{
 "slug": "contribution-attestation",
 "topic_id": "TOPIC-171",
 "cluster": "Higher Education, Research & Credential Identity",
 "tier": "Tier B",
 "title": "Authorship attestation: research integrity when the draft came from a model",
 "summary": "Authorship and data-integrity declarations are checkboxes in a submission portal. When a figure is later challenged, co-authors dispute who produced it and the institution has correspondence rather than evidence.",
 "lede": "A research misconduct inquiry begins with a simple question that turns out to have no recorded answer: who made this figure? The submitting author checked a box saying all authors approved. Six people are named. Four of them last saw a draft in March.",
 "date": "2024-06-27",
 "category": "Vertical",
 "author_id": "margot-reyes",
 "tags": [
  "research integrity",
  "authorship",
  "AI in research",
  "data provenance",
  "misconduct",
  "publication ethics"
 ],
 "image_title": "Contribution Attestation",
 "schema": "Article",
 "key_takeaways": [
  "Contribution is recorded as a role taxonomy asserted once, at submission, by one author on behalf of all.",
  "Generative tools have made the authorship question harder at precisely the moment institutions are settling misconduct matters in the millions.",
  "Per-contribution attestation binds each contributor to the data and figures they produced, with an explicit field for model assistance."
 ],
 "body": [
  {
   "type": "h2",
   "text": "The denominator nobody has"
  },
  {
   "type": "diagram",
   "kind": "flow",
   "alt": "The declaration and the dispute it is supposed to settle",
   "caption": "The record is a boolean. The question is about a specific figure and a specific person.",
   "nodes": [
    {
     "label": "Manuscript submitted",
     "note": "portal"
    },
    {
     "label": "Authorship declaration ticked",
     "note": "one checkbox",
     "bad": true
    },
    {
     "label": "Figure challenged",
     "note": "months or years later"
    },
    {
     "label": "Co-authors dispute authorship",
     "note": "record says nothing",
     "bad": true
    }
   ]
  },
  {
   "type": "p",
   "html": "Start with what is measurable. Research misconduct findings are rare relative to publication volume, and that ratio is frequently used to argue the problem is small."
  },
  {
   "type": "p",
   "html": "It is the wrong ratio. The relevant denominator is not publications; it is publications where a contribution question was asked. Inquiries are opened when something surfaces — an image duplication, a whistleblower, a failed replication. In every other case the question is never put, so the record's inability to answer it is never tested."
  },
  {
   "type": "p",
   "html": "What institutions can measure is the cost when it is tested: a January 2026 settlement of $2.3 million under civil monetary penalty authority following self-disclosure of falsified NIH grant data is a published data point, and investigation costs before any settlement are substantial."
  },
  {
   "type": "h2",
   "text": "How contribution is recorded today"
  },
  {
   "type": "p",
   "html": "Contributor role taxonomies assign roles — conceptualisation, methodology, formal analysis, writing — to named authors. It is a genuine improvement over an undifferentiated author list and it has been widely adopted."
  },
  {
   "type": "p",
   "html": "Its evidentiary properties are limited in three specific ways:"
  },
  {
   "type": "ul",
   "items": [
    "It is asserted at submission, by the submitting author, on behalf of everyone.",
    "It assigns roles to people, not artefacts to people. Nobody is bound to a specific figure or dataset.",
    "It is recorded in the journal's system, which is a third party to any later institutional inquiry."
   ]
  },
  {
   "type": "h2",
   "text": "What model assistance changed"
  },
  {
   "type": "p",
   "html": "Publisher policies now require disclosure of generative tool use, and authors disclose it in a methods statement. That is the right direction and it is coarse: a manuscript-level statement that a tool was used says nothing about which figure, which analysis, or which paragraph."
  },
  {
   "type": "p",
   "html": "The harder case is not disclosure but attribution. When a figure produced with substantial model assistance is later challenged, the question is who checked it — and the answer is currently a person's recollection of a workflow from eighteen months earlier."
  },
  {
   "type": "h2",
   "text": "The Contribution Attestation"
  },
  {
   "type": "p",
   "html": "Bind each contributor to the artefacts they produced, at the time they produce them."
