The 14 CRediT roles were designed around a fairly specific default picture of a
contributor: a named, professionally affiliated researcher who did some describable
piece of the work and can be held accountable for it. Real research increasingly
involves people, and non-human tools, who don’t fit that picture cleanly — community
and citizen-science participants, patient and public involvement (PPI) partners,
indigenous knowledge holders, and AI-assisted tools among them. This guide works
through how the CRediT taxonomy applies, applies awkwardly, or doesn’t apply at all
to each of these, and where CASRAI’s own controlled vocabulary already has language
for the gap.
AI-assisted work shows up here as one example of the broader pattern, not the main
subject. For the role-by-role mechanics of recording AI assistance within a CRediT
statement itself, see CASRAI’s dedicated guide,
How to Disclose AI
Assistance in a CRediT Statement, Role by Role; for the mechanics of AI-use
disclosure statements more generally (what needs disclosing, sample language, where
it goes in a manuscript), see
CASRAI’s AI disclosure guidance for authors.
This guide doesn’t repeat either — it covers the broader non-traditional-contributor
question, of which AI is one example among several.
What “non-traditional contributor” means for CRediT purposes
CRediT (Contributor Roles Taxonomy, formalized as ANSI/NISO Z39.104-2022,
originated by CASRAI in 2012–2014 and now stewarded jointly with NISO) describes
what someone did on a research output across 14 fixed categories:
Conceptualization, Data Curation, Formal Analysis, Funding Acquisition,
Investigation, Methodology, Project Administration, Resources, Software,
Supervision, Validation, Visualization, Writing – Original Draft, and Writing –
Review & Editing. It is a contribution-disclosure taxonomy, not an
authorship-qualification test — who
counts as an author is a separate question governed by criteria like ICMJE’s,
and a contribution can be real and CRediT-describable without meeting that bar. See
CASRAI’s CRediT overview and the 14
role pages for the taxonomy itself.
A “non-traditional contributor,” for this guide, is anyone whose contribution is
in principle describable by one or more of those 14 roles but who sits outside the
conventional employed/affiliated-researcher-as-author model CRediT was built around.
CASRAI’s Dictionary already tracks several such categories individually — see
non-author contributor and
acknowledged contributor for
the general pattern of crediting someone’s role without listing them as a formal
author.
Community and citizen-science contributors
CASRAI’s Dictionary defines a
citizen-scientist
contribution operationally as a “substantive contribution to research by members
of the public participating as volunteers in data collection, classification,
observation, analysis, or interpretation” — explicitly distinct from being a study
subject and distinct from professional research staff. The
European Citizen Science Association’s
(ECSA) Ten Principles of Citizen Science make recognition an explicit
commitment: one principle states plainly that citizen scientists are acknowledged in
project results and publications, though ECSA itself notes there’s no single
consensus mechanism for how that recognition should be given — acknowledgement
section, co-authorship, or something in between.
Mapped onto the 14 roles, some fit citizen-science contributions cleanly and some
don’t:
- Fits well: Investigation
(the actual data collection/observation work), Data
Curation (classification, tagging, annotation — the core task in many crowdsourced
projects), and Validation (verifying observations
or replicating a result) describe what volunteer participants are most often actually
doing. - Fits awkwardly, usually reserved for the professional team:
Conceptualization,
Methodology,
Funding Acquisition,
Project Administration, and
Supervision presuppose a degree of authority
over the study’s design and direction that a volunteer participant typically doesn’t
have — unless the project is genuinely co-designed with the community from the
outset, which is exactly what
Community-Based
Participatory Research (CBPR) is structured to do. In a real CBPR project, some
of these roles may legitimately apply to community co-investigators; in a
conventional citizen-science project where the public contributes data collection or
classification to a study designed and led by professional researchers, they usually
don’t.
A well-documented illustrative case (cited on CASRAI’s own citizen-scientist-contribution
entry): the Galaxy Zoo project, part of the Zooniverse platform, has credited its
most active volunteer classifiers as co-authors on discovery papers built from
crowdsourced galaxy classifications — a case where the contribution (large-scale,
sustained classification work) mapped recognizably onto Data Curation and
Investigation-type work, and the project chose co-authorship over acknowledgement
alone. The same Dictionary entry gives the boundary case on the other side: a
passive user of a fitness-tracking app whose data is later analysed by researchers
is a data subject, not a citizen-scientist contributor — no CRediT role
applies at all, because no substantive contribution to the research process
occurred.
