Skip to content

Key Concepts

Six foundational ideas that shape the scaffold's design -- from the Oracle Problem to conformal prediction.

Every design decision in this scaffold traces back to a small set of theoretical concepts from the companion paper. This section explains each one at three levels of depth: a plain-language analogy, a music-industry perspective, and an engineering deep-dive.

The Concepts

graph TD
    OP[Oracle Problem] -->|motivates| ABD[Attribution-by-Design]
    ABD -->|requires| AL[Assurance Levels A0-A3]
    ABD -->|requires| ER[Entity Resolution]
    AL -->|quantified by| CP[Conformal Prediction]
    ER -->|feeds| CP
    CP -->|exposed via| MCP[MCP Consent]
    TF[Two-Friction Taxonomy] -->|diagnostic for| ABD

    style OP fill:#fff3e0,stroke:#e65100
    style AL fill:#e8f5e9,stroke:#2e7d32
    style TF fill:#e1f5fe,stroke:#0277bd
    style CP fill:#fce4ec,stroke:#c62828
    style MCP fill:#f3e5f5,stroke:#6a1b9a
    style ER fill:#e0f2f1,stroke:#00695c
    style ABD fill:#fff8e1,stroke:#f57f17

Reading Order

Each page uses the same three-tier structure:

Section Audience What You Get
The Simple Version Anyone Plain-language analogy, zero jargon
For Music Industry Professionals Label ops, rights managers Real-world implications, industry context
For Engineers Developers, ML engineers Implementation details, code pointers, citations

Start Where You Are

You do not need to read these in order. If you are an engineer wanting to understand the confidence system, jump straight to Conformal Prediction. If you are a rights manager wondering how AI platforms query permissions, start with MCP Consent.

Concept Map

Concept One-Line Summary Key Question It Answers
Oracle Problem Digital systems cannot fully verify physical reality Why can't we just detect what was used to train a model?
Assurance Levels A0--A3 tiers of verification depth How much should I trust this attribution?
Two-Friction Taxonomy Administrative friction vs. discovery friction Which parts of music licensing should we automate?
Conformal Prediction Calibrated confidence with statistical guarantees What does "90% confident" actually mean?
MCP Consent Machine-readable permission queries for AI How do AI platforms ask "can I use this music?"
Entity Resolution Matching messy records across sources Is "Imogen Heap" on Discogs the same as on MusicBrainz?

How These Connect to the Pipeline

The scaffold's 5-pipeline architecture maps directly to these concepts:

Pipeline Primary Concept Boundary Object
ETL (5 sources) Entity Resolution NormalizedRecord
Entity Resolution Entity Resolution, Assurance Levels ResolvedEntity
Attribution Engine Conformal Prediction, Assurance Levels AttributionRecord
API / MCP Server MCP Consent, Two-Friction PermissionBundle
Chat Agent + Frontend Oracle Problem (transparency) Human-readable UI

Paper Reference

All concepts are developed in detail in: Teikari, P. (2026). Governing Generative Music: Attribution Limits, Platform Incentives, and the Future of Creator Income. SSRN No. 6109087.