The Peer-reviewed seal
Anyone can publish on Cabdell, and an article is readable the moment it is published. The Peer-reviewed seal tells reviewed work apart: an article holds it while enough qualified reviewers recommend it. This page gives the exact rules and values. For what to do as an author or a reviewer, see Get the Peer-reviewed seal and Review an article.
Why a seal, and not a gate
Section titled “Why a seal, and not a gate”The question was whether Cabdell should publish only what had passed a minimum of reviews. It does not, for three reasons:
- No gatekeeper. Anyone can take part from the first day. Making publication wait for reviewers would bring back the gate that Cabdell exists to remove.
- Readers still choose. The seal appears next to every article in lists, and search can show only sealed articles. A reader who wants reviewed work gets exactly that; a reader who wants the newest preprints gets them too.
- The rule is public and computed. No editor decides. Anyone can read the rules below and check the result for any article.
So publishing stays free and immediate, and the contract (the program on the blockchain) is unchanged. The status “closed version” (“final” on the blockchain) only stops new versions: an author can close an article without any review, and closing never gives or takes the seal.
When an article holds the seal
Section titled “When an article holds the seal”An article holds the seal while all of these are true, counting only the reviews that count:
- at least 2 of them are favourable: they recommend accept or minor revision;
- there are more favourable ones than unfavourable ones (major revision or reject);
- every author of the article (the submitter and the confirmed co-authors) has a verified ORCID iD (rule 3, D-136). The seal starts when the last of them is verified, if that comes after the reviews.
Rule 3 closes the last gap of the conflict checks: with an author’s verified iD, the OpenAlex co-authorship and the same-employer checks run against every reviewer. Co-authors declared but not confirmed are not authors yet and are not required.
The states
Section titled “The states”| Label in the web app | When |
|---|---|
| Peer-reviewed | at least 2 favourable reviews that count, more favourable than unfavourable, and every author with a verified ORCID iD |
| Seal awaits ORCID iDs | the reviews are enough, but an author has no verified ORCID iD yet (the panel names who) |
| Peer-reviewed, suspended | it meets the rule above, but the article is disputed |
| Changes requested | at least 2 unfavourable reviews that count, as many as the favourable ones or more |
| In peer review | it has reviews that count, but neither of the cases above |
| Not peer-reviewed yet | no review counts yet |
A retracted article holds no seal at all, whatever its reviews said.
When a review counts
Section titled “When a review counts”The indexer checks every review against these rules in this order. The first rule that fails is the reason shown on the article page, after “does not count:”. A review that passes them all “counts”.
| Order | Rule | The review does not count when | Shown as |
|---|---|---|---|
| 1 | Not an author | its reviewer is an author of the article | written by an author of the article |
| 2 | Verified ORCID iD | its reviewer has no verified ORCID iD | the reviewer has no verified ORCID iD |
| 3 | Expertise | its reviewer has neither reputation 10 in the article’s primary or secondary field (counted as below), nor 3 works in the article’s area on OpenAlex | the reviewer has neither 10 reputation in the article’s fields nor 3 works in its area on OpenAlex; or: most of it came from the article’s authors; or: it is less than 90 days old; or: it came from fewer than 3 people (who count as below) |
| 4 | No conflict: ORCID | the reviewer and an author have the same verified ORCID iD | the reviewer and an author have the same ORCID iD |
| 5 | No conflict: Cabdell | the reviewer and an author wrote another Cabdell article together, published within 3 years of the review | the reviewer and an author wrote article #N together (year), less than 3 years from the review |
| 6 | No conflict: OpenAlex | OpenAlex lists works by both of them in the co-authorship window (below) | the reviewer and an author have works together on OpenAlex within 3 years of the review |
| 7 | No conflict: institution | their public ORCID records name the same current employer (below) | the reviewer and an author work at the same institution, as their ORCID records say |
| 8 | No conflict: support | an author voted for the reviewer’s article, marked their review useful, voted for their comment or resolved it, in the 3 years before the review | an author supported the reviewer’s work less than 3 years before |
| 9 | Text read | its text has not been read from IPFS yet | its text has not been read from IPFS yet |
| 10 | Length | its text has fewer than 200 words | shorter than 200 words (the count) |
| 11 | Reciprocity, for favourable reviews only | an author of the article gave the reviewer a favourable review in the 2 years before | an author gave the reviewer a favourable review (article #N) less than 2 years before |
| 12 | One person, one review | another address of the same ORCID iD already has a review that counts on this article | this person’s review from another address already counts |
“Author” means the submitter or a confirmed co-author. “A person” means every address of one verified ORCID iD: the conflict checks look at all of a person’s addresses. The contract already refuses a review from a confirmed author, so rule 1 is a safeguard. Rule 2 is the second version of the rules (D-133, decided by the owner): without the reviewer’s verified ORCID iD, the conflict checks on OpenAlex and on employers could not run. Anyone can still review without one, and earn reputation when others find the review useful; the review just does not count toward the seal. Rules 3 (as counted below), 8 and 12 and the employer rule below are the fifth version (D-162, after a security audit showed how a seal could be bought cheaply). The sixth version (D-163, as amended) counts the people behind a reviewer’s reputation more strictly (below). Unfavourable reviews pass the same rules: “changes requested” needs unfavourable reviewers who pass them too.
