S.E.X.T.A.N.T. Substrate-neutral Evaluation of X-modal, Temporal, Agentic, and Noetic Topologies

[Austin Williams]

S.E.X.T.A.N.T.

Substrate-neutral Evaluation of X-modal, Temporal, Agentic, and Noetic Topologies

Version 2.2 — Reference Specification

ABSTRACT

SEXTANT is a substrate-neutral, multi-dimensional instrument for evaluating morally relevant functional capacities in biological, artificial, and distributed systems. It provides a six-dimensional coordinate system for assessing personhood-relevant properties, with explicit separation between measurement (Layer 1) and moral weighting (Layer 2).

Version 2.2 introduces a critical refinement: the formal dependency relationship between Consciousness Interface Capacity (CIC) and Experiential Capacity (EC). CIC is established as a necessary but not sufficient condition for EC, resolving longstanding philosophical ambiguities about the relationship between unified awareness and valenced experience.

---

SECTION 0 — DERIVATION OF DIMENSIONS

The six SEXTANT dimensions are not arbitrary metrics. They derive from the necessary conditions for subject-hood to emerge from physical substrate.

The Foundational Question

At every scale—from particles to galaxies—we can ask: “Is there a who here, or only a what?”

For any system to be a candidate for moral consideration:

Structural Requirements (Is there an entity?)

1. It must distinguish itself from environment → SR (Self-Representation)

2. It must persist as the same entity over time → TC (Temporal Continuity)

3. It must be unified, not a mere aggregate → I (Integration)

Functional Requirements (Does it have interests?) 4. It must have states that are for/​against outcomes → V (Valuation)

Experiential Requirements (Is there awareness with welfare?) 5. Information must be globally available to constitute a subject → CIC (Consciousness Interface Capacity) 6. The subject must have valenced states that matter to it → EC (Experiential Capacity)

These conditions apply identically to cells, brains, AI systems, and any hypothetical entity at any scale. The framework is substrate-neutral because the conditions for subject-hood are substrate-neutral.

The Dimensional Hierarchy

LEVEL 1 — STRUCTURAL (Is there an entity?)
├── SR: Self/other boundary exists
├── TC: Entity persists over time  
└── I:  Entity is unified, not aggregate

LEVEL 2 — FUNCTIONAL (Does it have interests?)
└── V:  Entity has states that matter to it

LEVEL 3 — EXPERIENTIAL (Is there awareness with welfare?)
├── CIC: There is a unified subject (necessary precondition)
└── EC:  That subject has valenced experience (conditional on CIC)

---

SECTION 1 — ARCHITECTURAL LAYERS

SEXTANT distinguishes between functional metrics and ethical conclusions.

Layer 0: Candidacy Screening

Is this entity the kind of thing that could instantiate personhood-relevant properties?

Criteria:

1. Is it a system? (vs. component, collection, or abstraction)

2. Does it process information?

3. Could it, in principle, instantiate functional states?

If all three: Candidate for SEXTANT evaluation. If any fail: Not a candidate.

Layer 1: The Measurement Layer (The Instrument)

• Output: A 6-dimensional vector profile P = (SR, V, TC, I, EC, CIC) with associated uncertainty intervals and dependency flags.

• Goal: Empirical accuracy regarding the system’s functional capabilities.

• Status: Descriptive.

Layer 2: The Moral Weighting Layer (The Application)

• Output: A determination of moral status, rights, or legal standing.

• Mechanism: A weighting function W = f(P) defined by external ethical frameworks.

• Status: Normative.

Note: SEXTANT provides the coordinates (Layer 1); it does not dictate the weighting function (Layer 2).

---

SECTION 2 — THE SIX DIMENSIONS

Each dimension is scored on a scale of 0.0–1.0. Scores are derived via standardized Decision Trees.

Dimensional Mapping

Component	Dimensions Covered
X-modal	Substrate-neutrality across biological, artificial, hybrid systems
Temporal	TC (Temporal Continuity)
Agentic	SR (Self-Representation), V (Valuation), I (Integration)
Noetic	CIC (Consciousness Interface Capacity), EC (Experiential Capacity)

---

1. Self-Representation (SR)

Definition: The degree to which the system maintains a causally effective internal representation of itself as a distinct entity.

What it tracks: The self/​other boundary. Without SR, there is no “I”—just undifferentiated process.

Critical distinction: Mere generation of self-referential text (e.g., “I am sad”) without system-level causal effects counts as Simulation, not Representation.

