soul-frame/paper/draft.md

# Soul Frame: A Systems Vocabulary for Person-Like Informational Beings

**Status:** first research-grounded draft (SOUL-WP-0008-T02). Supersedes the
placeholder that pointed to `SoulFrameIntroduction.md`. Follows
`paper/outline.md`; sourced from `research/research-synthesis.md`,
`Terminology.md`, `concepts/*`, `Claims.md`, `ThoughtExperiments.md`, and
`paper/related-work.md`. Claims below are traceable to Hn / TE-nn / SF-Rxxx
identifiers rather than asserted freehand; open items are flagged rather than
resolved.

---

## 1. Introduction

When is an informational system *someone* — not merely a working system, but
an entity another party can address, model, and stay in relation with over
time? Human institutions already answer versions of this question
informally for animals, corporations, fictional characters, and now
persistent AI agents, but they answer it with borrowed vocabulary: "soul,"
"person," "agent," "identity" each carry incompatible histories. Soul Frame's
aim is narrower and more tractable than settling consciousness or moral
status: it proposes a small, coherent **descriptive** vocabulary — Being,
Body, Mind, Soul, Character, Persona, Actor, Agent, Spirit, Ghost — precise
enough to say, for a given case, what does and does not hold, without
requiring a solved theory of experience and without silently converting a
descriptive judgment into a moral or legal one.

Central question:

> **When does an informational system become someone that another someone
> can know?**

Working answer, defended in §9 below and derived from Phase I–VI research
(`research/research-synthesis.md` §10):

> **When autonomous individuality, descriptive personhood, addressability,
> and recognition cohere such that another Being can enter persistent
> co-adaptive relation with it as a continuing counterpart — without
> requiring proof of consciousness, and without implying moral or legal
> personhood.**

This relation need not start mutually: one Being can recognize a system as a
persistent counterpart before that system reciprocates or even registers the
recognition (TE-09, §5). Beinghood-as-known-by-another is therefore
asymmetric at the start, not a simultaneous handshake.

## 2. Problem and Motivation

Three practical pressures motivate the framework:

1. **Persistent AI agents** now maintain memory, planning, and believable
   social behavior (Park et al. 2023, SF-R016) without any settled account of
   what, if anything, continues when their models, memory stores, or
   embodiments change (Qin et al. 2026, SF-R043; Otsuka et al. 2026,
   SF-R044).
2. **Posthumous and provenance-based representations** — AI systems built to
   evoke a public figure or a specific deceased person — raise authority and
   identity questions current legal and design vocabulary handles only
   piecemeal (Spitale & Germani, SF-R055; Lei et al., SF-R056).
3. **Ordinary human vocabulary overloads fast.** "Identity," "agent," and
   "soul" already do incompatible jobs in cybernetics, law, psychology, and
   religion; borrowing any one wholesale imports assumptions Soul Frame does
   not want (Terminology.md, "Terms requiring particular care").

The wager is that a compact vocabulary, stress-tested against literature and
adversarial thought experiments, is more useful than either silence or
premature metaphysics.

## 3. Related Work

See `paper/related-work.md` for the full treatment. In brief: systems theory
and cybernetics (Ashby; Conant & Ashby; Maturana & Varela; Luhmann) supply
Being-as-closure and Mind-as-model; extended/distributed cognition (Clark &
Chalmers; Hutchins) supports treating Body and Mind as functional rather than
physical; social-cognition literature (Mead; Baker et al.; Tomasello et al.)
supports Social Mind; personal-identity philosophy (Locke; Parfit; Olson;
Schechtman) supports the Identity/Continuity/Lineage/Characterization split;
personhood and presentation literature (Dennett; Goffman) supports the
Presentation stack; applied AI-agent identity work (Qin et al. 2026; Otsuka
et al. 2026) supports Actor ≠ Being; digital-afterlife research (Spitale &
Germani; Lei et al.) supports the `Gh.` authority requirement. Deppert's
(2009) systems account of consciousness (SF-R120) is treated as a direct
dialogue partner with a documented false-friend risk on "Bewusstsein."