Expertise
Section titled “Expertise”A reviewer is an expert for the seal in either of two ways:
- Reputation 10 in the article’s primary field or in its secondary field (see Reputation and vote weight), counted strictly: only reputation earned at least 90 days before the review, that came from at least 3 different people (a vote, a “useful” mark or a resolution; for reputation claimed from an article’s votes, the people who voted, each for their share), none of them an author of the article nor another address of the reviewer. Since rule version 6, each of those people counts only if they have a verified ORCID iD or are themselves experts in the field by these rules, and reputation that two addresses gave each other is ignored on both sides, so that a few accounts voting for one another cannot make each other experts. Reputation is a public record of who gave what, so the indexer reads this from the blockchain. A review starts to count once its author has 10 counted this way.
- At least 3 works in the article’s area on OpenAlex, an open index of scholarly works, found through the reviewer’s verified ORCID iD (see Link your ORCID iD). This path lets established researchers count from their first day on Cabdell.
How the OpenAlex works are counted: the indexer asks OpenAlex for the person’s works grouped by field, where each work
belongs to its main OpenAlex field. OpenAlex has 26 fields, and Cabdell classifies by the six areas of the OECD fields
(an article’s area is the first part of its field code: 0x0303 is in area 3). The match is made at the level of areas,
and an OpenAlex field may belong to two. The works of every OpenAlex field that belongs to the area of the article’s
primary field are added up, and the same is done for the area of its secondary field; 3 or more in either is enough.
| Cabdell area | OpenAlex fields whose works count |
|---|---|
| Natural sciences | Agricultural and Biological Sciences; Biochemistry, Genetics and Molecular Biology; Chemistry; Computer Science; Earth and Planetary Sciences; Environmental Science; Immunology and Microbiology; Mathematics; Physics and Astronomy |
| Engineering and technology | Chemical Engineering; Computer Science; Energy; Engineering; Environmental Science; Materials Science |
| Medical and health sciences | Biochemistry, Genetics and Molecular Biology; Dentistry; Health Professions; Immunology and Microbiology; Medicine; Neuroscience; Nursing; Pharmacology, Toxicology and Pharmaceutics |
| Agricultural sciences | Agricultural and Biological Sciences; Veterinary |
| Social sciences | Business, Management and Accounting; Decision Sciences; Economics, Econometrics and Finance; Psychology; Social Sciences |
| Humanities | Arts and Humanities |
The works per field are read again whenever the person’s ORCID record is checked, which happens weekly.
Conflicts of interest
Section titled “Conflicts of interest”The reviewer must have no conflict with any author of the article:
- The same ORCID iD. Two addresses whose verified ORCID iD is the same person.
- A Cabdell article written together. The reviewer and an author are both authors of another Cabdell article published less than 3 years before or after the review.
- Works together on OpenAlex. Both have verified ORCID iDs, and OpenAlex lists at least one work with both as authors, published between 3 years before the pair’s first review and 3 years after its last. “The pair” is this reviewer and this author, across every article (not retracted) where one of them reviewed an article of the other. The indexer counts these shared works again every 30 days, because OpenAlex adds works late, and at once when a new review moves the window.
- The same institution. The public ORCID records of both name the same current employer. Names are compared without case or accents. Of an author’s record, the employers an institution put there (through its own ORCID integration) always count, and of those the author entered themselves only the two most recent: listing many employers cannot push reviewers out.
- Support. In the 3 years before the review, an author voted for the reviewer’s article, marked one of their reviews useful, voted for one of their comments or resolved it. Support after the review never takes a review away.
The ORCID, OpenAlex and institution checks need both people to have verified ORCID iDs. The reviewer always has one (rule 2); for an author without one, only the Cabdell and support checks apply.
Text and reciprocity
Section titled “Text and reciprocity”The indexer reads each review’s text from IPFS and keeps it only if it matches the review’s fingerprint on the blockchain. Until then the review “has not been read from IPFS yet”, and it is judged again once the text is read. Words are counted as a reader sees them, with the Markdown formatting removed.
Reciprocity applies to favourable reviews only. If an author of the article gave the reviewer a favourable review (on an article the reviewer wrote) in the 2 years before, this favourable review does not count. It stops two people from sealing each other’s work in turn.