Patient and public involvement (PPI) and indigenous knowledge holders
Patient and public involvement research — see CASRAI’s
PPI (Patient and Public
Involvement) and patient partner
entries — has its own, older reporting standard: the
GRIPP2 (Guidance for
Reporting Involvement of Patients and the Public, version 2) checklists, a
consensus-developed reporting standard published in 2017 with a long-form (34-item)
and short-form (5-item) version. GRIPP2 and CRediT solve different problems: GRIPP2
describes the extent and nature of PPI involvement in a study (why patients
were involved, how, at what stage, with what impact); CRediT describes discrete
contribution types on a specific output. The two are complementary, not competing —
a study can report its PPI process via GRIPP2 and still use CRediT roles (most often
Investigation,
Validation, or
Writing – Review & Editing,
e.g. reviewing a lay summary for accuracy and tone) for a
patient-partner
contribution that rises to formal co-authorship or named acknowledgement.
Whether that contribution meets authorship criteria at all is, as above, a separate
ICMJE-governed question from which roles describe it.
Indigenous knowledge holders raise a related but distinct issue. CASRAI’s
indigenous
knowledge-holder contribution,
indigenous community
review, and community-
controlled research entries reflect a governance model where individual-level
role attribution isn’t always the right unit of recognition at all — the
CARE Principles for Indigenous Data
Governance (Collective Benefit, Authority to Control, Responsibility, Ethics —
published by the Global Indigenous Data Alliance in September 2019 to complement,
not replace, FAIR) explicitly center a community’s authority to control how its
knowledge and data are represented, which can include a preference for collective or
organizational credit over naming individual contributors in a byline or role list
at all. A CRediT statement built only for individually-named contributors doesn’t
have a native way to express that choice — worth flagging to a research office
before assuming role attribution is simply a paperwork step.
Where the role-based model structurally breaks down
Three recurring mismatches show up across every non-traditional contributor type
covered above, not just one of them:
- Accountability. Every CRediT role implicitly assumes the
credited party can be identified and, in principle, asked to stand behind their
piece of the work. A citizen-science volunteer, a PPI partner, or an AI tool may
contribute something real without being positioned (or, in the AI case, able) to
accept that accountability — the same underlying tension shows up at every point on
this spectrum, most acutely with AI, which is why CASRAI’s
CRediT-for-authors guidance is explicit that generative AI tools cannot be listed
as CRediT contributors at all: they are tools, not agents capable of accountability. - Granularity. A single classification, a single translated
survey instrument, or one round of community feedback on a draft often doesn’t map
cleanly onto exactly one of the 14 categories — it may partially fit several, or fit
none well. CASRAI’s Dictionary has started building vocabulary for these in-between
cases directly — see translator
contribution, pre-
submission feedback contribution, and the
“degree of contribution” qualifier (lead/equal/supporting), which adds intensity
to a role rather than inventing a new one. - Consent and preferred form of recognition. CRediT statements
assume individually-attributed roles are the desired outcome. As above, that’s not
universally true — some contributors (particularly in indigenous and
community-controlled research contexts) may prefer collective attribution, and
standard editorial practice already requires written consent before naming any
individual contributor in a paper regardless of role.
AI-assisted work: the far end of the same problem
AI tools sit at the extreme end of the accountability mismatch described above,
which is why the position across standards bodies is unusually consistent rather
than case-by-case. The
ICMJE’s Recommendations (updated 2023) state that AI tools do not meet the
requirements for authorship because they cannot take responsibility for the
accuracy, integrity, and originality of submitted work, and cannot agree to be
accountable for it. COPE’s
February 2023 position statement, “Authorship and AI tools,” reaches the same
conclusion for the same reason: AI tools cannot assert conflicts of interest, manage
copyright, or be held accountable, so they cannot be listed as authors — human
authors remain fully responsible for any AI-assisted portion of a manuscript. Both
positions are reflected in CASRAI’s own Dictionary — see
AI as author,
AI co-authorship
rejection (ICMJE 2023), and
acknowledgement vs.
authorship for AI.
