Skip to main content

PTI and Knowledge Graphs

Knowledge graphs represent entities and relationships — people, organizations, products, events — as nodes and edges for search, analytics, and inference. PTI's internal trust graph shares graph thinking but adds normative trust semantics: context isolation, provenance chains, consent gates, and consumer lookup contracts.

1. What knowledge graphs are

Knowledge graphs (KGs) are semantic data structures — often RDF, property graphs (Neo4j, Neptune), or vendor analytics graphs — that enable:

  • Entity resolution — link records across datasets
  • Relationship traversal — multi-hop queries (who knows whom, who owns what)
  • Inference and reasoning — ontologies, rule expansion, GNN embeddings
  • Enterprise search — unified discovery across siloed systems
  • Fraud and intelligence analytics — network exposure and community detection

Knowledge graphs answer: What entities exist, how are they connected, and what can we infer from the graph structure?

2. What problem knowledge graphs solve

ProblemKnowledge graph response
Data silos in enterprise analyticsUnified entity-relationship model
Complex dependency tracingMulti-hop graph queries
Recommendation and similarityGraph embeddings and paths
Investigative analysisLink chart visualization

Knowledge graphs excel at analytics and discovery. They typically lack normative trust contracts — consent-bound signal ingestion, context-scoped scoring APIs, and governed cross-institution exchange profiles.

3. What PTI adds

General knowledge graphs

  • Flexible entity-relationship modeling
  • Analytics and inference workloads
  • Internal enterprise scope

PTI adds

  • Trust graph model — normative subject, signal, and context edges
  • Context-scoped propagation — signals do not leak across life areas
  • Provenance on every edge — attributable trust events (RFC-012)
  • Lookup API contract — decision-time export, not ad-hoc graph queries

RFC-005 — Trust Graph defines PTI's graph semantics — not as a general-purpose ontology, but as a trust-specific subgraph with enforcement rules institutions can rely on for compliance.

4. How they compose together

Integration patterns:

  1. Side-by-side — enterprise KG for internal analytics; PTI trust graph for cross-institution portable trust. Entity resolution hints may flow one direction; PTI does not expose raw graph traversal to consumers.
  2. PTI as governed subgraph — institution deploys PTI-compatible registry; internal KG syncs derived features from trust lookup responses, not full partner graphs.
  3. Fraud analytics — enterprise KG runs link analysis; confirmed outcomes emit trust events into PTI for durable institutional memory.

PTI restricts open graph query access by design — consumers receive trust intelligence envelopes per RFC-004 — Trust Lookup API, preventing uncontrolled graph exfiltration.

5. When to use each

ScenarioKnowledge graphPTI trust graph
Internal AML link analysisKG idealEmit outcomes as signals
Cross-MFI portable trustKG alone insufficientPTI Required
Marketing customer 360KGNot applicable
Institution trust lookup at decisionAd-hoc queries riskyPTI lookup API
Research ontology developmentKGPTI profiles subset

Use knowledge graphs for analytics breadth; use PTI for governed trust depth with interoperability guarantees.

See also