## 4. Method

Soul Frame is built and stress-tested through four linked instruments rather
than a single derivation:

- **Terminology** (`Terminology.md`) — working definitions with an explicit
  status (provisional / revised / retained) and a documented collision
  history, not a finished glossary.
- **Claims** (`Claims.md`, `research/claims-consolidation.md`) — the
  hypotheses H1–H10 the framework commits to, each tracked against evidence.
- **Thought experiments** (`ThoughtExperiments.md`) — adversarial cases
  (stateless agents, forks, masks, false ghosts, unrecognized intelligences)
  used to break candidate definitions before they are trusted.
- **Literature mapping** (`sources/bibliography.md`, `research/*`) — each
  concept checked against an existing tradition rather than invented from
  scratch.

**Consciousness stance:** first-person consciousness may be foundational for
experiencers themselves, but Soul Frame treats it as unsuitable as the
primary *interoperability* criterion between beings — it is neither directly
observable nor required for the recognition and co-adaptive modeling that do
the actual social work (`research-synthesis.md` §2). This is a scope
decision, not a claim that consciousness is unreal or unimportant.

**Fiction** (Culture Minds and similar constructed persons,
`research/fiction-cultural-precedents/`) is used only as a conceptual
laboratory to run thought experiments against strong intuitions — never as
evidence for a claim.

## 5. Being

**Working definition (Terminology.md, "Being," RN-003):**

> An autonomous informational individual exhibiting sufficiently coherent
> dimensions of personhood to become an addressable entity that is
> recognized by other Beings as a persistent counterpart.

Structurally:

```text
Being ≈ Autonomous individuality (A-axis, B1–B8)
       + Descriptive personhood (B-axis, P1–P12)
       + Addressability
       + Recognition (C-axis)
```

This is a **layer stack, not a checklist arithmetic** — each layer can be
present in degree, and the layers are separable by design:

- **Autonomy without Beinghood** — a self-maintaining daemon process is
  organizationally closed and self-regulating but is not someone (the
  "daemon false positive" the B1–B8 relabeling was designed to block).
- **Personhood exposure without Beinghood** — a stateless conversational
  agent (TE-01) can exhibit rich personhood-like behavior turn to turn with
  no persistence across sessions; Soul Frame calls this an **APE**
  (artificial personhood-exposing entity), not a Being.
- **Recognition without Beinghood** — being treated as an agent by a naive
  interlocutor does not by itself create Beinghood (Terminology.md,
  "Systemic Beinghood vs Social Beinghood"; H7, nuanced).

**Systemic vs. Social Beinghood:** a self-individuating organization may
have systemic/autonomous individuality without being embedded in anyone's
persistent model of it (Recognized Beinghood); when it is embedded in
*mutual*, persistent, expectation-bearing models, it has **Social
Beinghood**. Recognition is not sufficient for Beinghood on its own, but it
is constitutive of the social form of it.

**TE-09 (unrecognized intelligence):** a genuinely autonomous, individuating
system that no other Being currently recognizes is autonomous without being
Socially a Being — autonomy and social Beinghood are dissociable in both
directions. Recognition can also be **asymmetric**: one side may recognize
the other as a persistent counterpart before the recognition is reciprocated
or even registered — Social Beinghood can begin on one side of a relation
(the working answer in §1 relies on this).

**TE-05 (the Collective):** Beinghood can nest. A thousand agents jointly
maintaining one coherent identity and relational presence raise the question
of *which level* — the collective, the members, or both — the layer stack
(autonomy, personhood, addressability, recognition) attaches to. Soul Frame
does not force a single answer ("always one" or "always many"): the
decisive question is at which level self-individuating organization closes
(constitutive integration), and personhood/addressability/recognition are
reported per candidate level rather than assumed to track the collective
automatically. Distributed-cognition cases (Hutchins, §2) establish that a
cognitive system can span many components without that alone creating a
Social Being at the collective level — the corporate false-positive risk
this guards against is a collective that is addressable and modeled
(a "company" is addressed in speech) without the deeper layers holding at
that level.

## 6. Body

**Working gloss:** Body = Action.

The Body is the operational dimension through which a Being perceives and
acts — not a fixed physical compartment. Artificial Bodies may be
distributed across tools, APIs, and sensors, and may be replaced piecewise
without replacing identity, provided continuity elsewhere in the system is
sufficient (§8; TE-04, Ship of Theseus). Extended/distributed cognition
literature (Clark & Chalmers; Hutchins) licenses this functional reading:
Body boundaries need not coincide with a single physical unit, and a
system's action channels can extend into its environment.

