RenaSciQuestion network · The Renaissance of Science E2Flood and Dykes

E2Publishing crisisFurther topic

Flood and Dykes

Papers and Journals in the Information Age

?Why are ever more papers ever harder to read and to trust?

Topic page
~1,130 words · 7 screens
Deep read
~8,340 words · 117 sources
Questions
16 out · 14 in
Start with the first question ↓Go straight to the deep read ↗
  1. 01Are there too many papers?
  2. 02In the age of the web, why is a paper still paper?
  3. 03After open access, who pays the bill?
  4. 04What is a Nature paper worth?
  5. 05Why do the dykes crack?
  6. 06Can AI-written papers be detected? Should they be?
  7. 07Is raising the dykes enough?

◇The network re-centres on “Flood and Dykes”. Every screen leaves a few follow-up questions; open one to walk into another topic.

01/07Flood

Are there too many papers?

No shortage of papers – only of time to read them

In 1945 Vannevar Bush was already complaining of “a growing mountain of research”. Eighty years on, the mountain has become a flood: some 3.3 million science and engineering papers were published worldwide in 2022, and Elsevier alone received more than 4.2 million submissions in 2025. The digital age promised to set the paper free. Instead we got more papers – harder to read, and harder to trust.

So we keep having to raise the dykes: journal selection, peer review, impact factors, plagiarism checks and retractions, and now AI detectors too. The earlier ones each have their uses; whether the last one works at all is an open question. And still the water rises. Perhaps the problem lies not in the dykes but in the channel.

Follow-up questions
Read deeperAre there too many papers?↘
Asked from elsewhereMFloodE1BooksE6MapE7Burden

02/07Digital paper

In the age of the web, why is a paper still paper?

The web was made for scientists; the paper stayed paper

The World Wide Web was born at CERN in 1989, and it was meant from the outset to let scientists everywhere share information automatically. Yet PDF, which arrived in 1993, soon took centre stage in academic publishing: it reproduces the printed page exactly, keeping volume, issue and page numbers. To this day the most prominent button on a journal’s website reads “Download PDF”.

The paper’s genre, length and structure were carried onto the screen unchanged, and once published it is as frozen as ever: a mistake can only be fixed by issuing a separate correction. The sciences at least have preprints and code repositories; in adapting to the internet age, the humanities have, if anything, gone backwards. To put it bluntly: print journals should have closed long ago, and research should be deposited on open platforms. Journals that want to stay should become windows for certification, curation and display.

Follow-up questions
Read deeperIn the age of the web, why is a paper still paper?↘
Asked from elsewhereMThe paperE6Paper shadow

Jost Amman, The Printer. Woodcut, 1568. Wellcome Collection · Public domain

03/07Open

After open access, who pays the bill?

Circulation changed; the unit did not

arXiv, founded in 1991, now holds more than three million preprints; in the first ten months of the pandemic, over 30,000 COVID-19 papers appeared first as preprints. The Budapest Initiative of 2002, PLOS and Plan S tore down the paywall in front of readers. Others simply went round it: at one point Sci-Hub’s download requests ran into the tens of millions.

But a new wall went up in front of authors. Article processing charges paid to six big publishers rose from $910 million in 2019 to $2.54 billion in 2023; in 2026 the Chinese Academy of Sciences stopped paying for about 30 high-priced journals. Preprints, open access, open review: all of them changed who can read and who pays. The unit was always the paper.

Follow-up questions
Read deeperAfter open access, who pays the bill?↘
Asked from elsewhereE6Paper shadow

04/07Counting

What is a Nature paper worth?

When the measure becomes the target

Citation indexes were invented for finding the literature; later they became rulers for measuring people. Impact factors, h-indexes and university rankings piled on the pressure, layer upon layer, and the paper turned from a tool of communication into the currency of an academic career. In 2016 the average bonus Chinese universities paid for a single paper in Nature or Science was about $44,000. Goodhart warned long ago that a statistical regularity, once it is used for control, tends to collapse.

The “Five Onlys” reform, the rules on using SCI metrics properly, DORA, the Leiden Manifesto for Research Metrics and CoARA all try to correct course; but what they change is how things are counted, and what gets counted is still the paper. A metric’s proper job is only to aid judgement. Plagiarism checks are the same: plagiarism has only two states, copied or not copied; there is no such thing as a few per cent of it. Substituting metrics for judgement is, in the end, institutional laziness.

Follow-up questions
Read deeperHow did the measure become the target?↘
Asked from elsewhereE1CountingE4SqueezeE7Perish

05/07Cracks

Why do the dykes crack?

