◎ OS PUB Apache 2.0 ← All specifications

P157 — AIEP — Evidence Attribution and Citation Protocol

Publication Date: 2026-03-27 Status: Open Source Prior Art Disclosure Licence: Apache License 2.0 Author/Organisation: Phatfella Ltd Schema: AIEP_OS_SPEC_TEMPLATE v1.0.1 — https://aiep.dev/schemas/aiep-os-spec-template/v1.0.1

Framework Context

[0001] This disclosure operates within an Architected Instruction and Evidence Protocol (AIEP) environment as defined in United Kingdom patent application number GB2519711.2, filed 20 November 2025, the entire contents of which are incorporated herein by reference.

[0002] The present disclosure defines a canonical protocol for attributing evidence artefacts to their source publications, datasets, institutional originators, or originating individuals through a structured CitationRecord schema, and for maintaining a CitationIndex in the Deterministic Dual-Ledger Memory Substrate (P80) enabling any downstream reasoning system or output artefact to carry verifiable, structured citations back to primary sources.

Field of the Disclosure

[0003] This disclosure relates to evidence attribution and formal citation protocols for governed artificial intelligence reasoning systems.

[0004] More particularly, the disclosure concerns a CitationRecord schema, a CitationIndex ledger structure, a citation extraction and attachment procedure applied during evidence ingestion, and a citation rendering interface allowing AIEP reasoning outputs to carry structured citations in standard formats including BibTeX, APA, Vancouver, and schema.org/ScholarlyArticle.

Background

[0005] AIEP-governed reasoning systems ingest evidence artefacts from journals, government datasets, web mirrors, federated providers, and other sources. When a reasoning output is produced by a chain that draws on these artefacts, it is essential that the reasoning output can carry formal attribution back to the original sources, enabling human reviewers and downstream systems to verify the evidentiary basis of any conclusion.

[0006] Existing AIEP specifications define how evidence artefacts are retrieved (P10), normalised (P14), hashed and stored (P38), and given a provenance chain (P150). However, none of these specifications defines a formal citation schema — the structured metadata needed to attribute an artefact to its originating publication, dataset, or author — or a mechanism for attaching and retrieving citations in standardised bibliographic formats.

[0007] Reasoning transparency requires that citations be first-class objects in the AIEP evidence layer: attached to artefacts at ingestion time, indexed for retrieval, and carried forward into reasoning outputs in a format usable by human reviewers, compliance systems, and downstream academic or regulatory processes.

Summary of the Disclosure

[0008] A CitationRecord is a structured metadata object attached to an EvidenceNode at ingestion time. It comprises:

citation_id — SHA-256 hash of the canonical serialisation of all other CitationRecord fields
evidence_id — identifier of the EvidenceNode this CitationRecord is attached to
citation_type — enumeration: JOURNAL_ARTICLE, DATASET, PREPRINT, BOOK, BOOK_CHAPTER, GOVERNMENT_REPORT, CONFERENCE_PAPER, WEB_PAGE, PATENT, STANDARD, OTHER
title — title of the source work
authors — ordered list of author names in FAMILY, GIVEN format
year — four-digit publication year
doi — DOI if available (null if not applicable)
url — canonical URL of the source (null if unavailable)
journal — journal or conference name (for JOURNAL_ARTICLE and CONFERENCE_PAPER)
volume, issue, pages — bibliographic locators (null where not applicable)
publisher — publisher or institution name
jurisdiction — ISO 3166 jurisdiction code for the originating institution (null if international)
access_date — ISO 8601 date on which the artefact was retrieved
licence — SPDX licence identifier or "PROPRIETARY" or "UNKNOWN"

[0009] Citation Extraction: During evidence ingestion, the AIEP normalisation pipeline (P10) executes a CitationExtractor step that attempts to populate a CitationRecord from structured metadata embedded in the source document (e.g. Dublin Core, OpenGraph, schema.org/ScholarlyArticle, DOI API response). Where structured metadata is absent, the CitationExtractor applies heuristic extraction from document headers, footers, and bibliographic sections.

[0010] CitationIndex: All CitationRecords are appended to the CitationIndex partition of the Deterministic Dual-Ledger Memory Substrate (P80). The index is queryable by doi, url, title, author, year, and evidence_id. Each CitationRecord entry in the index references the corresponding ProvenanceChain entry (P150) for the same artefact, enabling a single lookup to retrieve both the citation metadata and the full transformation history.

[0011] Citation Attachment: The CitationRecord is attached to its EvidenceNode as the citation field within the EvidenceNode schema. All downstream operations that consume the EvidenceNode carry the AttachedCitation reference, ensuring that citation metadata is not lost through normalisation, enrichment, or recall operations.

[0012] Citation Rendering: The CitationRenderer component produces formatted citation strings from a CitationRecord in the following standard formats, selected by the consuming system at render time:

BIBTEX — BibTeX entry conforming to BibTeX 1987 specification
APA7 — American Psychological Association 7th Edition
VANCOUVER — Vancouver style (numeric biomedical citations)
CHICAGO_AUTHOR_DATE — Chicago Manual of Style 17th Edition, author-date variant
SCHEMA_ORG — JSON-LD schema.org/ScholarlyArticle object
AIEP_CANONICAL — canonical AIEP serialisation (JSON, all fields)

[0013] Output Attribution: When a reasoning chain produces a ReasoningOutput, the ReasoningOutput schema includes a citations field containing the citation_id values of all EvidenceNodes accessed during the chain. A CitationList can be retrieved from the CitationIndex by resolving these IDs, enabling an output document to carry a fully formatted reference list.

ASCII Architecture

Evidence Artefact (retrieved)
        │
        ▼
┌───────────────────┐
│  CitationExtractor│ ◀── DOI API / schema.org / heuristic
│  (P10 step)       │
└─────────┬─────────┘
          │ CitationRecord
          ▼
┌───────────────────┐      ┌──────────────────────┐
│  EvidenceNode     │─────▶│  Ledger CitationIndex │
│  .citation = ref  │      │  (P80)               │
└───────────────────┘      └──────────────────────┘
          │
          ▼
┌───────────────────┐
│  CitationRenderer │
│  BibTeX / APA7 /  │
│  Vancouver / JSON │
└─────────┬─────────┘
          │
          ▼
   ReasoningOutput
   .citations = [...]

Operational Detail

[0014] DOI Resolution: Where a doi is present in the CitationRecord, the CitationExtractor queries https://doi.org/{doi} with Accept: application/vnd.citationstyles.csl+json to obtain a structured citation object. Fields from the CSL-JSON response are mapped to the CitationRecord schema. This query is performed once at ingestion; the result is cached in the CitationIndex to avoid repeated external HTTP calls.

[0015] Conflict Handling: Where CitationExtractor finds inconsistent metadata between embedded document metadata and the DOI API response, the DOI API response takes precedence and the conflict is logged as a CitationConflictRecord in the ledger, preserving both versions for manual review.

[0016] Privacy Consideration: Where evidence artefacts are retrieved from sources that restrict author-level attribution (e.g. anonymised peer review submissions), the CitationRecord authors field is set to ["[ANONYMOUS]"] and the citation_type includes an anonymised flag, ensuring no personally identifying authorship data is recorded against embargoed content.

Claims-Exclusion Notice

This specification is published as open-source prior art. No patent claims are asserted by the author in respect of the mechanisms described. Any third party seeking to patent mechanisms substantially equivalent to those described herein is placed on notice of this prior art disclosure.