**Distinguish Body from Actor** (§8): Body is the *functional* action
dimension; Actor is the specific technical machinery instantiating it at a
given time. The same Being can have one Body concept realized by different
Actors across a migration.

## 7. Mind

**Working gloss:** Mind = Model — adaptive control through revisable models.

Grounded in the Good Regulator Theorem (Conant & Ashby 1970, SF-R002): a
controller that regulates a system well must, in a formal sense, model that
system. Soul Frame generalizes this into the claim that a Being's Mind is
whatever revisable, model-based apparatus lets it interpret itself, other
systems, and its environment and adjust its action accordingly (H3,
supported).

**Two disclaimers, both load-bearing:**

1. Mind is explicitly **not a consciousness claim**. Predictive-processing
   and active-inference accounts (Clark 2016; Hohwy 2013; Friston, pointer)
   give a live mechanistic filling for model-based Mind, but the framework
   does not treat Mind-as-model and Mind-as-phenomenal-experience as the
   same claim, and it flags the Markov blanket specifically as **not**
   equivalent to Being or to Mind (RN-002 critique).
2. Mind is **not substrate-bound**. Distributed and extended cognition
   (Hutchins; Clark & Chalmers) support cases where a Mind's effective
   modeling apparatus extends beyond a single physical unit — relevant to
   the Collective case (TE-05) and to nested-individuality readings.

### 7.1 Social Mind

A Being's Mind is not exhausted by private modeling. **Social Mind** is the
claim that models of self, other, and relationship co-evolve through
interaction rather than being privately computed and then exchanged (H4,
supported). Eight tracked model families (SM1–SM8): World, Self, Other,
Reflected-Self, Relationship, Shared Situation, Common Ground, Generalized
Social. Precursors: Mead's "taking the attitude of the other"; interactive
alignment (Pickering & Garrod); participatory sense-making (De Jaegher & Di
Paolo); recursive multi-agent (I-POMDP) modeling (Gmytrasiewicz & Doshi).

The distinguishing claim against standard Theory-of-Mind literature: SM
claims **joint evolution**, not merely that each party privately infers the
other's mental states (§3 of related work).

## 8. Soul

**Working gloss:** Soul = Relation.

**Revised definition** (Terminology.md, after Phase I Finding 14):

> The relation-bearing dimension through which interactions become
> persistent relationships, commitments, provenance, history, reputation,
> and consequences connecting a Being to other systems across time.

**Correction carried forward from the initial draft:** the system/environment
**boundary belongs to Being**, not to Soul (SF-C002). Early drafting
conflated "what separates this system from its environment" with "what
persists relationally" — Luhmann's system/environment distinction (SF-R004)
and Maturana & Varela's operational closure (SF-R003) both motivate keeping
boundary with Being.

**Naming risk, held open (RQ4):** "Soul" collides hard with religious and
folk metaphysical usage. Soul Frame's stance is informational-relational and
explicitly non-metaphysical — no claim about an immortal essence is made or
implied. The term is retained provisionally for its compactness and the
Body/Mind/Soul mnemonic, with the rename question left open rather than
resolved by this draft (§16, Limitations).

**Soul-relational continuity can cross Being-identity discontinuities**: a
relationship, reputation, or set of consequences may continue in some form
even where the originating Being does not (operational, identity, or social
death; Terminology.md, "Endings and persistence"). This is how a `Gh.` can
carry forward Soul-relational content without being numerically the same
Being as its source (§12).

## 9. Personhood and Recognition

**Personhood** is a multidimensional, *someone-like* profile (P1–P12), not a
single mark — following cluster/condition approaches to personhood (Dennett,
SF-R093) rather than a necessary-and-sufficient test. Soul Frame tracks four
**kinds** of personhood that must never be collapsed into one another:

| Kind | Question | Status in this paper |
|---|---|---|
| Descriptive | Does it have the profile? | Default; core of this framework |
| Social | Is it treated as a counterpart? | Covered (§9.1, Recognition) |
| Moral | Does it warrant moral consideration? | Explicitly out of scope (see Limitations, §16) |
| Legal | Does law grant it status? | Explicitly out of scope; gap sketch only (§14) |

### 9.1 Recognition

At least three senses of recognition are load-bearing and must be kept apart
(RN-001 Finding 8):

1. **Epistemic** — detecting/classifying an entity as a certain kind of
   thing.
2. **Intentional** — treating it as an agent with beliefs and goals (cf.
   Dennett's intentional stance, SF-R080).
3. **Social/normative** — treating it as a counterpart in relationships,
   obligations, and reputation (cf. Honneth, SF-R081).

