Type: Protocol
Editorial status: Working
Epistemic status: Operational research architecture
Evidence scope: Project sources and page-specific references; cross-links do not supply independent corroboration
Last reviewed: 2026-09-16 — AI-assisted editorial check
Human reviewed: Human content-review pending for this revision
Current repository version: v0.3.2
CIP-KGE is the Cognitive Interview Protocol for Knowledge Graph Evolution.
Its current form is better understood as an epistemic governance protocol than as an autonomous interview agent.
The distinction matters.
The protocol is designed to turn tacit expert knowledge into bounded, traceable, human-reviewable proposals.
It is not designed to let an LLM update a knowledge graph by itself.
In the current architecture, AI assists with a limited extraction/classification stage.
The knowledge graph is not writable by the model.
Humans remain responsible for:
The protocol is intentionally slower than free-form AI synthesis.
That is a trade-off, not a defect.
CIP-KGE is a correctness mechanism, not a throughput mechanism.
The protocol currently requires:
Every proposed change must point back to specific evidence.
Updates should be decomposable into bounded local modifications.
Raw statements and derived conclusions must not be silently merged.
One patch should change one conceptual dimension at a time.
Uncertainty is structural metadata, not an optional disclaimer.
Rejected diffs, weak evidence, and failed interviews remain part of the record.
The formal output is a Knowledge Diff.
A diff is not an edit.
It is a structured proposal that says:
A human can accept, reject, defer, or arbitrate the proposal.
The point is not to automate expertise extraction.
The point is to make the movement from human experience to public knowledge inspectable.
Official repository:
https://github.com/pyragogy/Cognitive-Interview-Protocol
Evidence & Revision · Knowledge System · Handbook map · Home