Public story
A complete article written for an intelligent reader who has not already learned the field’s vocabulary.
- Human or practical opening
- Terms explained at first use
- Main findings and uncertainty
- Why the work matters
- What it does not show
THE TRANSDUCTIVE PUBLICATION FORMAT
One research object. Three reading depths.
The format does not ask a scientist to choose between accessibility and precision. It makes both compulsory — and links them so a reader can move deeper without starting over.
Proposal status: under active development within Transductive Science and publicly offered for consideration by researchers, funders, publishers, libraries, public institutions and citizen organizations. This is an invitation to evaluate the format, not a claim of external endorsement or adoption.
01
The layers are not a hierarchy of intelligence.
They correspond to how much detail a reader wants, not how much the reader deserves to know.
Informed explanation
For a reader who wants to understand the scientific reasoning without yet reading a specialist paper.
- Prior work and history
- Methods explained conceptually
- Competing explanations
- Uncertainty and limitations
- Inspectible figures and source links
Technical record
The exact record needed for expert review, reproduction, later reuse and formal citation.
- Methods, equations and parameters
- Data and code
- Full references and identifiers
- Version and correction history
- Review and challenge record
01A
Two choices, not one ladder
“How simply?” and “How deeply?” are different questions.
Evidence depth
How much of the work do you want to inspect?
Story → How we know → Exact work.
Language level
How should the same claim be explained?
Very simple (ELI5) → Clear (ELI12) → Young adult (ELI18) → General.
A reader can choose very simple language and still move all the way to methods and data. An expert can start with a simple explanation before checking equations. These controls must never imply an intelligence ranking.
LANGUAGE INVARIANT
Wording, examples and sentence length may change. Claims, numbers, uncertainty, scope and causal strength may not.
02
Every layer must stand on its own — and point to the others.
A reader should never hit a dead end merely because they started at the wrong depth.
The public article is not a teaser for the technical paper. It contains the full conclusion at its own level. The informed article is not a literature-review appendix. It gives the reader enough reasoning to identify where disagreement could enter. The technical record is not allowed to hide behind “see supplementary information” when an exact method, dataset or parameter can be linked directly.
The key design change is cross-layer addressability. A claim in the public article can link to the deeper explanation of that claim. The deeper explanation can link to a figure, method or dataset. A technical reviewer can link back upward to the public wording and flag a simplification that overstates the evidence.
Corrections propagate in the opposite direction. If an analysis changes, the technical record changes first, then every dependent statement in the informed and public layers must be reviewed.
INVARIANT
Different depth may change vocabulary and detail. It must not change the scientific meaning.
03
What the format refuses
Not everything inherited from academic publishing serves the reader.
| Common practice | Transductive rule | Reason |
|---|---|---|
| Abstract as public summary | Full public article | An abstract compresses for specialists; it does not teach the subject. |
| Jargon first, glossary later | Explain the thing before naming the term | The reader should understand the distinction before memorizing its professional label. |
| “Impact” written as promotion | Human stakes plus limits | Importance includes who is affected, what could change and what the result cannot justify. |
| Supplementary-material graveyard | Directly linked evidence | Data, code and detailed methods are part of the publication, not administrative leftovers. |
| One frozen paper | Versioned research object | Corrections, reviews, replication and new data should remain visibly attached to the record. |
| Public communication after publication | Public explanation required at publication | Explanation is part of completing the research, not marketing it afterward. |
04
AI changes the cost of explanation.
Earlier web-native journals discovered that great explanation could demand extraordinary editorial labor. That constraint is no longer fixed.
Distill’s editors described the cost directly: some early articles received more than fifty hours of editorial help, and producing high-quality interactive explanations required an unusual mixture of scientific, writing and design skill.
Generative systems change that production equation. They can draft alternative explanations, identify undefined terms, turn a technical figure into an annotated public figure, build interactive views, generate accessibility text, maintain links across versions and test whether two layers make inconsistent claims.
That does not make AI the authority. The researchers remain responsible for the scientific content. The important change is that translation between depths becomes cheap enough to require rather than admire.
A publication system can now run continuous checks: “this term appears in the public layer without explanation”; “this public claim has no technical anchor”; “the result changed in version 3 but the public story still uses the old number”; “this figure cannot be understood without color.”
What was once bespoke editorial craft can increasingly become ordinary publication infrastructure.
05
Different participants get different obligations.
Accessibility cannot be assigned to “communications” after everyone else has finished.
Explain before you compress.
Provide the public question, the deeper reasoning and the exact record. Confirm that the three layers say the same thing.
Review the translation too.
Technical correctness is not enough. Review whether the public and informed layers distort, oversell or omit material uncertainty.
Edit for traversal.
Make sure readers can move from a public claim to its evidence and back without hunting through unrelated files.
Fund intelligibility as output.
Treat public explanation as a deliverable of publicly supported research, not a discretionary outreach activity.
Preserve the whole object.
Archive layers, identifiers, code, data, versions and relationships — not only a final PDF.
Choose depth, not permission.
Start anywhere. Move deeper. Download evidence. Challenge a claim. No level is labeled as the “real” publication.
06
A publication should advertise how to challenge it.
Trust improves when the route to disagreement is part of the design.
Every Transductive publication should end its public and informed layers with a small, prominent set of routes: inspect the data, inspect the method, see known weaknesses, see corrections and versions, read reviews, and report a problem.
This changes the posture of publication. Instead of presenting a polished object that asks to be believed, the work exposes the places where belief can be tested.
The technical record can preserve the familiar academic apparatus — references, formal notation, detailed methods, statistical models — while adding stable anchors so criticism can point to an exact claim or parameter instead of vaguely “disagreeing with the paper.”
07
The format in one screen
A minimal publication can be simple.
Identity
Always visible
Title, authors/contributors, date, version, persistent identifier, funding, conflicts, review status and correction status.
Depth switch
Always visible
Read the story · Understand the science · Inspect the work. The switch keeps the reader on the corresponding claim or section where possible.
Evidence routes
Near claims
Figures, data, methods and references open in context rather than forcing the reader to lose their place.
Challenge routes
End of every layer
Known limitations, open questions, review, replication, corrections and a public issue/report mechanism.
Exports
Preservation
Accessible HTML is primary. Stable PDF and machine-readable exports remain available for archives, citation systems and offline use.
08
Reader testing changes the format
Accessibility is tested as a design problem, not declared by the authors.
The first presentation pass used a simulated focus group of eight reader profiles and a 100-turn discussion. The repeated result was that language complexity and evidence depth had been accidentally fused. This edition separates them.
The exercise also pushed question-based headings, a 20-second literal summary, visible objections, larger instructional text and direct routes from a public claim toward deeper evidence.
This was simulated design testing, not empirical human-subject research. It generates hypotheses for real reader testing rather than replacing it.