Abstract
A database stores records. An institution is not a set of records; it is a structure of relationships between entities, authorities, obligations, and dependencies. This note outlines a working framework for representing organizations and the environments around them as explicit, machine-readable graphs, rather than collections of unconnected tables.
The problem with records
Most enterprise and government systems store entities as isolated records: a company in one system, a permit in another, an agency in a third. The relationships between them — ownership, dependency, authority, obligation — exist in practice but rarely in any queryable form.
This produces a familiar failure mode: an organization can retrieve any individual fact quickly, but cannot answer questions that span several of them without a person manually assembling the picture.
Structure as a first-class object
An institutional graph treats relationships — not just entities — as data. A company owns a facility. A facility requires a permit. A permit is governed by an authority. An authority is constrained by a regulation. Each edge carries meaning, provenance, and a valid time range.
This structure is what allows a system to answer questions no single record can answer alone: which facilities are exposed if a given regulation changes, which suppliers sit upstream of a critical component, which approvals must precede a given action.
Toward a working framework
Our current thesis organizes institutional graphs around four primitives: entities (who or what exists), relationships (how they connect), constraints (what governs them), and provenance (how the system came to know any of it).
This is a working framework, not a finished ontology. We expect it to change as it is applied to real institutional environments across enterprise and government contexts.
Status note
This entry is published as a working thesis. It reflects current thinking inside Solomonic Intelligence and is expected to change as it is tested against real institutional environments.