Because an AI tool cannot hold accountability, no CRediT role gets assigned to it
— not even a role that superficially matches the task it performed (drafting text
resembles Writing – Original
Draft; running an analysis pipeline resembles
Formal Analysis). Instead, the
accountable human
author keeps the role and discloses the AI use directly, typically via a
generative-AI
disclosure statement separate from the CRediT contributor list. This guide
deliberately doesn’t repeat those mechanics — for exactly how to word that
disclosure within each of the nine roles it most often touches (Conceptualization,
Investigation, Formal Analysis, Software, Data Curation, Visualization, and both
Writing roles among them), see
How to Disclose AI
Assistance in a CRediT Statement, Role by Role; for disclosure statements more
generally, see CASRAI’s AI disclosure
guidance.
Practical guidance for research offices and editors
- Separate the two questions. Decide who qualifies as an author
(ICMJE-style criteria) and how to describe what each contributor did (CRediT roles)
as two distinct steps — collapsing them is the single most common source of
confusion in non-traditional-contributor cases. - Get explicit consent before naming anyone. Standard practice
for any named contributor, and especially important where a contributor may not
have anticipated being individually identified — citizen-science volunteers,
patient partners, and community members alike. - For community/citizen-science contributions, default to the roles that
actually describe the work performed — usually Investigation, Data
Curation, or Validation — rather than assigning a fuller set of roles for the sake
of a complete-looking statement. - For indigenous and community-controlled research, check governance
protocols before assuming individual role attribution is even the right approach
— CARE-aligned projects may call for collective or organizational credit instead of,
or alongside, individually-named CRediT roles. - For AI, disclose — don’t attribute a role. Keep the
accountable human author responsible for any AI-touched section, per ICMJE and
COPE.
Related CASRAI resources
- CRediT & Authorship Attribution —
the cluster hub for all of CASRAI’s CRediT and authorship content. - The 14 CRediT roles — individual definitions for
each role referenced above. - CRediT for authors — the general how-to for
writing a CRediT statement, including its own note on non-author acknowledged
contributors and why AI tools aren’t listed as contributors. - CRediT author statement
samples — worked statement formats for conventional multi-author papers. - How to Disclose
AI Assistance in a CRediT Statement, Role by Role — the role-by-role AI
mechanics this guide deliberately doesn’t repeat. - AI disclosure guidance for authors —
the mechanics of AI-use disclosure statements specifically.
Frequently asked
Can a citizen-science volunteer be listed as a CRediT author?
Yes, if their contribution meets the journal’s authorship criteria (typically
ICMJE’s) and they consent to being named — CASRAI’s own Dictionary example is Galaxy
Zoo crediting top classifiers as co-authors. Where the contribution doesn’t rise to
that bar, it’s still appropriate to name the role performed (Investigation, Data
Curation, Validation) in an acknowledgement rather than the author byline.
Which CRediT role fits an AI tool?
None. CRediT roles presuppose an accountable contributor, and AI tools cannot
accept accountability for a submitted work — the consistent position across ICMJE
(2023) and COPE (February 2023). AI use is disclosed separately, not attributed a
role; see how to
disclose AI assistance role by role for the mechanics.
Do patient and public involvement (PPI) partners need a separate CRediT
statement from GRIPP2 reporting?
They’re complementary, not substitutes. GRIPP2 reports the nature and extent of
PPI involvement across the study; CRediT roles describe specific, discrete
contributions to a given output, and apply only if the PPI partner is named as an
author or contributor on that output.
What’s the difference between an acknowledged contributor and a
CRediT-credited non-author?
They can be the same person described two ways. “Acknowledged contributor” is
about where they’re recognized (the acknowledgements section, not the byline);
CRediT role is about what they did. A contributor can have a CRediT role recorded in
an acknowledgement without being an author at all.
Does CRediT replace authorship criteria for non-traditional contributors?
No — CRediT has never been an authorship test for any contributor type. It
describes contribution type; ICMJE-style criteria (or a journal’s equivalent)
determines who qualifies as an author. This distinction matters more, not less, for
non-traditional contributors, because it’s exactly where the two questions most
often get conflated.