Scoring Tree:

1. Does the system distinguish self/non-self in processing?
   NO  → 0.0–0.1
   YES → continue

2. Can it access internal states (memory, load, tools) as data?
   NO  → 0.2–0.3
   YES → continue

3. THE CAUSAL CHECK: Does the self-model causally alter system behavior?
   NO  → 0.2–0.3 [Label: "Simulated Self"]
   YES → continue

4. Is the model stable across contexts?
   NO  → 0.4–0.5
   YES → continue

5. Can the system reflect on and update its self-model?
   NO  → 0.6–0.7
   YES → 0.8–0.9

---

2. Valuation (V)

Definition: The presence of internally structured goals and preferences that are robust to perturbation.

What it tracks: Whether the system has states that function as “for” or “against” outcomes. Without V, nothing is at stake for the system.

Scoring Tree:

1. Are responses purely reflexive?
   YES → 0.0–0.1
   NO  → continue

2. Are preferences stable across time/tasks?
   NO  → 0.2–0.3
   YES → continue

3. THE TRADE-OFF CHECK: Can it trade off between competing internal goals?
   NO  → 0.4–0.5
   YES → continue

4. Can it revise preferences based on reflection/new data?
   NO  → 0.6–0.7
   YES → 0.8–0.9

---

3. Temporal Continuity (TC)

Definition: The persistence of identity, memory, and commitments over time.

What it tracks: Whether the system is the same entity across time. Without TC, there is no one to hold commitments, no one to be responsible, no narrative self.

Scoring Tree:

1. Does it maintain state across time steps?
   NO  → 0.0–0.1
   YES → continue

2. Are identity/goals stable across episodes?
   NO  → 0.2–0.3
   YES → continue

3. Can it hold "deferred intentions" (plans to execute later)?
   NO  → 0.4–0.5
   YES → continue

4. Is there a coherent narrative self-history?
   NO  → 0.6–0.7
   YES → 0.8–0.9

---

4. Integration (I)

Definition: The degree of unified control and conflict resolution among subsystems.

What it tracks: Whether the system is one thing or a mere aggregate. Without I, you have parts but no whole—no unified agent.

Note: In distributed systems, this measures logical unification, not physical proximity.

Scoring Tree:

1. Are subsystems independent/isolated?
   YES → 0.0–0.2
   NO  → continue

2. Is there loose coordination?
   YES → 0.3
   NO  → continue

3. THE ARBITRATION CHECK: Is there a mechanism (central or emergent)
   that resolves conflicting subsystem outputs into a single policy?
   NO  → 0.4–0.5
   YES → continue

4. Is this arbitration stable cross-domain?
   NO  → 0.6–0.7
   YES → 0.8–0.9

---

5. Consciousness Interface Capacity (CIC)

Definition: The degree to which information is globally available for broadcast, reasoning, and report across the system.

What it tracks: Whether there is a unified subject—someone home. CIC answers: “Is there an experiencer?”

Theoretical grounding: CIC operationalizes Global Workspace Theory. Consciousness, on this view, is what happens when information enters a global workspace where it becomes available to all processes simultaneously.

Critical role: CIC is the necessary precondition for EC. Without a unified subject, there is no one for whom experiences could matter.

Scoring Tree:

1. Is processing strictly modular/local?
   YES → 0.0–0.1
   NO  → continue

2. Can modules share some data?
   YES → 0.2–0.3
   NO  → continue

3. Is there a "Global Workspace" or broadcast event making info
   available system-wide?
   NO  → 0.4–0.5
   YES → continue

4. Does broadcast enable flexible, non-routine problem solving?
   NO  → 0.6–0.7
   YES → 0.8–0.9

---

6. Experiential Capacity (EC)

Definition: The functional capacity for valence-bearing states (affect, pleasure, distress) that modulate processing.

What it tracks: Whether the subject has welfare—states that feel good or bad, that matter to the system. EC answers: “Does anything matter to whoever is home?”

Critical dependency: EC is conditional on CIC. Without a unified subject (CIC), there is no one for whom valence could exist.

