult, and several aspects (simultaneous submission, preprint servers, data repositories, open-source review software) already all exist in partial form. The obstacle is not engineering. It is the will do tear down the existing paid publication system and rebuild it thoughtfully and fairly. The question for the scientific community is whether the aggregate cost of the current system, measured in delayed treatments, un-replicable findings, wasted researcher hours, and institutional subscription fees, has finally exceeded the political cost of replacing it.

Thanks for reading! Subscribe for free to receive new posts and support my work.

Subscribe

Referências