PERSISTENT IDENTIFIERS · DOI · ORCID · ROR · RAiD · RESEARCH OBJECTS · IDENTITY RESOLUTION · KNOWLEDGE GRAPH

How Persistent Scholarly Identifiers Work

Scholarship has an identity problem. People share names. Organisations rename themselves. Projects start, merge and finish. Datasets acquire new versions. Articles move between web addresses. Software changes repositories. A paper can be corrected years after publication. If research infrastructure identifies these things only by text labels and URLs, the scholarly record slowly drifts into ambiguity.

A persistent identifier is not a permanent description. It is a durable handle that lets descriptions change without losing the identity of the thing being described.

Persistent identifiers, or PIDs, solve this by assigning stable, governed identifiers to entities in the research system. A DOI can identify a research object. An ORCID iD can identify a researcher. A ROR ID can identify a research organisation. A RAiD can identify a research activity or project. When these identifiers are connected through structured metadata, scholarship becomes a graph of people, organisations, projects, outputs, datasets, software, funding, versions and corrections rather than a pile of disconnected documents.

This article owns that identity layer. It is deliberately distinct from How ISBN, ISSN and Publication Identifiers Work, which covers publishing-supply identifiers; How Citations, References and Scholarly Linking Work, which owns citation relationships; and How Scholarly Indexing Works, which owns database ingestion and discovery.

The short answer

REAL-WORLD ENTITY OR RESEARCH OBJECT
  → IDENTIFIER SCHEME
  → UNIQUE PID ASSIGNED
  → PID RESOLVES TO A RECORD / LANDING PAGE
  → STRUCTURED METADATA DESCRIBES ENTITY
  → RELATIONSHIPS CONNECT OTHER PIDs
  → METADATA CHANGES OVER TIME
  → PID STAYS STABLE
  → CORRECTIONS / VERSIONS / SUCCESSORS ARE LINKED
  → MACHINES CAN RESOLVE THE GRAPH
  → HUMANS CAN RETURN TO THE SOURCE

1. The identity problem comes before the identifier

Research systems must answer questions such as: Is this the same researcher who published under another surname? Is this university the same organisation after a merger? Is this dataset version 3 or a different dataset? Is this project the parent of that grant? Is this article the corrected version of the earlier record?

A PID is useful only because these identity questions already exist.

2. Names are not identities

Two researchers can share the same name. One researcher can publish under initials, transliterations, different surnames or different ordering conventions.

Text matching alone therefore produces false merges and false splits.

3. URLs are not identities either

A journal can redesign its website. A university repository can migrate to a new platform. A dataset can move to a different storage system.

If the URL itself is treated as the identity, infrastructure change breaks scholarly continuity.

4. A PID separates identity from location

The persistent identifier remains the stable reference while the resolver or registry updates where that identifier should lead.

This separation is the same basic idea as keeping a person’s identity stable while their home address changes.

5. Persistent does not mean immutable

The identifier should remain stable. The metadata around it may need to change because scholarship changes.

Affiliations change. Titles are corrected. Organisations rename. Projects gain collaborators. A dataset receives a new version. Persistence protects identity through change rather than freezing the world.

6. An identifier is not the same thing as its metadata

The PID says which entity. Metadata says things about the entity.

IDENTIFIER
  → stable identity token

METADATA
  → title
  → name
  → affiliation
  → dates
  → type
  → licence
  → funding
  → relationships
  → versions
  → status

Good systems let metadata evolve without assigning a new identity unnecessarily.

7. A resolver turns an identifier into a route

A persistent identifier becomes practical when a human or machine can resolve it to a current record or landing page.

Resolution is what lets a stable identifier survive location change.

8. Persistence is a governance promise

A string of characters is not persistent by itself. Persistence depends on organisations agreeing to maintain resolution, metadata, ownership rules and succession over long periods.

The durable part is institutional commitment as much as technical format.

9. Good PID systems define who may create and update records

A useful identifier scheme needs rules for assignment, authority, metadata maintenance, duplicate prevention, correction and deprecation.

Without governance, identifiers multiply faster than identity becomes clearer.

10. Persistent identifiers are infrastructure, not decoration

An ORCID icon displayed beside an author name is useful. The deeper value comes when the authenticated ORCID iD travels through submission, publication, repository, funder and institutional systems.

