SIP-RPE-001 · PROPOSED EXPERIMENTAL PROTOCOL

Representation Perturbation Experiment

One question: How does a bounded change in representation propagate through a Scientific Intelligence knowledge architecture?

Why RPE exists

Scientific systems do not observe Reality directly; they observe through representations, classifications and relationships. RPE makes that mediation experimentally visible by changing one bounded representation layer and measuring downstream structural response.

10-step protocol

1. FreezeBaseline population + version + checksum/manifest

2. Define perturbationExplicit inclusion rule, changed field, old/new representation, scientific rationale

3. Clone branchExperimental population; canonical unchanged

4. Apply perturbationChange ledger with object IDs and state=AI_PREREVIEWED

5. Regenerate relationshipsBefore/after typed routes

6. Regenerate journeysBefore/after aggregate journey signatures

7. Measure sensitivityLayer rates, propagation ratios, RSI candidate metric

8. Inspect amplificationLargest downstream changes and affected claims/capability candidates

9. HI/domain reviewKEEP / REVISE / REJECT / NEED_MORE_EVIDENCE

10. Reality ReturnCanonical update proposal + lesson passport + reproducible audit

Non-negotiable boundary: experimental relabelling does not rewrite the canonical population. AI prereview proposes; HI/domain review decides whether a canonical change is justified.

Measurement family

Record Exposure · Classification Edit Rate · Relationship Turnover · Journey Sensitivity · Propagation Ratio · RSI v0.1

Later versions may add Signature Journey sensitivity, Claim sensitivity and Capability-candidate sensitivity once those promotion rules are reproducible.

First case

CIDM 2,231 is the first completed laboratory case: early-WGS and historical COVID/respiratory combined-label perturbations. The experiment changed representation in a cloned branch and measured the resulting relationship and Journey changes.

Case 001 → · RSI v0.1 → · Colleague Beta →