Этика · Начальный
Agree authorship, disclose support and use AI responsibly
Agree contributions and prepare truthful authorship, support and AI disclosures.
11 мин чтения · 26 окт. 2026 г.
Эта страница ещё не переведена, поэтому вы читаете английскую версию.
Главное
- Separate contributions from authorship criteria.
- Verify declarations with all authors.
- Retain human responsibility for AI assistance.
Содержание
Agree contributions and prepare truthful authorship, support and AI disclosures.
Authorship tells readers who takes credit and responsibility for a paper. Declarations explain contributions, support, relevant relationships, access to evidence and other conditions needed to interpret the work. These are part of research quality, not administrative boxes to complete after writing.
Discuss responsibilities at the start of a project and revisit them as the work develops. A student may contribute most of the analysis, a supervisor may help reshape the question, and another collaborator may identify a serious limitation. The final record should reflect actual contributions rather than an arrangement based only on seniority or friendship.
Separate authorship, contribution and acknowledgement
Authorship is a status defined by the destination's criteria and associated responsibilities. Contributorship describes what a person did. Acknowledgement credits help that does not meet the applicable authorship criteria. These categories are related but not interchangeable.
ICMJE's medical framework requires substantial intellectual contribution, participation in drafting or critical revision, final approval, and accountability. It also says that eligible contributors should have the opportunity to take part in drafting and approval. Check your journal's criteria because authorship conventions differ across disciplines. S10
Authorship should not be offered as payment, a favour or a reward unrelated to the work. A person's age, academic degree or job title does not automatically determine eligibility. A student who made the substantive contribution should not disappear from the author list because a more senior person controls submission.
Conversely, occasional advice, proofreading or access to equipment may deserve acknowledgement without authorship, depending on the complete contribution and relevant criteria. Do not assume that everyone in a laboratory must be listed or that a supervisor must occupy a particular position.
Write an agreement about responsibilities
At the beginning, record the intended question, expected tasks, data access, communication arrangements and how disagreements will be handled. Treat proposed author order as provisional if contributions are not yet known. Revisit the agreement at major milestones.
A useful responsibility record identifies who maintains original data, who reviews analysis code, who checks factual claims and references, who drafts sections and who coordinates submission. Specify who will remain able to answer questions after students graduate or move institutions. The corresponding author normally manages communication with the journal and readers, but this role does not transfer every other author's responsibility to one person.
Discuss conventions rather than assuming them. Some fields use contribution based order, some frequently use alphabetical order, and joint contributions may be recognised according to journal policy. If equal contribution is claimed, agree what that means and use the permitted format.
Describe contributions with an appropriate vocabulary
CRediT is a taxonomy of fourteen contributor roles. It provides a vocabulary for describing work, but it does not determine who qualifies for authorship. S11
| Role used in CRediT | Practical meaning in a student project |
|---|---|
| Conceptualization | Developing the central question and contribution |
| Methodology | Designing the approach used to answer it |
| Investigation | Performing the experiments or collecting evidence |
| Data curation | Organising and documenting the data for use |
| Formal analysis | Applying the analytical methods |
| Software | Creating or implementing research code |
| Visualization | Preparing figures or other analytical displays |
| Supervision | Providing substantive oversight and guidance |
Drafting and critical revision also have dedicated roles in the official taxonomy. When entering journal metadata, use the exact role labels required by the submission system. One person can have several roles, and several people can share a role. Avoid assigning every role to every author simply to make the list look balanced.
Work through an authorship case
The following people and project are fictional.
Mira develops a controller comparison, implements the evaluation and writes the first draft. Dani identifies a dependence problem in the analysis, helps redesign the evaluation and critically revises the interpretation. Alex provides equipment access and occasional general advice but has not yet reviewed the paper. Noor checks grammar after the final draft is written.
A good team does not immediately assign authorship from the job descriptions. It compares actual contributions with the destination's criteria and gives relevant intellectual contributors the opportunity to participate in the remaining responsibilities. Mira and Dani appear to have substantial intellectual roles, but final approval and accountability still matter. Equipment access alone and grammar checking alone do not automatically establish authorship under the ICMJE framework.
A transparent contribution statement could describe Mira's conceptual, implementation and drafting work and Dani's methodological, analytical and revision work. Alex and Noor might be acknowledged if that accurately reflects their roles and the destination permits it. If Alex subsequently makes a substantive contribution and meets the full criteria, the record can change. The decision should follow the work, not a desire to attach a prestigious name.
Handle disagreement without rewriting history
If authors disagree, return to the documented contributions and applicable policy. Avoid secretly removing a name or submitting before a dispute is resolved. Seek the appropriate institutional or supervisory process when direct discussion fails. Preserve records of contributions and agreed revisions.
Power differences can make these discussions difficult for early researchers. A mentor outside the immediate dispute, a research integrity contact or the relevant institutional process may help. The important point is that a student's lack of seniority is not a reason to invent their consent or exclude their contribution.
Before submission, every listed author should receive the actual final version and required declarations. Approval of an earlier draft is not necessarily approval of a later version with changed data or conclusions. Record the version that was approved.
Distinguish funding from competing interests
Funding identifies financial or material support for the work. A competing interest is a relationship or activity relevant to how the work might be influenced or perceived. Funding does not automatically invalidate a study, and a lack of grant funding does not establish the absence of other interests.