  },
  {
   "type": "code",
   "text": "{\n  \"type\": \"manav-stmt/1\",\n  \"action\": \"research_contribution\",\n  \"render\": [\n    \"Manuscript: [working title / identifier]\",\n    \"Artefact: [figure 3 | dataset S2 | analysis script]\",\n    \"Artefact hash: [value]\",\n    \"Role: [produced | analysed | verified]\",\n    \"Tool assistance: [none | model id/version, scope of use]\",\n    \"Source data: [hash of the underlying dataset]\"\n  ],\n  \"contributor\": \"[credential assertion]\"\n}"
  },
  {
   "type": "p",
   "html": "A manuscript then carries a set of attestations rather than a single submission-time declaration. An inquiry that asks who made figure 3 has an answer signed by that person at the time, referencing the exact file."
  },
  {
   "type": "h2",
   "text": "What this protects"
  },
  {
   "type": "p",
   "html": "Primarily the innocent co-author, which is the argument that makes this adoptable rather than resented."
  },
  {
   "type": "p",
   "html": "In a misconduct inquiry, a co-author who contributed one analysis to a paper whose problems lie elsewhere currently defends themselves by recollection and email archaeology. Attestations bound to artefacts establish the boundary of their contribution as a fact."
  },
  {
   "type": "h2",
   "text": "Disciplinary variation"
  },
  {
   "type": "p",
   "html": "Authorship norms differ enormously. High-energy physics collaborations list thousands of authors; a mathematics paper may have one. A mechanism that assumed one model of authorship would be useless outside it."
  },
  {
   "type": "p",
   "html": "This design does not propose an authorship standard. It proposes an evidentiary layer beneath whatever standard a field uses: whoever produced an artefact signs for that artefact, and how authorship credit is assigned remains a disciplinary question."
  },
  {
   "type": "h2",
   "text": "Where to start"
  },
  {
   "type": "p",
   "html": "One laboratory, one submission. Attest figures and datasets only — not text, which is where disciplinary norms are most contested and where the evidentiary value is lowest."
  },
  {
   "type": "p",
   "html": "If the laboratory finds the attestation burden material, the design is wrong and should be simplified before extending. If they find it useful during revision — which is the common early report, because it resolves internal questions about which version of a figure is current — extend it."
  },
  {
   "type": "h2",
   "text": "What a useful contribution record contains"
  },
  {
   "type": "table",
   "caption": "Per-contributor, per-artefact",
   "head": [
    "Element",
    "Why"
   ],
   "rows": [
    [
     "Which contributor",
     "By name, not a submitting account"
    ],
    [
     "Which artefact",
     "This figure, this dataset, this analysis"
    ],
    [
     "What role",
     "Conceived, produced, analysed, verified"
    ],
    [
     "Tooling involvement",
     "Generated, edited, or composed directly"
    ],
    [
     "When",
     "Bound inside the signature"
    ]
   ]
  },
  {
   "type": "p",
   "html": "The tooling row is the new one. It is also the one most likely to be asked about, and stating it plainly is more defensible than a general disclosure buried in a methods section."
  },
  {
   "type": "h2",
   "text": "Objections and honest limits"
  },
  {
   "type": "p",
   "html": "<strong style=\"font-weight:600\">“Contributor taxonomies already cover roles.”</strong> They do, at the paper level, self-reported by the submitting author. The gap is per-artefact attestation signed by each contributor."
  },
  {
   "type": "p",
   "html": "<strong style=\"font-weight:600\">“This will slow submission.”</strong> One gesture per contributor, on the artefacts they claim. Slower than a checkbox; substantially faster than an integrity investigation."
  }
 ],
 "faq": [
  {
   "q": "Does this imply researchers are dishonest?",
   "a": "No. Misconduct is rare. The argument concerns the evidentiary position of honest researchers when a question is asked, which is currently weak for everyone."
  },
  {
   "q": "Does the attestation contain research data?",
   "a": "No. It carries hashes of the artefacts and the source data, not the data itself."
