Evidence graph for AI R&D and assurance
A concept note scoping whether the evidence graph in ZAN-RN-011 can unify internal R&D evidence records with the public deployment-readiness evidence pack.
Whether the evidence graph schema from ZAN-RN-011 can be extended with R&D-specific fields (systematic progression, technical uncertainty framing) without breaking its use as a general-purpose assurance lineage record.
- Extended schema with optional R&D metadata
- Single-source-of-truth lineage test
- Comparison against maintaining two separate records
Extend the ZAN-RN-011 evidence node schema with optional systematic-progression fields and test whether the same graph can answer both an R&D evidence query and a deployment-readiness query without duplicated records.
A single evidence graph schema can serve both purposes if R&D-specific fields are added as optional metadata on existing node types rather than as a parallel schema.
One evidence graph, two audiences
Internal R&D evidence and public deployment-readiness evidence currently risk becoming separate record systems. This note scopes whether the ZAN-RN-011 evidence graph schema can serve both without duplication.
Related research
View allEvidence graphs for reproducible AI assurance
Can AI assurance evidence be represented as a queryable graph from hypothesis to experiment to result to decision?
A measurement model for AI deployment readiness
How should teams measure whether an AI system is ready for controlled production deployment?
RAG evaluation under domain shift
How should retrieval-augmented generation be evaluated when source documents, user questions, and domain language shift over time?