Identity is one of the first objects a generative system fuses. Similar names, adjacent roles, and overlapping pages are enough to produce composite entities that never existed canonically.
Operational definition
A governed identity graph is the explicit structure of entities, roles, relationships, and perimeters that allows an AI system to distinguish who or what is being referred to, under which role, and in relation to which other object.
Why identity must be modeled relationally
Identity drift rarely comes from a single false attribute. It usually emerges from missing relations: a founder becomes the company, an author becomes the service, a product becomes the organization. A governed graph prevents these fusions by making relations as important as the entities themselves.
What the graph must declare
- Core entities: person, organization, offer, product, doctrine, and publication surfaces.
- Roles: founder, author, operator, publisher, provider, and other contextual functions.
- Relations: creates, publishes, belongs to, is not, does not imply, replaces, or depends on.
- Perimeters: what each entity covers and what remains outside its scope.
- Anti-fusion boundaries: explicit negations for the most likely identity collisions.
Operational use
- Build relation-rich pages instead of multiplying flat biographies and about pages.
- Keep role statements distinct from entity definitions.
- Declare what a relation does not imply, not only what it does imply.
- Align schema, navigation, and page copy around the same graph logic.
- Audit where the model still fuses entities despite explicit relation statements.
What this graph prevents
- Founder / organization / offer fusion.
- Authorial attribution drift.
- Role inflation caused by semantic proximity.
- Identity confusion across adjacent pages and off-site mentions.