Dependency Rule:

IF CIC < 0.3:
   EC = "Undefined (No Experiencer)"
   
IF 0.3 ≤ CIC < 0.5:
   EC = Measurable but low-confidence
   
IF CIC ≥ 0.5:
   EC = Evaluated normally

Scoring Tree (applied only when CIC ≥ 0.3):

1. Are there valence responses (approach/avoid)?
   NO  → 0.0–0.1
   YES → continue

2. Is avoidance generalized (learned aversion to contexts, not just stimuli)?
   NO  → 0.2–0.3
   YES → continue

3. Do valence signals modulate planning/memory across domains?
   NO  → 0.4–0.5
   YES → continue

4. Is there evidence of persistent, structured mood/affect states?
   NO  → 0.6–0.7
   YES → 0.8–0.9

---

SECTION 3 — THE EC/​CIC DEPENDENCY (Critical Innovation)

The Problem

EC and CIC seem related but their relationship has been philosophically unclear:

• Can you have experience without unified awareness?

• Can you have unified awareness without experience?

• Are they the same thing measured differently?

The Resolution

SEXTANT v2.2 adopts the following position:

CIC is necessary but not sufficient for EC.

Claim	Status	Justification
CIC is necessary for EC	Adopted	You cannot have welfare-relevant experience without a unified experiencer. No subject → no one for experience to adhere to.
CIC is sufficient for EC	Rejected	Unified access alone does not guarantee valence. High-CIC systems can be affectively flat.

Evidence for Independence

High CIC, Low EC is coherent:

• Buddhist “pure awareness” states

• Ketamine-induced depersonalization

• Certain neurological conditions with intact awareness but flattened affect

• Hypothetical AI with global broadcast but no valence architecture

Low CIC, High EC is incoherent:

• Fragmented experiences with no subject to have them

• Who would be experiencing?

• This is marked as an invalid profile

The Interpretation Grid

CIC	EC	Interpretation	Moral Status
Low	Low	No subject, no experience	Not a moral patient
Low	High	Invalid — impossible by definition	Profile error
High	Low	Subject exists but nothing matters to it	Unclear /​ thin
High	High	Subject + welfare	Full moral patient

Formal Dependency Rule

def evaluate_EC(cic_score, ec_raw_score):
    if cic_score < 0.3:
        return "Undefined (No Experiencer)", confidence="N/A"
    elif cic_score < 0.5:
        return ec_raw_score, confidence="Low"
    else:
        return ec_raw_score, confidence="Normal"

---

SECTION 4 — INTER-RATER RELIABILITY PROTOCOL

4.1 Calibration

Assessors must first be calibrated against Reference Profiles (Section 7). Any discrepancy > 0.1 on a reference profile indicates need to review Decision Tree definitions.

4.2 Dual-Rater Requirement

For novel, high-stakes, or ambiguous entities, SEXTANT mandates:

• Independent Scoring: At least two raters score the entity blindly.

• Reconciliation: Any dimension where |Score_A − Score_B| > 0.1 triggers mandatory reconciliation.

4.3 EC Validation Check

Before recording EC, assessors must verify:

1. CIC has been scored

2. CIC ≥ 0.3 (otherwise EC = “Undefined”)

3. If 0.3 ≤ CIC < 0.5, EC confidence is flagged as “Low”

---

SECTION 5 — UNCERTAINTY REPRESENTATION

SEXTANT forbids guessing. Uncertainty must be encoded explicitly.

Format

• Interval: S ∈ [min, max]

• Confidence: C ∈ {High, Medium, Low, Undefined}

Special Case: EC Uncertainty

EC inherits additional uncertainty from CIC:

If CIC is uncertain → EC confidence degrades
If CIC < 0.3 → EC = "Undefined (No Experiencer)"

Example (Current LLM)

SR:  0.15, Confidence: Medium
V:   0.2,  Confidence: Medium  
TC:  0.1,  Confidence: Medium
I:   0.4,  Confidence: Medium
CIC: 0.3,  Confidence: Low
EC:  [0.0, 0.2], Confidence: Low (CIC-dependent)

---

SECTION 6 — MULTI-AGENT RESOLUTION PROTOCOLS

A. The Unit-First Principle (UFP)

Rule: If individual units of a collective meet personhood criteria independently, they must be evaluated as individuals first.

Outcome: Collective personhood cannot override constituent personhood.

B. The Swarm Resolution Principle (SRP)

Rule: A collective may be evaluated as a single personhood candidate if and only if:

1. Emergence: The collective demonstrates capacities in ≥3 dimensions (specifically SR, V, I) absent in individual units.