PID value grows when the identifier moves through workflows automatically.

11. DOI identifies research objects

The Digital Object Identifier system provides persistent identifiers for objects that need durable identification and resolution.

DOIs are widely used for journal articles, books, chapters, datasets, software, reports, grants, samples and other scholarly objects depending on the registration agency and community.

12. The DOI System is governed by the DOI Foundation

The DOI Handbook is the main technical and governance reference for the DOI System.

The DOI system conforms to the ISO 26324 standard and is implemented through authorised DOI Registration Agencies.

13. A DOI name is not merely a URL

A DOI can be expressed as a resolvable HTTPS URL, but the identity concept is deeper than the current web location.

The resolver can redirect the DOI to a new landing page while preserving the DOI identity.

14. DOI syntax has a prefix and suffix

10.xxxx / suffix

10.xxxx = registrant / registration prefix space
suffix   = locally assigned object identifier under the registration workflow

Users should not infer scholarly quality, object type or publisher prestige from the appearance of the DOI string itself.

15. DOI resolution uses Handle infrastructure

DataCite’s current DOI persistence documentation explains that DOI resolution uses the Handle System, administered and maintained by the DONA Foundation.

This separates persistent naming from any one publisher’s web server.

16. Registration Agencies specialise by community and service

DOI Registration Agencies operate under the DOI Foundation framework while serving different communities and workflows.

Crossref is deeply integrated with scholarly publishing. DataCite is especially prominent for datasets and other research outputs. Other agencies serve additional domains.

17. Crossref makes publication identity machine-readable

Publishers deposit bibliographic metadata with Crossref when registering scholarly content.

Crossref’s public REST API currently exposes titles, contributors, funding, licences, post-publication updates, ORCID iDs, ROR IDs, abstracts and other deposited metadata.

18. DataCite makes research objects part of the PID graph

DataCite registers DOIs for a wide range of research resources and uses rich metadata to connect outputs with people, organisations and other works.

Its current metadata system supports datasets, software, publications, samples, instruments and other resource types.

19. DOI persistence requires maintained metadata

A DOI that still resolves to a dead or irrelevant location has technically retained its string while failing its practical persistence job.

Registration workflows therefore include obligations to keep target URLs and metadata current.

20. DataCite treats DOI identity as durable

DataCite’s current guidance states that its DOIs are intended as permanent identifiers and are not deleted once created.

When a resource changes materially, the normal solution is versioning and relationship metadata—not pretending the earlier identifier never existed.

21. Versioning is an identity decision

Not every edit deserves a new DOI. Minor corrections may update metadata or version fields. Major changes may justify a new DOI linked to the earlier one.

The decision should preserve both continuity and the ability to cite the exact scholarly state used.

22. DataCite explicitly supports version relationships

Current DataCite guidance recommends a new DOI for major content updates and connecting versions through related-identifier metadata.

This creates a sequence rather than a silent overwrite.

23. A DOI can identify a version, not merely a conceptual work

Researchers often speak as though “the paper” were one object. In infrastructure terms, a preprint, accepted manuscript, version of record and corrected edition can be distinct scholarly states.

PID design should make that version structure legible.

24. Concept DOI and version DOI solve different needs

Some repositories create one identifier representing the conceptual object and separate identifiers for individual versions.

This allows a citation to target either the evolving resource generally or the exact version used.

25. DOI does not certify quality

A DOI proves persistent identification through a registration workflow. It does not prove peer review, methodological strength, indexing status or truth.

Identifier integrity and evidence quality are different layers.

26. ORCID identifies people in research

ORCID provides persistent identifiers for researchers and contributors.

The core problem is person disambiguation: the same name can belong to different people, while the same person can appear under several name forms.

27. An ORCID iD follows the researcher across career change

Institutions, countries, disciplines, names and job titles can change while the ORCID iD remains the stable researcher identifier.

This is especially valuable in careers spanning multiple organisations and publishing systems.

28. ORCID is not a citation score

An ORCID record can connect works, affiliations, funding, peer review and professional activities.

The identifier’s purpose is identity and connection, not ranking the researcher.

29. Authentication is stronger than copied ORCID text

Anyone can type an ORCID-like string into a form. Trusted scholarly workflows use ORCID authentication so the researcher actively connects their identity to the system.

