RFC-035: Mediation Vector Profile — A Validity-Aware Measurement Model for Domains of Concern — 1. Abstract
AIGP Specification › RFC-035: Mediation Vector Profile — A Validity-Aware Measurement Model for Domains of Concern › 1. Abstract
Section index · 2. Epistemological Foundation: What We Learn from Science →
RFC-035: Mediation Vector Profile — A Validity-Aware Measurement Model for Domains of Concern
PRIVATE AND PROPRIETARY — NOT A PUBLIC RFC. Owned by Kanjani AI Research & Causum. See NOTICE.md.
Status: Draft — conceptually stable Category: Empirical Evaluation Infrastructure Series: Mediated Intelligence Quality and Concern Evaluation Depends on: RFC-034 (Domains of Concern), RFC-032 (Post-Hoc Evaluation Loop), RFC-033 (Quantitative Outcome Evaluation), AIGP Core Defines: Mediation Vector Profile (validity-aware measurement model), Construct Declaration, Four-Layer Profile, Four-State Model, Gap Calculus, Measurement Typing, Evidence Admissibility, Provenance Requirements, Warrant Grounded in: Construct validity (Cronbach & Meehl), validity theory (Messick), measurement scales (Stevens), Goal-Question-Metric (Basili), provenance (W3C PROV), argumentation (Toulmin)
1. Abstract
This RFC defines the Mediation Vector Profile — a typed, multi-state, evidence-sourced variable structure that enables empirical concern calculation for governed AI artifacts within a declared Domain of Concern.
A Mediation Vector is the evaluative substrate on which VERIFY verdicts, Quality Moderator posture, and concern calculations operate. Without it, evaluation is opinion. With it, evaluation is calculable.
The central claim, borrowed from the Master Evidence Graph for Software Verdicts and applied to mediated intelligence:
A mediation concern cannot be calculated until its variables are declared, its evidence is sourced, and its states are distinguished.
This RFC establishes the variable model, the four-state architecture, the gap calculus, the evidence admissibility requirements, and the extraction method declarations that together make mediated intelligence concern posture empirically computable.
Section index · 2. Epistemological Foundation: What We Learn from Science →