Explainability

Most RAG frameworks are black boxes. You ask a question, you get an answer, and you have no idea why. TrustGraph takes a fundamentally different approach: every answer is traceable.

Why this matters

When an AI system gives you an answer, you need to be able to ask:

  • Which documents contributed to this answer?
  • Which facts were selected from the knowledge graph?
  • Why were those facts chosen over others?
  • What reasoning led to the final response?

In a demo, this doesn’t matter. In production — especially in regulated industries like financial services, healthcare, or legal — auditable AI reasoning is not optional. Even outside regulated contexts, explainability builds confidence and helps identify when the system is drawing on incorrect or outdated information.

How TrustGraph does it

TrustGraph records two layers of provenance:

Extraction provenance tracks how knowledge entered the system. When documents are processed, TrustGraph records the full derivation chain: document → pages → chunks → extracted knowledge. Given any fact in the knowledge graph, you can trace it back to the exact chunk of text it was extracted from.

Query-time explainability tracks how an answer was derived. When a query runs, TrustGraph records each stage: what concepts were extracted from the question, which knowledge graph nodes were visited, which relationships were selected and why, and how the final answer was synthesised.

The complete audit trail

The real power comes from connecting these two layers. For any answer, you can follow the full chain:

Your question → concept extraction → graph traversal → relationship selection (with reasoning) → specific knowledge → extraction chain → chunk → page → source document

Every link in this chain is recorded, queryable, and persistent. Reasoning traces aren’t ephemeral — they’re stored as standard RDF triples and remain available for later review, auditing, or reporting.

Deeper dive

For the full technical details on how explainability works — including named graphs, context graph architecture, and how extraction provenance connects to query-time traces — see the Explainability overview.

Next

Ontologies — how domain-specific schemas give you precision control over knowledge extraction.