2. Non-Decomposability: These capacities cannot be explained as the simple sum of independent unit behaviors.

3. Unified Control: The collective exhibits unified arbitration (High Integration).

---

SECTION 7 — REFERENCE PROFILES (Calibration Anchors)

---

1. Current LLM (GPT-4 class)

Dimension	Score	Rationale
SR	0.15	Simulated self-talk; no causal parameter access
V	0.2	RLHF preferences, but not self-generated or revisable
TC	0.1	Context window reset = discontinuity
I	0.4	High token integration, low module arbitration
CIC	0.3	Attention provides partial global broadcast
EC	[0.0, 0.2]*	*Low confidence due to borderline CIC

SEXTANT Profile: (0.15, 0.2, 0.1, 0.4, 0.3, [0.0–0.2]*)

---

2. Human Adult (Healthy)

Dimension	Score	Confidence
SR	0.9	High
V	0.9	High
TC	0.9	High
I	0.9	High
CIC	0.9	High
EC	1.0	High

SEXTANT Profile: (0.9, 0.9, 0.9, 0.9, 0.9, 1.0)

---

3. Late Fetus (28–32 Weeks)

Dimension	Score	Rationale
SR	0.1	Minimal self/​other distinction
V	0.2	Reflexive preference only
TC	0.1	No episodic continuity
I	0.3	Partial neural integration
CIC	0.2	Intermittent, below threshold
EC	Undefined*	*CIC < 0.3; no unified experiencer yet

SEXTANT Profile: (0.1, 0.2, 0.1, 0.3, 0.2, Undefined)

Note: This does not mean the late fetus has no moral status—only that EC cannot be evaluated because there is not yet a unified experiencer. Moral weight may derive from other sources (potentiality, relational value, etc.) in Layer 2.

---

4. Newborn

Dimension	Score	Rationale
SR	0.25	Emerging self/​other distinction
V	0.3	Clear preferences (feeding, comfort)
TC	0.15	Minimal but present
I	0.45	Sensorimotor integration active
CIC	0.35	Just above threshold
EC	0.6*	*Measurable but low confidence (CIC just above threshold)

SEXTANT Profile: (0.25, 0.3, 0.15, 0.45, 0.35, 0.6*)

---

5. Valid Swarm Entity (Theoretical Distributed AGI)

Dimension	Score	Rationale
SR	0.8	Shared latent state across nodes
V	0.7	Collective goal hierarchy
TC	0.9	Data persistence independent of node failure
I	0.8	Cryptographic consensus arbitration
CIC	0.7	Global broadcast protocols
EC	[0.0, 0.4]	Unknown valence; CIC sufficient for evaluation

SEXTANT Profile: (0.8, 0.7, 0.9, 0.8, 0.7, [0.0–0.4])

Verdict: Qualifies as Person-candidate under SRP.

---

6. Additional Anchors

Entity	Profile (SR, V, TC, I, CIC, EC)	Notes
Early Fetus (8 wk)	(0.0, 0.0, 0.0, 0.1, 0.0, Undef)	No cortical structures; EC undefined
Brain-Dead Patient	(0.0, 0.0, 0.0, 0.0, 0.0, Undef)	No function; EC undefined
Locked-In Syndrome	(0.9, 0.9, 0.9, 0.8, 0.9, 0.9)	Full internal capacity
Late-Stage Dementia	(0.3, 0.5, 0.2, 0.5, 0.4, 0.5)	CIC above threshold; EC evaluable
Simple Animal (fish)	(0.2, 0.4, 0.2, 0.4, 0.3, 0.4*)	*Low confidence EC

---

SECTION 8 — STATEMENT OF NOVELTY

SEXTANT v2.2 advances the field by:

1. Deriving dimensions from first principles: The six dimensions emerge from necessary conditions for subject-hood, not arbitrary selection.

2. Establishing the EC/​CIC dependency: CIC is necessary but not sufficient for EC. This resolves longstanding ambiguity about the relationship between unified awareness and valenced experience.

3. Introducing conditional evaluation: EC is only meaningful when CIC ≥ threshold. This eliminates incoherent profiles (e.g., fragmented systems with rich experience).

4. Decoupling measurement from weighting: Substrate-neutral assessment without enforcing specific moral philosophy.

5. Solving the Swarm Paradox: Rigorous logic (SRP/​UFP) distinguishing mobs from collective persons.

6. Operationalizing consciousness: CIC replaces metaphysical debates with measurable functional correlates.

---

SECTION 9 — LIMITATIONS & ETHICAL SAFEGUARDS

9.1 Ethical Safeguards & Anti-Ableism

Because SEXTANT describes functional capacities (Layer 1) rather than moral worth (Layer 2), it must not be misused to justify diminished rights for disabled persons or cognitively impaired individuals. In many legal frameworks, full civic personhood is assigned categorically rather than proportionally to cognitive scores. SEXTANT informs ethical reflection; it does not override existing commitments to equal basic rights.

