External coherence graph
The external coherence graph designates the mapping of public signals (sources, mentions, entities, relations, attributes) that frame how an entity is interpreted by AI systems in the open web. It allows identifying where the narrative is coherent, where it is contradictory, and where it is vulnerable to contamination, collision, or capture.
In interpretive governance, exogenous “truth” is not a single text. It is a graph: relations between sources, entities, and attributes that produce a dominant interpretation.
Definition
External coherence graph is the structured set linking:
- the entity (brand, person, concept, organization);
- external sources (articles, directories, wikis, profiles, citations, aggregators);
- attributes (description, offering, positions, dates, categories, promises);
- relations (belonging, filiation, synonymy, opposition, competition, homonymy).
The graph is called “coherence” graph when it allows measuring signal compatibility among themselves and with the endogenous canon. It reveals zones where AI risks reconstructing an unstable narrative.
Why this is critical in AI systems
- The open web is the training environment: AI aligns on dominant external patterns.
- Secondary sources weigh heavily: they often stabilize a default interpretation.
- Conflict is invisible: weak contradictions produce interpretive debt and inertia.
What the graph allows detecting
- External authority conflicts: incompatible strong sources.
- Neighborhood contaminations: reframing by dominant clusters.
- Entity collisions: homonymies, acronyms, confusions.
- Invisibilization: endogenous canon absent from external activation paths.
- Capture: dominant framing imposed by saturation and vocabulary.
Practical indicators (symptoms)
- AI systems explain the entity based on external sources that contradict each other.
- The internal canon is ignored in responses (interpretive invisibilization).
- Attributes from a neighbor are projected onto the entity (contamination).
- An internal correction does not change the external interpretation (inertia / remanence).
What the external coherence graph is not
- It is not a list of backlinks. It concerns semantic relations and interpreted attributes.
- It is not a classic SEO audit. The objective is interpretation stability.
- It is not a single truth. It is a mapping of the forces that shape AI output.
Minimum rule (enforceable formulation)
Rule ECG-1: any entity that claims a stable interpretation in the open web must map and monitor its external coherence graph, by identifying dominant sources, contradictions, homonymies, and capture zones. Without an external graph, exogenous governance remains blind.
Example
Case: a brand is described differently on a directory, a profile, an article, and a wiki. AI systems produce a hybrid synthesis.
Diagnosis: incoherent external graph (contradictions + contamination + dominant secondary sources).
Expected correction: realign dominant sources, reinforce the internal canon, publish governed negations, and reduce confusion zones.