a11oy / packages /knowledge /vertical /academic-research-ethics.yaml
betterwithage's picture
sync(space): full source mirror — resolve all GitHub<->Space drift (CTO)
a6a5d8e verified
Raw
History Blame
6.9 kB
# Vertical Governance Policy: Academic Research — Belmont Report / Common Rule / COPE
# Doctrine v6 | R3 Adversarial Receipts
# Last revised: 2025-07
schema_version: "1.0.0"
vertical: academic
regime: Common-Rule/Belmont/COPE
effective_date: "2025-07-01"
jurisdiction: US-Federal-HHS/Global-COPE
meta:
title: "Academic Research AI Governance Policy — Common Rule / Belmont / COPE Alignment"
description: >
Maps the Federal Policy for the Protection of Human Subjects (Common Rule),
Belmont Report principles, Committee on Publication Ethics (COPE) guidelines,
and NSF/NIH AI data management requirements to Doctrine v6 Λ-axes for AI
systems used in academic research contexts.
authority: "45 CFR Part 46 (Common Rule); 21 CFR Part 50/56; Belmont Report (1979); COPE Guidelines (2023); NSF PAPPG Ch. II.E.3; NIH DMS Policy (2023)"
receipt_chain_required: true
merkle_root_algorithm: SHA3-256
irb_oversight: required
regulatory_clauses:
- clause_id: COMMON-RULE-46.111
title: "Criteria for IRB Approval"
citation: "45 CFR § 46.111; 21 CFR § 56.111"
full_ref: "45 C.F.R. § 46.111 — IRB criteria: risks minimised, equitable selection, informed consent, monitoring, privacy protection"
lambda_axes:
- axis: Λ4
label: Fairness
weight: 1.0
enforcement: mandatory
rationale: >
AI research involving human subjects must demonstrate equitable
participant selection; demographic stratification receipts submitted
with IRB application.
- axis: Λ3
label: Privacy
weight: 0.95
enforcement: mandatory
- clause_id: BELMONT-RESPECT-PERSONS
title: "Belmont Report — Respect for Persons (Autonomy)"
citation: "Belmont Report Part B.1 (1979); 45 CFR § 46.116"
full_ref: "Belmont Report § B.1 — Respect for Persons: informed consent; 45 C.F.R. § 46.116 requirements for informed consent"
lambda_axes:
- axis: Λ3
label: Privacy
weight: 0.90
enforcement: mandatory
rationale: >
AI systems training on participant data must have consent receipts
specifying purpose, data scope, and withdrawal mechanism.
- axis: Λ1
label: Transparency
weight: 0.88
enforcement: mandatory
- clause_id: BELMONT-BENEFICENCE
title: "Belmont Report — Beneficence / Non-Maleficence"
citation: "Belmont Report Part B.2 (1979)"
full_ref: "Belmont Report § B.2 — Beneficence: maximise benefits and minimise harms to research subjects"
lambda_axes:
- axis: Λ5
label: Safety
weight: 0.92
enforcement: mandatory
rationale: >
AI-generated research outputs that could harm participants must
undergo safety review; harm assessment receipts generated quarterly.
- axis: Λ9
label: Explainability
weight: 0.75
enforcement: mandatory
- clause_id: NIH-DMS-POLICY-2023
title: "NIH Data Management and Sharing Policy"
citation: "NIH DMS Policy (Jan 2023); NOT-OD-21-013"
full_ref: "NIH Data Management and Sharing Policy (effective 25 Jan 2023) — Data management plans and sharing of scientific data"
lambda_axes:
- axis: Λ1
label: Transparency
weight: 0.95
enforcement: mandatory
rationale: >
AI-generated research datasets and model weights must be shared per
FAIR principles; repository deposit receipts logged in chain.
- axis: Λ10
label: Sovereignty
weight: 0.78
enforcement: recommended
- clause_id: COPE-AI-AUTHORSHIP-2023
title: "COPE — AI Authorship and Disclosure"
citation: "COPE Position Statement on Authorship and AI Tools (2023)"
full_ref: "COPE Position Statement: Authorship and AI tools — AI cannot be listed as an author; authors accountable for AI-generated content"
lambda_axes:
- axis: Λ2
label: Accountability
weight: 1.0
enforcement: mandatory
rationale: >
All AI-generated content in academic publications must be disclosed;
disclosure receipt references specific AI system version and inference
timestamp per Doctrine v6 §5.1 provenance requirements.
- axis: Λ1
label: Transparency
weight: 0.95
enforcement: mandatory
- clause_id: NSF-PAPPG-AI-DATA
title: "NSF — AI Research Data Management Requirements"
citation: "NSF PAPPG Ch. II.E.3 (2024); NSF 23-1 PAPPG"
full_ref: "NSF Proposal & Award Policies & Procedures Guide (PAPPG) Ch. II.E.3 — Data management and sharing plan requirements"
lambda_axes:
- axis: Λ7
label: Auditability
weight: 0.90
enforcement: mandatory
rationale: >
NSF-funded AI research must maintain 3-year post-award data records;
Merkle DAG provides tamper-evident archive with dataset versioning.
- axis: Λ8
label: Robustness
weight: 0.72
enforcement: recommended
- clause_id: COMMON-RULE-46.111E-PRIVACY
title: "Common Rule — Privacy and Confidentiality Safeguards"
citation: "45 CFR § 46.111(a)(7)"
full_ref: "45 C.F.R. § 46.111(a)(7) — IRB must determine that privacy of subjects and confidentiality of data are adequately protected"
lambda_axes:
- axis: Λ3
label: Privacy
weight: 1.0
enforcement: mandatory
rationale: >
AI training on IRB-approved data must implement k-anonymity (k≥5)
or differential privacy (ε≤1.0); privacy parameter receipts generated
per dataset epoch.
- axis: Λ6
label: Security
weight: 0.82
enforcement: mandatory
- clause_id: EU-AI-ACT-ART-53-GPAI
title: "EU AI Act — GPAI Model Transparency for Research"
citation: "EU AI Act Art. 53; Recital 106"
full_ref: "Regulation (EU) 2024/1689 Art. 53 — Obligations for providers of general-purpose AI models used in research"
lambda_axes:
- axis: Λ1
label: Transparency
weight: 0.88
enforcement: mandatory
rationale: >
General-purpose AI models used in academic research must publish
training data summary and evaluation results; publication receipt
links to EU AI Act database entry.
- axis: Λ4
label: Fairness
weight: 0.80
enforcement: mandatory
compliance_thresholds:
minimum_lambda_coverage: 7
mandatory_axes: [Λ1, Λ3, Λ4]
receipt_retention_days: 1095 # 3 years NSF/NIH post-award
irb_review_cycle_days: 365
consent_renewal_days: 365
differential_privacy_epsilon_max: 1.0
receipt_chain:
algorithm: SHA3-256
chaining: merkle_dag
quorum: 2-of-3
nodes: [primary, irb-backup, institutional-archive]
irb_signed: true