P201 — AIEP — Entity State Registry

Applicant: Neil Grassby Classification: Patent Application — Confidential Priority: Claims priority from GB2519711.2 filed 20 November 2025 Architecture Layer: AIEP AGI Cognition Layer — Phase 2 Dependencies: P200 — Causal World State Graph

Framework Context

[0001] This specification operates within an Architected Instruction and Evidence Protocol (AIEP) environment as defined in United Kingdom patent application number GB2519711.2, filed 20 November 2025, and GB2519798.9, filed 20 November 2025.

[0002] The present specification defines a canonical mechanism for deterministic entity identification and state tracking within the Phase-2 AIEP world modelling architecture, extending the provenance model of Phase-1 to persistent entity identity across multiple evidence artefacts.

Field of the Invention

[0003] The present invention relates to artificial intelligence reasoning systems and knowledge representation architectures.

[0004] More particularly, the invention relates to deterministic mechanisms for identifying entities and maintaining persistent state records derived from evidence artefacts within an AIEP environment.

Background

[0005] Knowledge representation systems frequently require the identification and tracking of entities such as persons, organisations, locations, and objects. Existing knowledge graph systems typically rely on heuristic identity matching or probabilistic entity linking techniques.

[0006] Such approaches introduce inconsistencies including duplicate entity representations, incorrect entity merging, loss of provenance, and inability to deterministically reconstruct entity identity.

[0007] The Phase-1 AIEP architecture provides evidence provenance, deterministic replay, and reasoning traceability. However, Phase-1 architectures do not define a canonical mechanism for maintaining persistent identity for entities referenced across evidence artefacts.

[0008] Without such a mechanism, world modelling systems such as the Causal World State Graph (P200) cannot maintain consistent entity representations across evidence-ingestion cycles.

[0009] There therefore exists a need for a deterministic entity registry capable of tracking entity identity and state evolution across evidence records.

Summary of the Invention

[0010] The invention provides an Entity State Registry (ESR) capable of maintaining deterministic identifiers for entities referenced in evidence artefacts. The registry assigns a stable identifier to each entity and records state attributes associated with that entity.

[0011] Entity states are derived exclusively from admissible evidence artefacts recorded within the AIEP evidence ledger. State transitions are preserved as historical records rather than overwriting previous states, maintaining a complete temporal history.

[0012] The ESR assigns entity identifiers deterministically. For a given set of input signals — name, classification, first-seen evidence hash — the same entity identifier is always produced, enabling cryptographic verification of entity identity.

[0013] The invention therefore enables deterministic reconstruction of entity identity and state across time, supporting the CWSG (P200) and hypothesis simulation (P204) with consistent entity semantics.

ASCII Architecture

Evidence Artefact
       |
       v
+-------------------------------+
| Entity Extraction Subsystem   |
+---------------+---------------+
                |
                v
+-----------------------------------------------+
|           Entity State Registry (ESR)          |
|                                                |
|  EntityID = hash(name + class + origin_hash)  |
|                                                |
|  [EntityID] --> [StateRecord_t0]               |
|             --> [StateRecord_t1]               |
|             --> [StateRecord_tN]  (current)    |
+-----------------------------------------------+
                |
                v
     CWSG Entity Nodes (P200)

Detailed Description

[0014] Entity Identifier Derivation. The primary entity identifier is derived as a SHA-256 hash over the concatenation of: entity canonical name string; entity classification type; and the SHA-256 content hash of the first evidence artefact in which the entity was referenced. This derivation ensures that the same entity always produces the same identifier regardless of subsequent state changes, and that the identifier is verifiably linked to the originating evidence artefact.

[0015] State Record Structure. Each state record in the ESR contains: entity_id (derived identifier); state_seq (monotonically increasing ordinal); attributes (key-value map of entity attributes at this state); evidence_hash (content hash of the evidence artefact from which this state was derived); and recorded_at (ISO 8601 timestamp of admission).

[0016] State Transition Protocol. When a new evidence artefact references an existing entity with changed attributes, a new state record is appended to the entity’s state chain. The previous state record is never modified. The current state of an entity is always the state record with the highest state_seq.

