External coherence graph: mapping an entity’s active sources

External coherence graph

The external coherence graph designates the mapping of sources, attributes, versions, collisions, and proximities that participate in the exogenous reconstruction of an entity.

This page does not define an authority hierarchy. It defines a mapping surface. The canonical admissibility of external authorities falls under External Authority Control (EAC).

1. Function

The external coherence graph first answers a descriptive question: what is actually circulating around the entity?

It identifies:

• which external sources are active;
• which attributes they expose;
• where collisions, omissions, contradictions, and ambiguities appear;
• which nodes risk contaminating the overall reconstruction.

The graph is therefore not a judgment about source quality. It is a map of the active forces inside the interpretive environment.

2. What the graph does not settle

The graph does not answer the normative question: what should count as admissible authority?

A source may be active, visible, redundant, or popular without being canonically admissible. Mapping precedes qualification. This is why the external coherence graph must be articulated with EAC.

3. What must actually be mapped

A useful mapping is more than a URL list. It should render visible at least:

• editable and non-editable nodes;
• surviving versions, archives, and republications;
• semantic proximities likely to produce collision;
• strong, weak, contested, or residual attributes;
• republication chains in which secondary information ends up dominating the origin.

4. Relationship with EAC and governed negation

• The graph identifies active nodes and reconstruction tensions.
• EAC qualifies which external authorities may actually constrain interpretation.
• Governed negation bounds unresolved conflicts or non-editable surfaces.
• The Q-Layer keeps the final decision to authorize, suspend, or refuse a response.

5. Associated doctrinal deliverables

Work on the graph may produce, depending on context:

• a mapping of active nodes;
• an external attribute matrix;
• a list of collisions, conflicts, or drifts;
• an admissibility qualification via EAC;
• a register of negations or arbitrations when a conflict persists.

6. What this page establishes

• External mapping is necessary but not sufficient.
• Visibility is not admissibility.
• Activity is not authority.
• The graph does not arbitrate; it makes the necessary arbitrations visible.

7. Why the graph matters operationally

The external coherence graph is not an optional diagnostic. It is a prerequisite for any serious governance of an entity’s interpretive environment. Without it, the entity governs its own canonical surface but remains blind to the forces that actually shape its reconstruction in AI outputs.

Timing. The graph must be established before remediation begins. Attempting to correct distortions or reduce interpretive debt without first mapping active sources leads to misdirected effort: corrections may target nodes that are not active in the reconstruction, while the actual sources of contamination remain unaddressed.

Scope. The graph extends beyond the first ring of sources. Republication chains, aggregator surfaces, and cached versions often carry more weight in LLM training data than the original. A source that appears dormant on the live web may be the dominant version inside a model’s context window. This is why the mapping must include archives, residual temporalities, and surviving authority — not only what is current, but what persists.

Connection to governance layers. The graph feeds into multiple governance mechanisms. It informs EAC decisions about which authorities to recognize or reject. It provides the raw material for interpretive collision detection. It supports the Q-Layer by revealing when a response would necessarily rely on contested or contradictory inputs.

Evolution. The graph is not static. Sources appear, disappear, and change. Attributes drift. Republication chains extend. The graph must therefore be maintained as a living instrument, updated when the entity’s external environment shifts. A fossilized graph is worse than no graph at all, because it creates false confidence in a mapping that no longer corresponds to the active reconstruction surface.