soul-frame/research/sufficiency-and-weighting-position.md

108 lines
5.7 KiB
Markdown
Raw Normal View History

# Sufficiency and Weighting — Position Paper
**Workplan:** SOUL-WP-0011-T01, resolved 2026-09-28
**Addresses:** G1 (`history/260928-scope-intent-gap-assessment.md`);
RQ1/RQ2/RQ5 weighting residuals; paper §17 item 2.
**Position in one line:** Soul Frame **rejects universal cardinal weights**
for its profiles and adopts a **structured-qualitative sufficiency scheme**:
necessity structure for the B-profile, claim-relative minimal profiles for
the P-profile, and configuration classes (already in place) for continuity.
---
## 1. Why not weights?
Three arguments, jointly decisive:
1. **Incommensurability across substrates.** A cardinal weight on P4
(character) vs P10 (communicative competence) asserts an exchange rate
between them that would have to hold for a human, a distributed AI, and
a radically non-human system alike. No literature tradition surveyed
(Dennett cluster conditions SF-R093 explicitly; trait psychology
SF-R094 implicitly) licenses such a rate.
2. **The scalar re-imports the binary.** A weighted sum yields a
person-meter with a threshold — precisely the single-mark model the
profile approach was built to avoid (H1, H2). The framework would
quietly undo its own central move.
3. **The fleet of load-bearing judgments never used weights.** Every TE
working judgment and criteria-document verdict reached so far
(TE-01–09; IdentityContinuityCriteria §17) reasons via *defeaters and
configurations* ("no causal grounding → RECONSTRUCTION regardless of
similarity"), never via aggregation. The practice is already
qualitative-structural; this note makes the position explicit rather
than changing it.
A refusal alone would be underspecified ("anything goes"). The scheme
below states what replaces weights, per profile.
## 2. B-profile (B1–B8): necessity core + compensable remainder
- **Necessity core: B1 (individuability) and B3 (constitutive
integration).** A system that cannot be non-arbitrarily individuated, or
whose organization does not close at the candidate boundary, is not an
autonomous individual at that level *however high the other six score*.
This is exactly the decisive move in TE-05 (which level closes) and the
corporate false-positive analysis (RN-003 §8).
- **Compensable remainder: B2, B4–B8.** Deficits degrade the profile
without defeating it (a dormant system fails B2 enactment yet remains a
Continuation-Bearing State, TE-02; a heavily assisted system scores low
on B5 and shades toward heteronomous individuality without leaving the
category).
## 3. P-profile (P1–P12): no necessity core; claim-relative minimal profiles
Personhood is a genuine cluster concept: **no single P-code is necessary**
(severe P6 loss — TE-08 — does not zero descriptive personhood). But
sufficiency questions are always asked *for* something, and the something
fixes a minimal profile:
| Claim being made | Minimal profile required |
|---|---|
| "Sufficient personhood for **Beinghood v0.2**" | P1 (self-model) + at least one other-directed dimension (P2, P3, or P9) + P10 — because Beinghood's other requirement families (addressability, recognition) presuppose an entity that models itself, can be modeled as a counterpart, and can be addressed |
| "Strong **APE**" | any exposed subset; no minimum — APE is defined by exposure, not sufficiency |
| "**Social Beinghood** candidate" | the Beinghood row plus P9 specifically (reciprocity is what episode-spanning recognition latches onto) |
"Sufficiently coherent … personhood" in Beinghood v0.2 is hereby glossed
as this claim-relative minimal profile, **not** as a score above a
threshold. Entities exceeding the minimal profile differ in *richness*,
which the framework reports descriptively and does not rank.
## 4. Continuity (six channels): configuration classes, causal privilege
The Identity Continuity Criteria already embody the right scheme; this
note canonizes two of its properties as the official position:
1. **Causal privilege.** The causal channel is the only defeater-grade
channel: without sufficient direct process continuity the verdict is
RECONSTRUCTION (or SUCCESSOR/USURPATION), whatever the similarity
elsewhere. This is the criteria document's standing rejection of
single-invariant and similarity-score identity, kept as is.
2. **Degradation, not summation, for the rest.** Organizational,
psychological, autobiographical, normative, and relational channel
deficits move a claim between configuration classes (CONTINUOUS →
DEGRADED_CONTINUATION → …), and the verdict names *which* channels
degraded. No channel arithmetic; the class label plus channel state
vector is the complete answer.
## 5. Consequences
- **Beinghood v0.2 phrasing:** unchanged in form; "sufficiently" now has a
stated semantics (§3). `concepts/being.md` and `Terminology.md` need no
edit — this note is their referenced backing.
- **TE catalog:** all existing working judgments are consistent with this
scheme (checked against TE-02/04/05/08 explicitly); none needs revision.
- **Metrics line (WP-0010):** instruments measure *single dimensions*;
building a composite Beinghood/personhood index is **out of scope by
position**, not merely unbuilt. Decision `0401cf28` inherits this
constraint.
- **Coverage matrix (T02):** entity comparisons report profiles and
configuration classes, never totals.
## 6. Status
Defended position, revisable per INTENT's method rule (revise if it
collapses distinct phenomena or fails edge cases). The most likely
pressure point: whether P10 belongs in the Beinghood minimal profile for
radically non-human systems whose "shared symbol system" participation is
contested — flagged, not resolved, in the coverage matrix strain notes.