Core Working Paper v6.5
The rationale, architecture and research argument for claim-level assessment assurance.
AIAP v6.5 · Open-publication edition 1
The Artificial Intelligence Assurance Protocol for AI-mediated higher education.
AIAP asks what an assessment allows an institution to claim about a student—and what evidence, safeguards and decision rules are needed to support that claim. Explore the working paper, normative standard, practical instruments, machine-readable schemas and verification tools.
James McGaughran · Research version 6.5 · Open edition published 28 September 2026
Open source and open research. Original code and schemas use Apache-2.0. Author-controlled papers, standards, figures, data, workbooks and templates use CC-BY-4.0. The new open-license grant supersedes earlier restrictive author notices preserved in historical documents. Third-party rights remain with their owners.
The rationale, architecture and research argument for claim-level assessment assurance.
The version-matched rules and requirements governing AIAP conformance claims.
A secured assessment route.
An open route with competence authentication.
An open route with a non-certifying claim boundary.
An open route integrating AI into the assessed task.
These are assurance and claim-status routes. They are not a scale of increasing AI permission or educational value. Read the standard before applying any route or making a conformance claim.
Guidance and operational instruments for examining how the specification could be implemented.
Pilot protocols and validation materials with their own evidence requirements.
Download the workbooks and examine formulas, examples and intended use.
Machine-readable assurance records, examples, a Python validator and tests.
Publication materials · Original verification pack · Current automated checks.
The open licenses permit reuse and adaptation, including commercial use, under their applicable terms. Retain required notices, provide attribution and identify modifications. Referenced third-party works are not relicensed, and reuse does not imply author or institutional endorsement.
License scope and supersession notice · Full license texts · Citation metadata · Contribution guide.
Suggested identification: James McGaughran, AIAP v6.5 (2026), open-publication edition 1. Cite the Core Working Paper for its research argument; identify the release and any modified files when sharing an implementation or adaptation.
The complete open edition includes the research files, full license texts, preserved provenance and unchanged original release ZIP. The scientific version remains 6.5; the new edition changes licensing and publication guidance.
After extraction, use Python 3.12 or later:
python -m pip install -r requirements.txt "python-docx>=1.1,<2"
python -B scripts/verify_publication.py --archive provenance/AIAP_v6.5_original.zipThe checker verifies preserved research contents, checks the original ZIP hash, runs its unchanged full verifier in isolation, and runs the assurance-record tests. The original strict verifier is intended for the original extracted archive; the maintained repository contains deliberate licensing and navigation changes.
AIAP v6.5 is published for scholarly evaluation and governed pilot preparation. The integrated intervention is not empirically validated. Successful file-integrity checks or schema tests do not establish educational effectiveness, psychometric validity, legal authority, certification or readiness for consequential deployment.
AIAP has its own validation burden. Its relationship to the broader MathGov research programme does not transfer validity or authority from other collections.
Original research release notes · Report a precise issue or reproducible failure.