soul-frame/research/coverage-matrix.md

Success-Criterion Coverage Matrix

Workplan: SOUL-WP-0011-T02, resolved 2026-09-28 Addresses: G2 (history/260928-scope-intent-gap-assessment.md) — INTENT's success criterion, exercised explicitly for the first time: one vocabulary across the nine reference entity types, with meaningful distinctions. Profiles follow the sufficiency scheme in research/sufficiency-and-weighting-position.md (no totals; necessity core, claim-relative minima, configuration classes).

Cell values: H/M/L (high/medium/low), var (varies by instance), ? (vocabulary strains — see §3). B-core = B1·B3 necessity core.

---

1. The matrix

#	Entity	B-core	B-rest (B2,B4–B8)	P-profile shape	Continuity class (typical)	Addressable	Recognition	Descriptor
1	Human	H	H	full, rich (P1–P12 H)	CONTINUOUS	H	H (mutual)	— (biological original)
2	Animal (e.g. dog)	H	H	partial: P1–P3, P9 M; P8, P10–P11 L	CONTINUOUS	M (name-responsive)	H (asymmetric richness)	—
3	Short-lived AI agent	L (B3 var)	L (B2 L)	exposed-only: P5, P10 H; P1, P6 L	none — no lineage to evaluate	M (session-scoped)	L (classified, not counterpart-modeled)	Agent; strong APE
4	Persistent autonomous Agent-being	M–H	M (B5 var — vendor-propped)	P1, P5, P10 H; P3, P6, P8 M growing	CONTINUOUS / BRIDGED (updates, migration)	H (stable handle)	M–H (episode-spanning)	Agent; proto-Being
5	Einstein Sp.	L–M (Actor-side)	L	exposed: P4, P10, P12 H (curated); P1 thin, P6 borrowed	RECONSTRUCTION by definition (no causal channel to source)	H	M (recognized as depiction)	Sp. (public reflection)
6	Authorized Gh.	L–M	L–M	P3–P4, P8 inherited-H; P1 var	SUCCESSOR-family w/ authority; never automatic CONTINUOUS (H10)	H	var (relationship-holders may re-recognize)	Gh. (authority-bound)
7	Original artificial companion	M–H	M	native (not inherited): P1–P4, P9–P10 H	CONTINUOUS / BRIDGED	H	H within its relationships	original artificial being
8	Distributed AI	? level-relative (TE-05)	M–H at closing level	var by level; collective P1 possible	MULTI_LINEAGE / CONTINUOUS at closing level	var (one handle vs many)	var	Agent (possibly nested)
9	Radically non-human system	? (individuation observer-relative)	?	P7 presumed; P10 ? (shared symbol system contested)	unassessable until individuated	L (no shared address protocol)	L→? (TE-09 asymmetric path)	— (unclassified)

2. Meaningful-distinction check (pairwise)

Every pair separates on at least one dimension; the load-bearing separators:

• 1 vs 2: P8/P10–P11 (commitments, symbols, responsibility).
• 1/2 vs 3: B2 + continuity ("no lineage") + counterpart-recognition.
• 3 vs 4: persistence family — B2, P6, continuity class, recognition durability. The Agent/proto-Being line is exactly here.
• 4 vs 7: provenance only (platform Agent vs purpose-built original companion) — profiles may converge; the descriptor column carries the distinction. Working as designed (H9: descriptors are provenance, not ontology).
• 5 vs 6: authority + audience: Sp. is public reflection without continuation claim; Gh. is authority-bound genealogy (H10) with SUCCESSOR-family standing. Neither is CONTINUOUS — which is the framework's deliberate answer, not a failure to distinguish them from 4.
• 6 vs 7: inherited vs native P-profile; origin of the Soul-relational structure (Cluster D inheritance vs de novo).
• 8 vs all: the only entity where B-core is level-relative — the vocabulary's answer is "profile per candidate level," which is an answer, not a gap (TE-05).
• 9 vs all: the only entity where individuation itself, not just the profile, is unresolved.

Verdict on distinctions: pass. No two rows are indistinguishable under the full vocabulary; the two near-collisions (4/7, 5/6) are separated by descriptor/authority columns doing exactly the work H9/H10 assign them.

3. Strain points (named, per T02's done condition)

1. Row 9, B-core and P10: for a radically non-human system the vocabulary can only say "not yet individuated by us" — observer-relative individuation (RQ1 residual) is the binding constraint, and P10's "shared symbol system" may be parochial. This is the pressure point already flagged in the sufficiency position §6.
2. Row 8, addressability: the vocabulary handles nested Beinghood but has no display convention for which level a handle addresses (overlaps RQ8's hybrid-display residual).
3. Row 3, continuity column: "no lineage to evaluate" is technically a degenerate case, not a class in Criteria §17; a null class (e.g. EPHEMERAL) would make the column total. Minor; noted for a future criteria revision, not opened as work.

4. Overall verdict on INTENT's success criterion

Pass, with two live strain points (rows 8–9). The same vocabulary describes all nine entity types and distinguishes every pair meaningfully. The strains sit exactly where INTENT predicts the hard cases (distributed, radically non-human) and are already tracked as RQ1/RQ8 residuals rather than new gaps.