The values
Section titled “The values”| Value | Meaning | Rule version 6 |
|---|---|---|
| required | favourable reviews that count needed for the seal | 2 |
| minimum reputation | reputation in the article’s primary or secondary field | 10 |
| OpenAlex works | works in the article’s area, found through the reviewer’s verified ORCID iD | 3 |
| co-authorship years | window for Cabdell and OpenAlex co-authorship conflicts | 3 |
| minimum words | length of a review’s text | 200 |
| reciprocity years | look-back for an author’s favourable review of the reviewer | 2 |
| granters | different people the reputation must come from, each with a verified ORCID iD or expertise of their own | 3 |
| reputation age | days the reputation must be old at the review | 90 |
| support years | look-back for an author’s support of the reviewer | 3 |
| own employers | an author’s most recent employers they entered themselves that count | 2 |
A year here is 365 days. In the indexer’s code these values are SEAL_PARAMS, and the rule version is SEAL_RULE.
A living seal
Section titled “A living seal”The seal is not a stamp given once. It follows the article:
- Later reviews can take it away. Unfavourable reviews that count can bring the article to “In peer review” or “Changes requested”.
- It is recomputed whenever its inputs change: reviews, their texts, reputation, authors, statuses, versions, and the ORCID and OpenAlex checks. A review can therefore start to count (its text is read, its reviewer reaches reputation 10) or stop counting (a shared work appears on OpenAlex).
- It is suspended while the article is disputed and comes back if the dispute is cleared or expires (see Flags, disputes and governance).
- It is gone if the article is retracted.
- The article page says since when the article has held the seal, counted from the review that last earned it.
The reviewed version
Section titled “The reviewed version”The seal names the version that was current when the latest review that counts was written. The article page says either “The reviews that count were written on version 2, the current one” or, after a newer version, that the current version “has not been reviewed yet”. A new version does not remove the seal; it tells readers what was reviewed.
Worked example
Section titled “Worked example”An article by Ana (the submitter) and Ben (a confirmed co-author) has its primary field in Health sciences (area: Medical and health sciences) and its secondary field in Psychology (area: Social sciences). Three reviews arrive:
| Review | Recommendation | Reviewer | Result |
|---|---|---|---|
| 1, Carla | accept | a verified ORCID iD and reputation 14 in Health sciences, from five people over the past year; 640 words | counts (expertise from reputation): favourable |
| 2, Dan | minor revision | no reputation, but a verified ORCID iD with 12 works in Medicine on OpenAlex; he and Ben published a Cabdell article together two years before this review | does not count: the reviewer and an author wrote an article together, less than 3 years from the review |
| 3, Eva | major revision | a verified ORCID iD with 4 works in Psychology on OpenAlex; 320 words | counts (expertise from works on OpenAlex, through the secondary field): unfavourable |
Dan passes the expertise rule (Medicine belongs to Medical and health sciences), then fails the Cabdell conflict rule, which is the reason shown. Two reviews count: 1 favourable, 1 unfavourable. The article is In peer review.
Then Farid, with a verified ORCID iD and reputation 10 in Psychology (from three people with verified ORCID iDs, months ago), writes 230 words recommending accept. It counts: 2 favourable against 1 unfavourable, so the article is Peer-reviewed from Farid’s review on, and the seal names the version Farid reviewed. Had Farid written 180 words, his review would show “shorter than 200 words (180)” and the article would stay In peer review.
Later Gil, an expert with a verified ORCID iD and no conflict, recommends reject in a review that counts. Now there are 2 favourable and 2 unfavourable: the unfavourable ones are at least 2 and as many as the favourable, so the article shows Changes requested and no longer holds the seal.
Computed by the indexer, and reproducible
Section titled “Computed by the indexer, and reproducible”Nothing about the seal is written on the blockchain; the contract does not know it exists. The Cabdell indexer (the service that reads the blockchain and builds the website’s data) computes it from:
- the blockchain: reviews and their recommendations, authors, statuses, versions and reputation;
- the review texts on IPFS, each checked against its fingerprint;
- public ORCID records and OpenAlex, for verified ORCID iDs.
Anyone who runs an indexer gets the same result from the same inputs; the ORCID and OpenAlex checks are simply asked again from those services (see Verify without trusting cabdell.press). Each result names its rule version, version 6 today, so a change of rules can never pass unnoticed.
Where you see it:
- a badge next to every article in the feed, search results and lists;
- a “Peer review” panel on the article page, with every review marked “counts” or “does not count” and why, and “How the seal works”;
- seal counts on profiles;
- the “Peer-reviewed” filter of Search, for example https://cabdell.press/testnet/search?sealed=true;
- the indexer API:
GET /articles/:idreturnsseal_detailand lists returnseal(see Indexer API).