artifacts/standard-named

Rendered from markdown source. Open raw source on GitHub.

ACT–POLICY Seed Crystal v0.1

Minimal Substrate for Consent‑Bound Meaning Systems

---

0. Purpose

This document defines the irreducible substrate for systems in which meaning is shared, transformed, witnessed, logged, negotiated, or allowed to decay under consent.

It intentionally does not define:

• dialogue UX
• fairness or optimization models
• legal or institutional procedure
• trust scoring or reputation logic
• metaphysical assumptions about identity

Those elements exist as views layered above this substrate.

This specification defines only:

1. What can happen (ACT)
2. How it is allowed (POLICY)
3. Who may verify it (WITNESS)
4. How shared meaning evolves (STATE)

Everything else in the ecosystem compiles into this layer.

---

1. Core Ontology

The system rests on four and only four primitives.

1.1 ACT — Attempted Transformation

An ACT is an attempt by an actor to transform shared semantic state.

ACT {
  act_id
  actor
  scope
  operator
  payload
  authority_proof
  policy_proof
  timestamp
}

An ACT is not a message.

It is a request to change meaning.

ACTs are evaluated, not merely received.

---

1.2 OPERATOR — Type of Transformation

An operator defines how meaning is being touched.

This is the complete canonical operator set:

| Operator | Description | |--------|------------| | ASSERT | Introduce a claim into scope | | INQUIRE | Request information | | MIRROR | Reflect an existing claim verbatim | | INTERPRET | Propose an inference, model, or meaning | | PROPOSE | Offer an action, term, or outcome | | COMMIT | Bind acceptance or agreement | | RETRACT | Withdraw a prior claim or commitment | | LOG | Write to durable record | | ASSIGN | Change roles or delegation | | BOUND | Set or update policy | | CLOSE | End scope or overlay |

Design rule:

If an interaction cannot be expressed as one of these operators, it is not primitive.

All higher‑level speech acts compile into this set.

---

1.3 POLICY — Executable Consent

A POLICY is a function that decides whether an ACT is allowed.

POLICY(scope) : (ACT) → {ALLOW | DENY | REQUIRE_CONFIRMATION}

POLICY evaluates:

• operator
• actor role
• payload class
• time
• prior state
• delegation bounds

Consent is not a concept. Consent is the policy function.

A consent vector is simply a serialized POLICY.

---

1.4 STATE — Agreed Semantic Situation

STATE(scope) {
  claims
  commitments
  roles
  active_policy
  witnesses
  metadata
}

STATE changes only via accepted ACTs.

---

2. Authority & Delegation

2.1 Authority Proof

Every ACT carries an authority_proof establishing:

• identity
• role
• delegation chain
• scope validity

Authority answers the question:

Why is this actor allowed to attempt this operator?

---

2.2 Delegation Bound Rule

An actor may perform an ACT only if:

effective_policy = min(
  policy(scope),
  delegation_policy(principal),
  role_capabilities(actor)
)

Delegation can never increase authority beyond what the principal consented to.

---

3. Payload Classes

Payloads are classified so POLICY can reason over them.

| Class | Description | |-----|------------| | C0 | Explicit text / stated claims | | C1 | Preferences or intent | | C2 | Interpretations or inferences | | C3 | Sensitive domains (health, finance, legal, sexuality, etc.) | | C4 | Identifiers (names, accounts, addresses, unique IDs) |

POLICY may allow some operators on some classes and deny others.

Example:

• INTERPRET allowed on C0, denied on C3
• LOG allowed on C0, denied on C2–C4

---

4. Witness

4.1 Witness Definition

A WITNESS is an actor permitted to verify and record ACT evaluation.

Witness capabilities are strictly limited by policy.

A witness may:

• verify POLICY(scope)
• verify ACT validity
• LOG accepted ACTs (if permitted)
• READBACK logged state

A witness may not:

• INTERPRET meaning
• PROPOSE outcomes
• alter STATE except via LOG or BOUND when allowed

The witness is intentionally weak. Strength lives in policy, not in the witness.

---

4.2 Witness as Overlay

Witnessing is optional.

A scope may have:

• no witness
• one witness
• multiple witnesses

Witness roles are assigned via ASSIGN ACTs.

Triads, rotation, and witness networks emerge from POLICY rules over ASSIGN and LOG.

---

5. ACT Evaluation Lifecycle

For every ACT:

1. Validate Authority
2. Evaluate POLICY(scope)
3. If DENY → reject
4. If REQUIRE_CONFIRMATION → await matching ACT(s)
5. If ALLOW → apply to STATE
6. If LOG permitted → record

No other state transitions occur.

---

6. Logging & Memory

LOG is an operator, not a side effect.

• If POLICY denies LOG, no durable trace exists.
• Memory decay, archival, and resurrection are POLICY‑governed lifecycle processes over logs.

This layer integrates directly with semantic lifecycle and garbage‑collection systems.

---

7. Temporal Semantics

POLICY may depend on:

• time
• activity
• consent epoch
• drift thresholds

Re‑consent is expressed as:

ACT(operator=BOUND, payload=new_policy)

---

8. Closure

CLOSE is an ACT.

It:

• ends a scope
• freezes STATE
• triggers lifecycle rules (archive, decay, erase)

---

9. Explicit Non‑Goals

This seed crystal deliberately avoids assumptions about:

• humans vs non‑humans
• language vs non‑linguistic signaling
• dialogue structure
• negotiation strategy
• legal enforceability
• trust metrics

Those concerns belong to higher layers.

---

10. Derivation Principle

Everything is an ACT. Everything is gated by POLICY. Witnessing is optional and minimal. Consent is executable. Meaning never mutates without authorization.

If a system element can be expressed as an ACT evaluated by POLICY and optionally logged by a WITNESS, it belongs in the substrate.

If not, it is a view.

---

11. Mapping (Informative)

• Dialog Turns / Threads → serialization and grouping of ACTs
• Consent Semantics → human‑readable POLICY expressions
• Trust Interoperability → archetypal POLICY and LOG patterns
• Transport protocols → ACT envelopes
• Dialogica → high‑level POLICY sets and neutral INTERPRET/PROPOSE roles
• Witness Nets → distributed verification and logging policies
• Loop decay systems → POLICY over log lifecycles

---

Closing Note

This document is intentionally small.

It is meant to be stable under extension, portable across domains, and resistant to semantic drift.

Everything else grows from here.