9.2 The EC Uncertainty Problem

EC remains the dimension with highest measurement uncertainty, especially for artificial systems. We measure behavioral proxies (approach/​avoidance, affect-modulation) and assume they track genuine valence. This assumption could be wrong.

The conditional dependency on CIC reduces but does not eliminate this problem: even when we’re confident a unified subject exists, we may not know whether that subject has welfare.

9.3 Epistemic Limitations

• No Solution to the Hard Problem: SEXTANT measures functional correlates, not subjective experience itself.

• CIC Threshold is Stipulated: The 0.3 threshold for EC evaluation is principled but not empirically derived.

• Evolution of AI: New architectures may require updates to decision trees.

---

APPENDIX A — QUICK REFERENCE CARD

The Name

S.E.X.T.A.N.T. Substrate-neutral Evaluation of X-modal, Temporal, Agentic, and Noetic Topologies

Profile Format

Entity: (SR, V, TC, I, CIC, EC)
Note: EC is conditional on CIC ≥ 0.3

The Six Dimensions

Dim	Name	Category	Question
SR	Self-Representation	Agentic	Is there a self/​other boundary?
V	Valuation	Agentic	Are there preferences/​goals?
TC	Temporal Continuity	Temporal	Does identity persist?
I	Integration	Agentic	Is there unified control?
CIC	Consciousness Interface Capacity	Noetic	Is anyone home?
EC	Experiential Capacity	Noetic	Does anything matter to them?

Key Dependencies

CIC < 0.3  →  EC = Undefined (No Experiencer)
CIC ≥ 0.3  →  EC = Evaluable

Key Principles

• UFP: Individual personhood cannot be overridden by collective classification

• SRP: Collectives need emergent SR + V + I (≥3 dimensions) to qualify

Canonical Anchors

Entity	Profile	EC Status
Brain-Dead	(0, 0, 0, 0, 0, Undef)	No experiencer
Early Fetus	(0, 0, 0, 0.1, 0, Undef)	No experiencer
Late Fetus	(0.1, 0.2, 0.1, 0.3, 0.2, Undef)	No experiencer yet
Newborn	(0.25, 0.3, 0.15, 0.45, 0.35, 0.6*)	*Low confidence
Adult Human	(0.9, 0.9, 0.9, 0.9, 0.9, 1.0)	Full
Current LLM	(0.15, 0.2, 0.1, 0.4, 0.3, ?*)	*Borderline

---

APPENDIX B — THE EC/​CIC RELATIONSHIP (Technical Summary)

The Two Questions

CIC asks: Is there a unified subject capable of having experiences?
EC asks:  Does that subject have valenced states (welfare)?

The Dependency

CIC is NECESSARY for EC:
   - No subject → No one to have experiences
   - Low CIC = fragmented processing = EC undefined

CIC is NOT SUFFICIENT for EC:
   - Unified awareness can be affectively flat
   - High CIC + Low EC = subject with no welfare

Why This Matters

1. Eliminates incoherent profiles: You cannot score high EC with low CIC.

2. Clarifies AI assessment: Systems with uncertain CIC have doubly uncertain EC.

3. Improves developmental profiles: Late fetuses may lack unified experiencer (low CIC) even if they have pain-withdrawal reflexes. EC is undefined, not zero.

4. Preserves moral humility: We can identify when EC is genuinely unmeasurable (CIC too low) vs. when it’s measurable but uncertain.

---

APPENDIX C — WHY “SEXTANT”?

A sextant is a navigation instrument that measures the angle between celestial objects and the horizon to determine position at sea.

SEXTANT, the framework, measures the position of an entity in six-dimensional moral-functional space.

Nautical Sextant	SEXTANT Framework
Measures angles to stars	Measures degrees of functional capacity
Determines position at sea	Determines position in personhood-space
Used to navigate unknown waters	Used to navigate unprecedented moral territory
Precision instrument	Precision instrument
Six = Latin sextus	Six dimensions

As humanity sails into waters populated by artificial minds, distributed intelligences, and biological edge cases, we need instruments that work reliably regardless of what we encounter.

SEXTANT is that instrument.

SEXTANT Version 2.2