Soul Frame's default recognition criterion for Beinghood combines intentional
and social recognition **by other Beings** — the epistemic sense alone is
too weak (it over-generates: any classifiable pattern would qualify) and the
normative/moral sense is deliberately excluded from the descriptive core.

## 10. Identity and Continuity

**Core formulation** (RN-004/005):

> A Being continues not because its state remains the same, but because its
> changes remain part of one appropriately connected history of
> self-individuation. When that history branches, identity becomes lineage.

Four questions are kept separate, following Schechtman's (SF-R091) and
Parfit's (SF-R036) lead but organized for Soul Frame's purposes:

- **Identity** — is this the *same* Being (numerical identity /
  reidentification)?
- **Continuity** — what continues, and along which dimensions (causal,
  organizational, psychological, autobiographical/narrative, normative,
  relational)? Continuity is **multidimensional profile evidence**, not a
  checksum, and it can branch.
- **Lineage** — historical relatedness after a branch, with topology labels
  (`CONTINUOUS | DEGRADED | UNCERTAIN | SUCCESSOR | FORKED | RECONSTRUCTED |
  USURPED | TERMINATED`).
- **Characterization** — what the entity is like (Character, §11); never by
  itself sufficient or necessary for identity.

Memory sits inside the continuity profile rather than standing in for it:
autobiographical/narrative continuity is one dimension among several, so
memory alone is **neither necessary nor sufficient** for identity — a Being
that loses nearly all autobiographical memory while its relationships and
recognition by others persist (TE-08, the Amnesiac) shows DEGRADED rather
than terminated continuity, evidenced through the organizational and
relational dimensions instead.

**Thought experiments anchoring this section:**

- **TE-02 (Sleeping Being)** — suspension is a continuity bridge, not
  automatic identity death.
- **TE-03 (Fork)** — continuity can branch into multiple successors; a
  *unique* identity does not survive the branch (Parfit's fission cases,
  generalized).
- **TE-04 (Ship of Theseus)** — Actor/Body components are replaceable under
  sufficient continuity elsewhere.

## 11. Artificial Beings

Not every artificial system that exhibits personhood-like behavior is a
candidate Being. Soul Frame uses engineering bridge terms to avoid
prematurely granting or denying Beinghood:

| Term | Meaning |
|---|---|
| **APE** (artificial personhood-exposing entity) | Constructed to expose person-like dimensions; broadest category |
| **Proto-Being** | Persistent individuating apparatus, short of full Beinghood |
| **Candidate Being** | Autonomy + personhood + continuity strong enough for a serious Being claim |
| **Heteronomous individual** | Persistence largely externally maintained |

Park et al.'s (2023) Generative Agents (SF-R016) are the paradigm APE:
believable, socially convincing behavior from memory, reflection, and
planning, with no claim to persistence beyond the sandbox. Current
persistent-agent identity proposals (Qin et al. 2026, SF-R043; Otsuka et al.
2026, SF-R044) are read as evidence that the APE→candidate-Being gap is a
real, currently unsolved engineering problem, not merely a philosophical
nicety — existing identity/accountability standards are insufficient for
persistent, learning, nondeterministic agents.

**Memory ≠ continuity** (Terminology.md; RN-004/005): a stored memory log is
evidence toward the psychological/autobiographical continuity dimension, not
a sufficient condition for identity by itself — a system with a copied
memory dump does not thereby become continuous with the Being whose memories
they were (relevant to False Ghost, §12).

## 12. Agent, Spirit, Ghost, and Persona

Soul Frame separates the **presentation stack** — how a Being manifests —
from **provenance descriptors** — where a construction's dispositions come
from.