Report support accurately, including relevant equipment, materials, services or access supplied by an organisation. Explain the funder's role in the design, analysis, writing and publication decision where required. ICMJE also addresses authors' access to study data in its disclosure guidance. S210
For a hypothetical engineering project, a company loaning a sensor is material support. If an author also works for that company, that relationship may require separate disclosure. For a social science project, an author evaluating a programme they manage should check the journal's requirements for relevant nonfinancial interests. Ask each author rather than writing “none” on behalf of everyone without confirmation.
A useful funding statement names the actual support and role. Avoid automatically writing that the funder “had no role” if it selected the question, supplied analysis or controlled release of the data. A disclosure should describe reality even when the reality is complicated.
Make data and code statements actionable
A data availability statement should say what is available, where, in which version and under what conditions. If access is restricted, identify the legitimate reason and the actual request or governance process. “Available on request” is not useful when no one is authorised to respond or the underlying files have not been prepared.
Distinguish raw data, cleaned analysis data, metadata and code. A PDF containing a final graph is not equivalent to the observations used to create it. If data cannot be public, code, variable definitions and a suitable synthetic demonstration may still help readers understand the analysis, provided each is labelled accurately.
Do not promise to share participant information beyond the consent and permissions that apply. Removing names may not remove identifying details from a rich interview or a small population. Article 6 explains how to organise data and document access decisions before submission.
Write ethics statements from actual records
Report the actual review body, decision, identifier and relevant consent arrangements when required. If an authorised exemption or determination applies, describe it accurately. Do not copy a placeholder committee name from a template or invent an approval number.
Different research activities require different oversight. A public dataset, an interview project, an animal experiment and a classroom engineering test do not become equivalent because all are described as student research. Establish the applicable process before data collection or access, with appropriate supervision. Article 5 covers this planning in more detail.
For a synthetic teaching example, a direct statement that the values were constructed and that no participants or animals were studied is more informative than a fabricated approval statement. Do not imply that the example underwent empirical peer review or institutional endorsement.
Use AI with a specific, documented purpose
AI can assist with brainstorming, language revision, code drafting or other tasks, but its output must be checked against the evidence and the relevant publication rules. A fluent answer may contain a nonexistent citation, an incorrect calculation or a method that does not match the design.
ICMJE's current guidance requires disclosure of relevant AI use and retains human responsibility; AI systems do not meet its authorship requirements. It does not accept AI generated material as a substitute for a primary source. S12
Policies differ. IEEE requires disclosure of generated content in the acknowledgements, including identification of the system, affected sections and extent of use. Its policy distinguishes generated content from ordinary editing and grammar assistance, for which disclosure is recommended. These details were checked for this edition on 11 November 2026; consult the destination's current instructions before submission. S203
A disclosure does not make an inappropriate use acceptable. Generated observations, fabricated quotations or altered experimental images cannot be presented as authentic evidence merely because AI use is mentioned somewhere else. Clearly label simulated or synthetic material and explain its purpose.
Check AI assistance according to what it changed
For language revision, compare the proposed wording with the original meaning. Watch for “associated with” becoming “caused”, a limitation disappearing, or an uncertain result becoming definitive. For references, open the actual source and verify both existence and relevance. A real DOI attached to the wrong title is still a defective reference.
For code, understand the analysis and test it on cases where the expected behaviour is known. Check data selection, grouping, missing values, model specification and output interpretation. Reproducing an AI generated result by running the same incorrect code twice is not an independent check.
For figures, retain the connection to the underlying data and preserve original scientific images. Distinguish an explanatory illustration from an image offered as an experimental observation. For confidential material, use only a service and workflow permitted for that material; do not assume that removing an author's name makes a manuscript or dataset public.
Keep a brief record of the system, version if available, date, task, material supplied, substantive changes and verification performed. This helps write an accurate disclosure and answer later questions.
Write a disclosure that a reader can understand
The following is a fictional example of adequate specificity, not wording to copy unless every statement is true:
During manuscript preparation, the authors used an AI language tool to suggest revisions for clarity in the Introduction and Discussion. The authors compared each suggestion with the original meaning and retained only revisions they could verify. The tool was not used to generate observations or participant quotations. All cited sources were opened and checked by the authors. The authors revised the final text and take responsibility for its content.
If AI contributed to analysis or figure generation, the disclosure should identify that use and the relevant method details rather than describing everything as language editing. Include the actual tool and version when available and required. Do not claim that all sources or calculations were checked if that work was not performed.
Complete the final accountability audit
Check that the author list, contribution statement, acknowledgements, funding, interests, ethics, data and AI declarations agree with each other and with the manuscript. Confirm that all author names and affiliations are accurate and that no institution or person is described as endorsing the work without a basis.
Practice: A team writes “No competing interests” despite one member owning the device being evaluated, and describes AI produced analysis code as “proofreading assistance”. What needs revision? Model answer: Collect the relevant relationship disclosure from the author and report it according to the destination's policy. Describe the AI contribution to analysis accurately, verify the code and document the checks. Neither declaration should be selected because it sounds simpler. Complete Template 12 using the project's actual records and contributions.
Sources
S10: ICMJE. Defining the role of authors and contributors
S11: NISO. CRediT role descriptors
S12: ICMJE. Use of AI by authors
S203: IEEE Author Center. Submission and Peer Review Policies.