This reduces transcription errors and identity impersonation.

30. ORCID records contain provenance

An ORCID record can distinguish information added by the researcher from information written by trusted member organisations through authorised integrations.

That source provenance matters because verified affiliation data and self-entered biography text carry different evidential weight.

31. ORCID reduces administrative re-entry

When publishers, funders and institutions integrate ORCID properly, researchers can reuse identity and contribution information rather than retyping it into every system.

Persistent identity becomes workflow infrastructure rather than another profile to maintain manually.

32. ORCID should travel with every publication version

Author identifiers are most valuable when captured at submission, deposited in publication metadata, displayed on the article and carried into repositories and indexing systems.

Late addition after publication is useful but loses the opportunity for automatic propagation.

33. ROR identifies research organisations

The Research Organization Registry provides open persistent identifiers for organisations involved in research.

Its scope includes organisations that produce, fund, facilitate, manage, publish or otherwise participate in scholarly research.

34. Organisation names are unusually messy

Universities have abbreviations, translated names, former names, parent bodies, campuses, hospitals, research institutes and faculties.

Text-only affiliation matching therefore creates substantial ambiguity.

35. ROR turns affiliation into structured identity

A ROR ID can sit beside the textual affiliation in article or dataset metadata.

That lets systems recognise that “NUS”, “National University of Singapore” and another legitimate language form refer to the same organisation when the metadata points to the same ROR record.

36. ROR is deliberately open

ROR provides its registry data under CC0 through a web search interface, REST API and downloadable data dump.

This openness allows publishers, libraries, repositories, funders and research systems to reuse organisation identity without negotiating a proprietary licence for every integration.

37. ROR is community curated

Anyone can request a new organisation or an update. Proposed changes are reviewed through ROR’s curation process before release.

Current ROR documentation describes registry updates as a rolling process with releases approximately monthly.

38. ROR IDs persist through organisational change

An organisation can rename, merge, split or become inactive.

The registry needs to preserve historical identity relationships rather than simply rewriting the present name over the past.

39. Parent-child organisation relationships matter

A university system, campus, hospital or research centre may have hierarchical relationships.

Persistent organisation identifiers help systems distinguish the exact institutional level represented in a publication or grant.

40. Crossref supports ROR inside publication metadata

Crossref’s current metadata infrastructure supports ROR IDs for author affiliations and funders.

This allows research outputs to connect explicitly to the organisations behind them rather than relying only on institution-name text.

41. DataCite also connects ROR with research objects

DataCite metadata supports organisation identifiers so datasets and other objects can connect to affiliated or funding organisations.

When DOI, ORCID and ROR meet in one metadata record, a machine can resolve object, person and organisation independently.

42. RAiD identifies research activities

RAiD, the Research Activity Identifier, provides persistent identification for research projects and activities.

Its current system documentation describes RAiD as a global identifier framework governed by ISO 23527:2022.

43. Projects are not reducible to grants

A research project can involve multiple grants, organisations, people, datasets, instruments and outputs over several years.

Grant identifiers identify funding awards. RAiD is designed to identify the research activity that connects those resources.

44. Project identity persists while membership changes

Researchers join and leave. Organisations change roles. New datasets appear. Software is released. Funding is extended.

A project PID lets these changes accumulate around a stable activity identity.

45. RAiD metadata turns projects into graph nodes

Current RAiD documentation describes project metadata recording who was involved, which organisations participated, what resources were used and what the project produced.

The project becomes addressable by humans and machines rather than inferred indirectly from a pile of output titles.

46. Research activity identity fills a major scholarly gap

Traditional scholarly infrastructure is strong at identifying publications and increasingly strong at identifying people and organisations.

Projects have often remained implicit. RAiD makes the middle layer explicit.

47. A PID graph is more useful than a PID list

Having DOI, ORCID, ROR and RAiD identifiers in separate databases is helpful. Connecting them is much more powerful.

PERSON — ORCID
  ↓ participates in
PROJECT — RAiD
  ↓ hosted by
ORGANISATION — ROR
  ↓ produces
ARTICLE — DOI
  ↓ uses
DATASET — DOI
  ↓ generated by
INSTRUMENT — PID / DOI
  ↓ funded by
GRANT — DOI / award identifier

48. Relationships need types, not only links

A bare hyperlink says two things are connected. Structured metadata should say how.