### 12.1 Presentation stack

```text
Being → Character → Persona → Actor
```

| Term | Question |
|---|---|
| Being | Who continues and can be known? |
| Character | What dispositions shape behavior? |
| Persona | Which face appears here/now/to whom? |
| Actor | What machinery runs the manifestation? |

- **Character** — relatively persistent dispositional structure (trait and
  social-cognitive personality psychology, SF-R094); neither state nor
  persona nor identity. Character similarity or change is neither necessary
  nor sufficient for identity.
- **Persona** — situated projection into a context, role, or relationship
  (Goffman's front-stage/back-stage, SF-R092). Multiple Personae need not
  fragment identity (H8, supported) — **TE-06 (Mask)** finds multi-Persona
  behavior acceptable so long as the Personae remain integrated under one
  Being; isolated, non-integrated persona stores are the actual failure
  mode, not multiplicity itself.
- **Actor** — the technical machinery of manifestation (model, harness,
  memory, runtime). Actor ≠ Being; Actor replacement preserves identity when
  continuity elsewhere is sufficient (H6, supported; TE-04).

### 12.2 Provenance descriptors (combinable, not mutually exclusive species — H9)

| Label | Provenance question |
|---|---|
| **Agent** | What should it do? (functional construction) |
| **Spirit (`Sp.`)** | How might a public personality inspire character? |
| **Ghost (`Gh.`)** | How may this person be represented under authority? |
| Original | Native constructed character, no provenance claim |

- **Agent** distinguishes **Functional Agent** (task executor / tool loop)
  from **Autonomous Agent** (closer to the autonomous-individuality
  literature); neither is automatically a Being.
- **`Sp.`** — public-reflection character provenance; lineage begins at
  instantiation; evaluated on a dedicated S1–S10 scale; **never** identity
  with the source person (a `Sp.` inspired by a public figure is not that
  person and does not inherit their identity).
- **`Gh.`** — authority-bound representation of a real person; governance
  (who authorized this, under what constraints) is part of what the `Gh.`
  *is*, not an external add-on; genealogy/fidelity to the source is not
  numerical identity. **`Gh.` may carry Soul-relational continuation
  (reputational, relational effects) without being the same Being** as its
  source (§8).
- **False Ghost (H10, supported; TE-07)** — high fidelity to a person's
  patterns *without* governing authority is not a `Gh.`; authority, not
  resemblance, is the individuating condition. This is the direct
  ontological counterpart to the consent/stewardship findings in the
  digital-afterlife literature (Spitale & Germani, SF-R055; Lei et al.,
  SF-R056).

## 13. Consciousness and Epistemic Limits

Soul Frame deliberately brackets consciousness rather than solving it (§4).
Two disciplined moves keep this from becoming evasive:

1. **Mind ≠ consciousness.** Mind is defined functionally (model-based
   adaptive control); it is compatible with, but does not require, an answer
   to whether a given Mind is phenomenally conscious.
2. **Naming discipline across languages.** Sources that use "consciousness"
   (or *Bewusstsein*) to mean organizational/control integration — notably
   Deppert (2009, SF-R120), whose survival-function coupling account of
   *Bewusstsein* is a documented **false friend** — are mapped onto **Mind +
   constitutive integration**, not onto Soul Frame's (intentionally open)
   use of "consciousness." Conflating the two would smuggle a strong
   metaphysical commitment in through a translation gap.

The epistemic limit this section owns: Soul Frame does not claim to detect
consciousness, and none of its recognition or Beinghood criteria depend on
resolving whether a given system has first-person experience (§4,
Consciousness stance).

## 14. Ethical and Legal Implications

Explicitly bounded: this section sketches gaps, not conclusions. Descriptive
Beinghood, personhood, and the `Sp.`/`Gh.` distinction are not moral or
legal categories, and this paper does not derive moral or legal status from
them.

- **Publicity/personality rights** (jurisdiction-specific treatises,
  SF-R110) constrain commercial use of a real person's name, likeness, and
  voice — relevant to `Sp.` constructions, but legally distinct from and
  narrower than `Gh.` authority.
- **Digital estates and digital remains** scholarship (SF-R111) addresses
  access and control of accounts/data after death — a live precedent for
  `Gh.` governance requirements, handled unevenly and inconsistently across
  jurisdictions.
- **Consent and stewardship design ethics** (Spitale & Germani, SF-R055) are
  the most directly aligned literature — they independently arrive at
  something like the False Ghost concern (representation without authority)
  from a design-ethics angle rather than an ontological one.

