RFC-035: Mediation Vector Profile — A Validity-Aware Measurement Model for Domains of Concern — 12. Extraction Methods

AIGP Specification › RFC-035: Mediation Vector Profile — A Validity-Aware Measurement Model for Domains of Concern › 12. Extraction Methods

← 11. Evidence Admissibility · Section index · 13. Minimal Mediation Vector: Autonomous Systems →

12. Extraction Methods

Each variable declares how its value is extracted from evidence:

Extraction Type	Description	Automation Level
deterministic	Direct computation from structured evidence (boolean logic, arithmetic)	Fully automated
rule_based	Pattern matching or threshold application against evidence fields	Fully automated
model_derived	LLM-as-judge or classifier applied to unstructured evidence	Semi-automated (requires calibration)
human_assessed	Human observer assigns value based on professional judgment	Manual
composite	Combines multiple extraction methods with declared weights	Mixed

8.1 Extraction Method Declaration

{
  "extraction_method": {
    "type": "composite",
    "components": [
      {"type": "deterministic", "weight": 0.6, "logic": "scope_envelope.decision == ALLOW"},
      {"type": "rule_based", "weight": 0.2, "logic": "budget_remaining.actions > 0"},
      {"type": "human_assessed", "weight": 0.2, "source": "reviewer_authority_rating"}
    ],
    "aggregation": "weighted_mean"
  }
}

---

---

← 11. Evidence Admissibility · Section index · 13. Minimal Mediation Vector: Autonomous Systems →