Examples include IsPartOf, HasPart, IsVersionOf, Cites, IsCitedBy, IsSupplementTo, IsDerivedFrom and other typed relationships depending on the schema.

49. Crossref supports typed scholarly relationships

Crossref’s current relationship metadata can connect scholarly objects to datasets, software and other related content using identifier types and relationship types.

This turns the publication record into a machine-readable network.

50. DataCite does the same across research objects

DataCite’s RelatedIdentifier property supports citation, reference, part, version and supplementary relationships.

The current DataCite documentation explicitly describes these connections as forming a PID Graph.

51. Relationship direction matters

“A is cited by B” and “A cites B” describe the same edge from opposite directions.

Typed directional relations let machines traverse the graph reliably.

52. Versions require explicit relationships

A new dataset release can contain corrected values while remaining part of the same conceptual research output.

Version relationships should tell a future user whether one DOI replaces, extends or merely differs from another.

53. Parts and wholes need identity too

A book has chapters. A dataset can contain component files. A collection can contain samples. A project can contain subprojects.

Part-whole relations prevent identifiers from becoming disconnected islands.

54. Derived-from relationships preserve provenance

A cleaned dataset can derive from a raw dataset. A figure can derive from analysis code and data. A model can derive from a training corpus.

Where infrastructure permits, typed provenance links make transformations auditable.

55. Citation is only one relationship in the graph

Citation graphs answer who referenced whom. PID graphs can also answer who created what, where they worked, which project produced an object, which dataset supports a paper and which version replaced another.

This richer graph is closer to the actual structure of research.

56. Metadata quality decides whether graph edges are trustworthy

A DOI connected to the wrong ORCID or a ROR connected to the wrong institution creates structured misinformation.

Machine readability increases the value of good metadata and the damage of bad metadata.

57. Authentication should happen upstream

The best place to capture identity is at the workflow point where the responsible person or organisation can authenticate it.

Cleaning identity later through probabilistic matching is more expensive and less reliable.

58. Never ask humans to retype identifiers unnecessarily

Manual typing introduces transposition errors, wrong profiles and stale information.

Search, select and authenticate against the authoritative registry whenever the workflow supports it.

59. Person, organisation and object identifiers should stay separate

An ORCID identifies a person. A ROR ID identifies an organisation. A DOI identifies an object or resource under the relevant DOI registration workflow.

One universal identifier for everything would lose the governance and metadata rules specific to each entity class.

60. Identifier pluralism can be healthy

Scholarly systems also use PMID, PMCID, ISBN, ISSN, Handle, ARK, accession numbers and domain-specific identifiers.

The goal is not to eliminate every identifier except one. It is to connect appropriate identifiers and understand their scope.

61. PMID and DOI identify from different infrastructures

A biomedical article can have both a DOI and a PMID.

The DOI belongs to a persistent object-identifier system; the PMID identifies the record inside PubMed’s NLM ecosystem. Both can coexist and cross-reference the same article.

62. PMCID identifies full-text records in PubMed Central

A PMCID identifies an article in the PubMed Central full-text archive.

It should not be confused with PMID, DOI or journal identity.

63. ISBN and ISSN solve publishing-supply identity

ISBN identifies editions and formats of books and monographic publications. ISSN identifies serial publications.

They sit beside scholarly PIDs rather than being replaced by them.

64. Accession numbers often belong to disciplinary systems

Genomic sequences, protein structures, clinical trials and archival collections may have domain-specific accession or registration identifiers.

These specialised IDs can be more meaningful inside their domain than a generic DOI alone.

65. A robust graph keeps domain IDs and global PIDs together

Global scholarly infrastructure should not erase specialised systems.

It should connect them so an object remains discoverable from multiple legitimate routes.

66. Persistent identifier records need lifecycle states

Entities can be active, inactive, withdrawn, merged, superseded, corrected or deprecated depending on the identifier system.

Persistence does not mean pretending every entity is permanently current.

67. A retracted article keeps its identity

Retraction changes trust status, not historical existence.

The DOI should continue resolving to a record clearly marked with the retraction state and linked notice where publisher infrastructure permits.