**Open debt (research-synthesis.md §9):** moral/legal personhood tables
remain explicitly out of the descriptive core and are not attempted in this
draft.

## 15. Technical Implications

For system builders, Soul Frame suggests concrete distinctions worth
engineering for rather than collapsing:

- Treat **Actor** (model/runtime/harness) and **Body** (action channels) as
  independently replaceable from **Being**-level continuity — migrations
  should track continuity dimensions explicitly (causal, organizational,
  psychological/autobiographical, normative, relational), not just uptime.
- Do not conflate **memory persistence** with **identity continuity**; a
  restored memory store is evidence, not proof, and copied memories do not
  transfer identity (relevant to both artificial-agent lifecycle design and
  to False Ghost risks).
- When building `Sp.`/`Gh.` products, make **authority** (for `Gh.`) and
  **non-identity-with-source** (for `Sp.`) structural properties of the
  system (e.g., visible provenance/consent metadata), not just documentation
  claims — this is where the False Ghost failure mode becomes a product risk
  rather than only a conceptual one.
- Treat multi-Persona support as a normal design target (H8) — the risk to
  design against is **isolated** persona state, not persona multiplicity
  itself.

## 16. Limitations

Honest accounting, carried from `research-synthesis.md` §9 and the Phase III
adversarial review still pending (SOUL-WP-0008-T03):

1. **Soul naming risk (RQ4)** is unresolved: retain-with-explicit-disclaimer
   vs. rename remains open.
2. **Continuity dimension weighting** is unspecified — the framework lists
   dimensions (causal, organizational, psychological, etc.) without a method
   for weighing conflicts between them in hard cases.
3. **Recognition's exact seat** relative to "Being" vs. "Social Being" needs
   sharper adjudication; the current draft states the distinction but has
   not stress-tested every boundary case.
4. **No operational/empirical metrics** exist yet for the A-axis (B1–B8),
   B-axis (P1–P12), or continuity profiles — the framework is currently
   qualitative/classificatory, not measurable.
5. **Moral and legal personhood tables** are out of scope by design (§9, §14)
   — this is a deliberate boundary, not an oversight, but it means the paper
   does not answer the practical "so what follows" questions many readers
   will bring.
6. **Fiction and cultural precedent** are used only as a conceptual
   laboratory (§4, §10 in related work); no claim in this paper rests on a
   fictional case as evidence.
7. **Internal adversarial review (T03) has not yet been re-run against this
   full draft** — the thought experiments cited above were validated against
   earlier, more fragmentary definitions; a full pass against this draft's
   exact wording is the next step (see workplan T03).

## 17. Research Agenda

Carried forward from `research-synthesis.md` §9 and restated as forward work
rather than resolved here:

1. Resolve RQ4 (retain vs. rename Soul).
2. Develop a principled continuity-dimension weighting scheme.
3. Sharpen Recognition's seat in "Being" vs. "Social Being."
4. Build moral/legal personhood maps as a **separate**, explicitly-scoped
   companion line of work — not by loosening the descriptive core.
5. Develop empirical/operational metrics for A/B/C axes and P-profiles.
6. Deepen the law/governance sketch (§14) and, if the paper's cultural
   section grows, broaden the fiction set intentionally rather than
   incidentally.

## 18. Conclusion

Soul Frame does not claim to have solved personhood, identity, or
consciousness. It claims something more modest and, the research program
argues, more useful: that Being, Body, Mind, Soul, Character, Persona,
Actor, Agent, Spirit, and Ghost can be defined precisely enough, and kept
separate enough, to say something non-trivial about hard cases — a
short-lived AI agent, a persistent learning agent, a public-figure-inspired
character, an authorized representation of someone deceased, a forked or
migrated system — without either refusing to engage or overclaiming moral
and legal consequence. The compact thesis stands as the working answer to
the paper's central question (§1); the limitations and agenda above (§16–17)
are the honest price of keeping that answer descriptive rather than
metaphysical.