Three cracks in the dyke

The first is replication: when a hundred psychology studies were repeated, only 36% produced a significant result again. The second is review: in 2020 the world’s reviewers gave more than 100 million hours of unpaid work, yet it is ever harder to find anyone willing to accept an invitation; in one experiment, reviewers found on average just over two of the nine major errors planted in a paper. The third is fraud: paper mills have ridden the wave, and in 2023 retractions worldwide passed 10,000 for the first time – more than 8,000 of them from a single publisher.

A paper tells only the story that worked; failure has nowhere to go. And as long as a paper can be cashed in for bonuses, promotions and degrees, someone will mass-produce it. The arms race between detection and fraud has no end. It is like the melamine in China’s milk powder: once the test becomes a checkpoint anyone can run for themselves, what gets mixed in is no longer water.

Follow-up questions
Read deeperWhere does failed research go?↘
Asked from elsewhereE7Perish

John Martin, The Deluge. Mezzotint. Wellcome Collection · Public domain

06/07Sluice gate

Can AI-written papers be detected? Should they be?

A paper in fifty-four seconds

Less than a year after ChatGPT appeared, the words “Regenerate response” – the label on one of its buttons – were turning up in papers. One detection company’s model estimates that about 21% of the reviews at ICLR 2026 were written entirely by AI. ICML planted hidden cues in submissions that only large language models could read, to catch reviewers who had broken their pledge and handed whole manuscripts to an LLM. In February 2026 a researcher used a new tool to write a paper about a fictional experiment in fifty-four seconds.

When the cost of writing approaches zero, “being able to write it” no longer vouches for anything. The hidden cues catch broken promises, not bad reviews; scoring texts for their “AI rate” is both inaccurate and beside the point. The question to ask is not “Who wrote it?” but “Is it any good?” AI has lifted the average to fluent, orderly prose, and what that exposes is a shortage of judgement.

Follow-up questions
Read deeperCan AI-written papers be detected? Should they be?↘
Asked from elsewhereMFloodE5Flood

07/07Channels

Is raising the dykes enough?

Rather than raise the dykes, dig new channels

DOIs, ORCID, Registered Reports, open peer review, overlay journals, OpenAlex: all are real advances, yet all of them patch things up around “the paper”. In Chinese legend, Gun “dammed the flood” and after nine years had nothing to show for it; Yu “dredged the nine rivers”. Yu did not tear down all of his father’s works. What he changed was where the water went.

Faced with a new technology, we need neither reject it outright nor obey it mindlessly. The flood cannot be stopped, but it can be steered by working with the current – and the bold can even ride the crest of the wave. Today there are three channels to dig: give knowledge a new unit; make verification, replication and clear explanation count as achievements; and let the public in where research happens.

Follow-up questions
Read deeperIs the open-science toolbox enough?↘
Asked from elsewhereE5Shift

Rebuilding the regulating sluice of the Menufia canal in Egypt: the gantry for the new gates. Photograph. Wellcome Collection · Public domain

E2The discussion goes on

The discussion goes on

?Why are ever more papers ever harder to read and to trust?

More than thirty years ago the World Wide Web software was opened to the public, and PDF appeared in the same year. Digital networks promised to free the scientific paper from paper, postage and subscription fees. Today the paper really no longer needs paper – yet there are more papers than ever, they are harder to read, and they are harder to trust. This essay looks back over the information age. PDF carried an old genre onto the screen unchanged; preprints and open access changed circulation but not the unit called “the paper”; metrics, the replication crisis, the reviewing bottleneck, paper mills and generative AI have piled onto that unit, layer upon layer. We have built dyke after dyke. They help, and they have their limits.

Deep readFlood and Dykes: Papers and Journals in the Information Age~8,340 words · 11 questions · 117 sources↗
  1. 1Are there too many papers?
  2. 2In the age of the web, why is a paper still paper?
  3. 3After open access, who pays the bill?
  4. 4Where does the surge come from?
  5. 5How did the measure become the target?
  6. 6Where does failed research go?
  7. 7Who reviews, and can they keep up?
  8. 8Why can papers be mass-produced?
  9. 9Can AI-written papers be detected? Should they be?
  10. 10Is the open-science toolbox enough?
  11. 11Raise the dykes, or dig new channels?

→Questions from here

The follow-up questions left on this page, grouped by the topic they lead to.

←Questions from elsewhere

Questions asked in other topics that lead here. Open one to return to the screen where it was asked.