68. A defunct organisation keeps historical identity

If an institute closes or merges, publications produced under its earlier identity still need to resolve correctly.

Historical affiliation cannot be repaired by replacing every old organisation name with the current successor.

69. A completed project remains citable as a project

Research activities end, but their outputs, participants and funding history remain relevant.

Project identifiers should preserve that completed state rather than disappear when active work stops.

70. PID persistence needs succession planning

What happens if a registry operator restructures, a repository closes or a publisher disappears?

Trustworthy PID infrastructure plans for organisational continuity, data export, governance transfer and resolver persistence.

71. Open metadata improves survivability

When metadata can be reused broadly, many systems can preserve references to the scholarly graph rather than concentrating all interpretation inside one proprietary database.

Open reuse is therefore a resilience feature as well as an access feature.

72. Crossref exposes much of its scholarly metadata openly

Crossref’s REST API currently requires no sign-up and exposes deposited scholarly metadata for broad reuse.

This enables libraries, research tools, indexers and AI systems to resolve publication identity at scale.

73. ROR data are openly reusable

ROR distributes organisation IDs and metadata under CC0 through search, API and downloadable datasets.

Open organisation identity reduces duplicated proprietary affiliation dictionaries.

74. DataCite metadata support interoperable connection data

DataCite’s schema makes related identifiers first-class metadata, enabling research outputs to connect to works, people and organisations.

Its current metadata documentation identifies Schema 4.7 as the latest minor version.

75. PID graphs help attribution

A dataset DOI linked to ORCID iDs and a ROR ID gives clearer evidence of who created the resource and which organisation hosted the work.

This is more robust than reconstructing attribution from free-text acknowledgements years later.

76. PID graphs help funding analysis

When grant, project, organisation, person and output identifiers connect, funders can trace portfolios with less dependence on title matching.

The same graph can support compliance, reporting and public accountability.

77. PID graphs help reproducibility

A paper can point to the exact dataset, software release, protocol and project that produced it.

This does not guarantee reproduction, but it reduces the identity ambiguity that blocks reproduction before methods are even tested.

78. PID graphs help correction propagation

If a dataset is corrected, systems can identify dependent outputs. If a paper is retracted, connected repositories can update status.

Typed relationships create the possibility of machine-assisted downstream warning.

79. PID graphs help literature discovery

A researcher can move from one article to its author’s ORCID record, then to related datasets, the hosting institution, the project and other outputs.

Discovery becomes traversal rather than keyword search alone.

80. PID graphs help systematic reviews detect duplicates

Preprints, accepted manuscripts and versions of record can be linked rather than counted as independent studies.

Identifiers do not solve study-level deduplication automatically, but they make the evidence graph easier to reconstruct.

81. PID graphs help bibliometrics clean identities

Author and institution disambiguation are major sources of bibliometric error.

ORCID and ROR reduce reliance on probabilistic string matching when adoption is strong.

82. PID graphs help institutional repositories

Repositories can connect deposited outputs with the correct researcher, organisation, version and publication record.

This improves both local stewardship and global discoverability.

83. PID graphs help publishers reduce metadata repair

Authenticated author and institution identifiers captured at submission can flow into production metadata automatically.

Every manual re-entry step removed is one fewer opportunity for identity corruption.

84. PID graphs help AI retrieve more accurately

An AI research system can use identifiers to distinguish two authors with the same name, separate versions of a dataset and resolve the current institution behind an affiliation.

This is one of the strongest practical reasons to improve PID coverage now.

85. AI needs entity resolution before synthesis

A model cannot reason reliably over a scholarly graph if it has merged the wrong people or treated several versions of one paper as independent studies.

Identity resolution is therefore upstream of trustworthy synthesis.

86. Names should become labels, not primary keys

Humans need readable names. Machines need stable identifiers.

A well-designed knowledge graph stores the human-readable label alongside a persistent ID rather than using the label itself as the unique key.

87. AI should expose identifier evidence

If an assistant claims two papers were written by the same researcher, it should prefer authenticated or strongly supported identity links over name similarity alone.

Where ORCID or another authoritative PID is available, the return path should be visible.

88. AI should distinguish current and historical organisation identity

A research institute can merge into another organisation while older papers remain affiliated with the predecessor.

