RFC-035: Mediation Vector Profile — A Validity-Aware Measurement Model for Domains of Concern — 17. Design Principles
AIGP Specification › RFC-035: Mediation Vector Profile — A Validity-Aware Measurement Model for Domains of Concern › 17. Design Principles
← 16. Temporal Model · Section index · 18. Non-Goals →
17. Design Principles
17.1 Variables are not scores
A variable is a measurement. A score is a judgment. The Mediation Vector provides the former; the Domain of Concern defines how to derive the latter.
17.2 Four states prevent circularity
By distinguishing anticipated from actual from evidenced from perceived, the system cannot claim “what happened is what should have happened.” Each state is populated from a different source at a different time.
17.3 Gaps demand verdicts, not conclusions
A gap signal does not assert failure. It asserts that a qualified observer must render a verdict. The gap is the question, not the answer.
17.4 Evidence admissibility prevents false confidence
A variable populated from unsigned claims carries different weight than one populated from D-DNA signed evidence. The admissibility system makes this distinction explicit and enforceable.
17.5 Extraction methods enable progressive automation
Start with human_assessed extraction (Level 4). Calibrate model_derived extraction (Level 3) against human baselines. Graduate to deterministic extraction (Level 1–2) where the domain permits. The same variable structure supports all automation levels.