1. Why Method Matters
Music history rarely arrives as one complete and uncontested account. Information can be distributed across physical releases, catalogues, photographs, interviews, printed material, private collections, websites, and human memory. Sources sometimes agree, sometimes complement one another, and sometimes conflict.
MK Music Archive and MK Music Registry therefore do not treat research as simple entry of “correct data.” The aim is to preserve a traceable path: what was found, where it came from, what it supports, what remains uncertain, and how a decision was reached.
The Registry's credibility is intended to come from method, evidence, and accountability—not from an assumption that every historical question has one readily available answer.
2. A Source Is Not the Same as Evidence
A Source identifies where information came from: for example a physical release, catalogue, book, interview, website, photograph, or other documentary material.
An Evidence Link explains how a particular version of a Source is relevant to an entity, relationship, identifier, claim, or Research Item. It can retain stance, confidence, and visibility rules.
This distinction is fundamental. A Source may contain many pieces of information; the Evidence Link explains what we are using from it and for which research decision.
3. Provenance Must Remain Visible
Provenance means that the Registry knows not only what has been recorded, but also where the information came from and the context in which it was observed.
Stable Registry identities, versioned Sources, and Evidence Links make it possible to reconstruct later why a particular value or relationship was accepted. External identifiers can assist matching, but they do not replace the Registry's own stable internal identity.
Provenance also matters for future archive, museum, and AI possibilities: without it, information may be easy to copy but difficult to verify.
4. Fidelity to the Original Source
The Registry preserves original wording, script, and historical naming wherever possible. Translations, transliterations, normalized forms, and search representations are additional access layers; they do not replace the original.
This applies to Artist names, Release and Track titles, credited-as wording, Label names, and other historical forms.
“Easier to search” and “historically recorded” are therefore not the same field serving the same purpose.
5. Uncertainty Is Valid Knowledge
Incomplete information is not a defect that must be hidden. If only a year, decade, approximate period, or range is known, the system should preserve exactly that rather than inventing an exact date.
When Sources conflict, the Registry can retain a preferred value, alternative values, the Source for each value, confidence, research rationale, and—where relevant—a public uncertainty note.
Example: one Source gives 1984, another gives 1985, while the physical Release shows no year. A responsible system does not choose an arbitrary date merely to make the field look complete.
Unsupported precision is worse than honestly represented uncertainty.
6. Research Questions and Human Judgment
A Research Item retains unresolved work: the question, candidate values or possible identities, findings, evidence, and resolution.
A Research Item does not automatically change Registry facts. Candidates and machine- or human-generated suggestions remain suggestions until they are evaluated through the appropriate process.
This is also important for future AI assistance. OCR, transcription, translation, transliteration, duplicate detection, and entity reconciliation may support research, but a machine suggestion must remain identifiable as a suggestion, retain provenance, and pass human review before becoming approved knowledge.
7. From Research to Approved Knowledge
The simplified model is:
Research and discovery → Sources and observations → Registry draft with evidence and uncertainty → Submitted revision → Review → Approved Snapshot
Review accepts a particular revision. An Approved Snapshot is a preserved, immutable representation of the information accepted at that review, including the relevant Evidence Link versions and uncertainty decisions.
Later editing does not rewrite that Snapshot. This allows the Registry to preserve how knowledge developed over time.
Approval does not claim absolute and permanent truth. It means that a particular version has been accepted according to the available evidence and applicable rules.
8. Approved Does Not Automatically Mean Public
A Public Projection is a selected, public-safe representation prepared from approved Registry knowledge.
Approval and publication are separate decisions. Rights, privacy, language readiness, public qualifications, and other readiness rules may matter before public presentation.
The public website is therefore not a direct window into the entire internal research state. It is a deliberately selected presentation of approved knowledge.
9. Corrections Do Not Erase History
Responsible archival work expects correction. A new Source or stronger evidence may show that an earlier interpretation should change.
Material factual corrections return through controlled review before republication. Minor editorial corrections remain traceable through an authorized path.
Withdrawal from public view does not mean that the internal historical record is discarded. Duplicate resolution should likewise preserve identities, provenance, and continuity rather than collapsing genuinely different editions or recordings.
10. Controlled Terms Without Losing History
Genres, formats, roles, Source types, and other cultural classifications use controlled vocabularies. They provide consistency but must remain governed and extensible because cultural terminology evolves.
Changing a label or classification term should not retroactively rewrite a historical Approved Snapshot. A new projection is required to present updated terminology publicly under the applicable rules.
A controlled vocabulary is not the same as a fixed software enum: cultural terms evolve through governance, while system states follow separate rules.
11. Principles for Responsible Knowledge
- Preserve the original before normalizing it.
- Connect important facts to Sources and evidence.
- Do not convert unknown or approximate information into fabricated precision.
- Preserve conflicting accounts when the conflict is real.
- Distinguish identity from similarity; a matching name does not automatically mean the same entity.
- Separate research work from approved knowledge.
- Separate approval from public publication.
- Preserve the history of corrections, decisions, and withdrawals.
- Apply rights, privacy, and access restrictions independently of metadata approval.
- Keep human review as the boundary before a suggestion becomes an approved fact.
12. Why This Matters for the Future
The same principles that make the Registry trustworthy today create a foundation for future possibilities. Stable identities, structured relationships, original-language forms, provenance, confidence, rights restrictions, and human-reviewed preferred values could later support links to archival holdings, institutional collections, photographs, posters, oral histories, and other heritage material.
They also matter for future AI tools: the system can use automation without losing the distinction between suggestion, evidence, and approved fact.
These possibilities are long-term potential, not claims that all of them are already implemented.
13. What This Document Is Not
This document is not a step-by-step research manual, a complete evidence-policy specification, or a definition of database, API, or workflow-engine structures.
It is a public methodological overview: it explains why MK Music Archive treats provenance, evidence, uncertainty, human review, and historical continuity as part of the knowledge itself.
Document Governance
This public document is derived from the authoritative Registry Philosophy & Domain Model, Metadata Model, Domain Entity Specification, Back Office IA & Workflows, and Controlled Vocabularies specifications, together with the approved Concept Overview.
Public authority: informational. If there is a difference, the approved numbered specifications remain the source of truth for implementation and governance.
Update triggers: a material change to the evidence/provenance model, uncertainty rules, review/approval model, correction/history rules, Public Projection model, or AI-governance principles.