Graph traversal should preserve historical truth rather than rewrite every past affiliation into the present.

89. AI should respect version identity

When a model summarises a paper, dataset or software package, it should know which version it used.

Otherwise it can merge claims from incompatible versions into an output that never existed.

90. AI should propagate correction states through PID links

If a DOI resolves to a retracted publication, a retrieval system should not quietly cite it as ordinary current evidence.

Identifier resolution should be followed by status resolution.

91. Identity graphs can create privacy risks

Connecting people, organisations, grants, locations and activities at scale can reveal patterns not obvious from any single record.

Open scholarly infrastructure still needs careful decisions about personal data, sensitive projects and lawful disclosure.

92. Public identifier does not mean all metadata must be public

An entity can need stable identity while some associated metadata remain restricted or private.

Identity persistence and data minimisation can coexist.

93. Sensitive research activities need careful project metadata

Defence, health, indigenous knowledge, commercially sensitive research and work involving vulnerable groups can contain project information that should not be exposed indiscriminately.

PID governance must separate what is needed for identity from what is unsafe to publish.

94. Duplicate identifiers weaken trust

Two PIDs accidentally assigned to the same entity can fragment history. One PID incorrectly reused for two entities is worse.

Assignment workflows need duplicate checking and explicit merge or relationship policies.

95. Identifier reuse is dangerous

A persistent identifier should not be recycled from an old entity to a new one merely because the earlier entity became inactive.

Historical references depend on the identifier continuing to mean what it meant when cited.

96. Broken resolution is a maintenance failure

If a DOI or other resolvable PID stops leading users to a meaningful record, the persistence promise is weakened.

Registrants and registry operators need monitoring and repair workflows.

97. Metadata drift is quieter than a broken link

A PID can resolve successfully while displaying stale authors, an obsolete organisation name or missing correction relationships.

Resolution health and metadata health should be checked separately.

98. Canonical source matters for each identifier class

DOI SYSTEM → DOI Foundation + Registration Agency metadata
PERSON → ORCID
RESEARCH ORGANISATION → ROR
RESEARCH ACTIVITY / PROJECT → RAiD
DOMAIN OBJECT → relevant authoritative registry

Third-party profiles can enrich discovery but should not silently override the canonical identity source.

99. Search interfaces are not always canonical registries

Google Scholar, library discovery systems and bibliometric platforms may display PID-linked information.

When identity is contested, return to the identifier registry or authoritative metadata source.

100. A PID receipt should travel with consequential assertions

PID RECEIPT

Entity type:
Persistent identifier:
Registry / Registration Agency:
Resolved landing page:
Canonical label:
Status:
Version if applicable:
Related identifiers:
Source of metadata:
Last verified date:
Known ambiguity / limitation:

This makes identity claims auditable instead of merely plausible.

101. A practical researcher checklist

Register and maintain an ORCID iD.
Authenticate ORCID rather than retyping it when possible.
Check that publications deposit your ORCID correctly.
Use DOI links rather than unstable publisher URLs in formal citations.
Check which dataset or software version a DOI identifies.
Use the correct ROR ID for your organisation.
Preserve project identifiers where a RAiD or equivalent exists.
Link datasets, software and protocols to publications.
Check correction and retraction states.
Do not treat a PID as proof of quality.

102. A practical publisher checklist

Register persistent identifiers at the appropriate publication stage.
Capture authenticated ORCID iDs from authors.
Capture ROR IDs for affiliations and funders where supported.
Deposit complete DOI metadata.
Deposit references and related research objects.
Update landing-page URLs after platform migrations.
Update metadata when author or title corrections occur.
Link corrections, retractions and new versions explicitly.
Preserve old identifiers rather than reusing them.
Audit resolver health periodically.

103. A practical repository checklist

Assign identifiers according to a documented granularity policy.
Decide when versions receive new identifiers.
Expose creator ORCID iDs.
Expose organisation ROR IDs.
Link related publications, datasets and software.
Preserve metadata provenance.
Maintain persistent landing pages.
Track withdrawals and tombstone states.
Export metadata in interoperable formats.
Plan for repository succession.