  },
  {
   "q": "How does this handle large collaborations?",
   "a": "Attestation is per-artefact, so it scales with artefacts rather than with author count. A thousand-author paper does not require a thousand attestations."
  },
  {
   "q": "What if a figure is revised during peer review?",
   "a": "A new attestation is produced for the revised artefact. The chain shows the revision and who made it, which is information reviewers and editors currently lack."
  },
  {
   "q": "Don't contributor taxonomies solve this?",
   "a": "They describe roles at paper level, self-reported by the submitting author. They are not per-artefact and not individually signed."
  },
  {
   "q": "Why record tooling involvement?",
   "a": "Because it is what integrity panels now ask about, and a plain statement is more defensible than a general disclosure in the methods."
  },
  {
   "q": "Does this slow submission?",
   "a": "One gesture per contributor on the artefacts they claim — slower than a checkbox, far faster than an investigation."
  }
 ],
 "sources": [
  {
   "t": "2 CFR Part 200 — Uniform Administrative Requirements for Federal Awards",
   "u": "https://www.ecfr.gov/current/title-2/subtitle-A/chapter-II/part-200"
  },
  {
   "t": "COPE — Committee on Publication Ethics guidelines",
   "u": "https://publicationethics.org/guidance"
  },
  {
   "t": "Publisher policies on disclosure of generative AI use in manuscripts."
  }
 ],
 "related": [
  {
   "slug": "effort-certification-evidence",
   "title": "Time-and-effort certification",
   "category": "Compliance"
  },
  {
   "slug": "ambient-scribe-attestation",
   "title": "Who signed the note the model wrote?",
   "category": "Vertical"
  },
  {
   "slug": "annex-22-ai-oversight-receipt",
   "title": "EU GMP Annex 22",
   "category": "Developer"
  },
  {
   "slug": "synthetic-resume-storm-ai-generated-portfolios-overwhelm",
   "title": "When every credential a candidate presents can be manufactured",
   "category": "Future of Work"
  }
 ],
 "image": "https://cdn.twc.sh/images/igcache/Contribution%20Attestation/1500_900/blog.jpg",
 "wordcount": 965,
 "url": "/blog/contribution-attestation.html",
 "reading_time": "4 min read",
 "seo_title": "Authorship attestation and research integrity",
 "meta_description": "Authorship and data-integrity declarations are checkboxes in a submission portal. When a figure is challenged, co-authors dispute who produced it.",
 "hub": {
  "slug": "topics/credential-identity",
  "title": "Education and credential identity"
 },
 "answer": "Authorship and data-integrity declarations are checkboxes in a submission portal. When a figure is later challenged, co-authors dispute who produced it, and the record says only that a box was ticked. Generative drafting makes that dispute more common and the record no more informative.",
 "answer_q": "Who wrote this, when the draft came from a model?",
 "glossary": [
  {
   "term": "Contributor taxonomy",
   "def": "A structured vocabulary for describing author roles, usually reported at paper level by the submitting author."
  },
  {
   "term": "Research integrity panel",
   "def": "The body investigating allegations about a published work, which currently reconstructs authorship from memory and email."
  },
  {
   "term": "Per-artefact attestation",
   "def": "A signed claim about one figure, dataset or analysis rather than about the paper as a whole."
  }
 ],
 "checklist": {
  "title": "Making authorship attestable",
  "id": "attest",
  "desc": "Four steps.",
  "steps": [
   {
    "name": "Attest per artefact, not per paper.",
    "text": "Figures, datasets and analyses are what get challenged."
   },
   {
    "name": "Have each contributor sign their own claims.",
    "text": "Not the submitting author on everyone's behalf."
   },
   {
    "name": "Record tooling involvement explicitly.",
    "text": "Generated, edited or composed directly."
   },
   {
    "name": "Keep the receipts with the submission record.",
    "text": "So an integrity panel retrieves rather than investigates."
   }
  ]
 },
 "cta": {
  "title": "Where this fits in Manav",
  "html": "Manav binds the person to the exact record, credential or submission being authorised, and produces a receipt a receiving institution or an integrity panel can verify without contacting the issuer.",
  "href": "../docs.html",
  "label": "See issuance receipts"
 }
}