[0017] Cross-Artefact Entity Resolution. The ESR matches entity references across evidence artefacts by comparing canonical name strings and classification types. When a match is found, the existing entity identifier is used rather than creating a new entity. Name canonicalisation applies standard normalisation rules (case folding, whitespace normalisation, common abbreviation resolution).

[0018] ESR Query Interface. The ESR exposes: get_entity(entity_id) returning the current state; get_state_history(entity_id) returning the ordered state chain; find_by_name(name) returning matching entity identifiers; and get_evidence_chain(entity_id) returning all evidence artefact hashes referenced in the entity’s state history.

[0019] Integration with CWSG. The ESR is the authoritative source of entity identity for the CWSG. All entity nodes in the CWSG reference the ESR entity identifier rather than maintaining independent identity state. This ensures that the CWSG and the ESR remain consistent without replication of entity state.

Technical Effect

[0020] The invention provides a deterministic, cryptographically verifiable entity identity system for evidence-bound AI, ensuring that references to the same real-world entity across multiple evidence artefacts always resolve to the same identifier regardless of admission order. By deriving entity identifiers from a SHA-256 hash over canonical name, classification type, and origin evidence hash, the system makes entity identity independently reproducible and verifiable. The append-only state chain ensures complete temporal entity state history is preserved for audit and reconstruction, while integration with the CWSG as the authoritative identity source prevents duplicate entity representations in world state modelling.

Claims

1. A method of deterministic entity identification and state tracking in an evidence-bound artificial intelligence system, comprising the steps of: (a) on extraction of an entity reference from an admitted evidence artefact, computing a primary entity identifier as SHA-256(canonical_name || classification_type || origin_evidence_hash), where origin_evidence_hash is the content hash of the first evidence artefact in which the entity was referenced; (b) checking the registry for an existing entry matching the derived identifier; if an entry exists, appending a new state record to the entity’s state chain; if no entry exists, creating a new registry entry; (c) constructing the state record as: entity_id (derived identifier), state_seq (monotonically increasing ordinal), attributes (key-value map from the evidence artefact), evidence_hash (content hash of the admitting artefact), recorded_at (ISO 8601 timestamp); (d) never modifying existing state records; resolving the current entity state as the record with the highest state_seq; (e) providing query operations: get_entity returning current state, get_state_history returning ordered state chain, find_by_name returning matching identifiers, and get_evidence_chain returning all evidence hashes in an entity’s state history.

2. The method of claim 1, wherein name canonicalisation applies case folding, whitespace normalisation, and common abbreviation resolution to enable cross-artefact entity matching.

3. The method of claim 1, wherein the entity identifier derivation at step (a) is deterministic, such that independently processing the same set of evidence artefacts always produces the same registry contents in the same order.

4. The method of claim 1, wherein the CWSG references ESR entity identifiers for all entity nodes, with the ESR as the authoritative identity source, preventing duplicate entity representations.

5. The method of claim 1, wherein get_evidence_chain returns the complete ordered list of evidence artefact content hashes from first_seen to current state, enabling full provenance tracing for any entity.

6. An entity state registry for an evidence-bound artificial intelligence system, comprising: an entity extraction subsystem parsing evidence artefacts for entity references; a deterministic identifier derivation module; an append-only state chain store; a cross-artefact entity resolution module applying name canonicalisation; and a query interface.

7. A computer-readable medium carrying instructions for implementing the method of any preceding method claim.

Abstract

An entity state registry for evidence-bound artificial intelligence assigns stable deterministic identifiers to entities referenced in evidence artefacts, deriving each identifier as SHA-256 over canonical entity name, classification type, and origin evidence hash. State records are appended to an immutable state chain on each new artefact reference, preserving complete temporal state history. Current entity state is always the highest-sequence record. A query interface provides current state, full state history, name-based search, and complete evidence chain retrieval. The registry serves as the authoritative entity identity source for the CWSG, preventing duplicate entity representations.