104. A practical institution checklist

Maintain the institution’s ROR record.
Integrate ORCID into research-information systems.
Write verified affiliations to ORCID where authorised.
Use PIDs in grant and project workflows.
Connect repositories to authoritative identifier registries.
Train staff not to treat names as unique keys.
Preserve historical organisation identity after restructuring.
Use persistent identifiers in reporting and assessment.
Audit duplicate and orphaned records.
Document correction workflows.

105. A practical funder checklist

Collect authenticated ORCID iDs from applicants and awardees.
Identify organisations with ROR IDs.
Assign stable award identifiers where supported.
Connect awards to research activities and outputs.
Carry grant identifiers into publication metadata.
Preserve project history after funding ends.
Use identifiers to reduce repeated manual reporting.

106. A practical AI-system checklist

Resolve identifiers before merging entities.
Prefer authoritative registry evidence over name similarity.
Keep person, organisation, project and object IDs distinct.
Track exact versions of cited research objects.
Check correction and retraction status after resolution.
Expose identifier provenance in high-stakes answers.
Preserve historical affiliations rather than rewriting them.
Do not infer that a DOI proves peer review or quality.
Use typed relations when building knowledge graphs.
Flag unresolved identity ambiguity rather than guessing.

107. Failure modes

Failure	What breaks
Name used as unique identity	Different people or organisations are merged.
URL used as identity	Platform migration breaks continuity.
PID treated as immutable metadata	Legitimate change cannot be represented.
DOI treated as quality certification	Identity infrastructure is confused with peer review.
ORCID copied rather than authenticated	Wrong person can be attached to the record.
Affiliation stored only as text	Organisation disambiguation becomes probabilistic.
Project has no stable identity	Outputs and funding remain disconnected.
Version overwritten silently	Citation and provenance history disappear.
Retraction changes not linked	Old trust state persists downstream.
Duplicate PIDs for one entity	History fragments across records.
Identifier reused	Historical references become ambiguous.
AI merges entities before PID resolution	Knowledge graph errors propagate into synthesis.

108. The deeper model: persistent identifiers manufacture continuity

Research is dynamic. People move. Projects evolve. Organisations restructure. Data change version. Publications are corrected. Infrastructure migrates.

Persistent identifiers reduce the variance created by all that motion.

CHANGE IN NAME
CHANGE IN LOCATION
CHANGE IN VERSION
CHANGE IN ORGANISATION
CHANGE IN PROJECT MEMBERSHIP
CHANGE IN PUBLICATION STATUS

        ↓

PERSISTENT IDENTITY
        +
VERSIONED METADATA
        +
TYPED RELATIONSHIPS
        =
CONTINUITY THROUGH CHANGE

This is why PIDs are not clerical conveniences. They are part of the manufacturing line that turns unstable research activity into a reliable long-term scholarly record.

109. The Wintour rule: identity must return to its canonical owner

Every consequential identity claim should have a return path to the registry or authoritative metadata source that owns that identifier class.

CLAIM
  → PID
  → AUTHORITATIVE REGISTRY
  → METADATA RECORD
  → RELATIONSHIPS
  → CURRENT STATUS
  → VERIFIED DATE

If the return path breaks, confidence should fall.

A knowledge graph becomes trustworthy when every edge can be traced back to an identity that is governed, resolvable and correctable.

110. Why persistent identifiers matter to civilisation

Civilisation depends increasingly on distributed knowledge systems. A paper is published in one country, archived in another, cited by a researcher at a third institution, reused in a dataset, incorporated into software and summarised by an AI system years later.

That chain only works reliably when the participants and objects remain distinguishable through time.

Persistent identifiers give civilisation stable joints in an otherwise moving knowledge machine. They do not make research true. They make it possible to keep asking which research object, which person, which organisation, which project and which version we are actually talking about.

Current authority routes

Continue the Archives and Publishing series

Publication control: Wintour House · eduKate Publishing · identity, authority, metadata, version, provenance, relationship, correction, freshness and archive gates.

World Return: The next time a research system shows you a name, title or institution, ask whether you are seeing a label or an identity. Find the persistent identifier, resolve it to its authoritative record, check its relationships and version state, and preserve enough provenance that a future reader—or machine—can follow the same route back.

How Persistent Scholarly Identifiers Work | From DOI, ORCID and ROR to RAiD, Research Objects, Identity Resolution and the Knowledge Graph