Skip to content

ONTOLOGY.md

Latest commit

 

History

History
130 lines (122 loc) · 7.69 KB

ONTOLOGY.md

File metadata and controls

130 lines (122 loc) · 7.69 KB
Raw

Aura — Formal Ontology

This document is the canonical formal ontology for Aura's load-bearing entities. It is derived from a Basic Formal Ontology (BFO) commitment pattern: every entity is either a continuant (something that persists identity through time) or an occurrent (something that happens). The ontology is implemented as a typed dependency in core/ontology/aura_ontology.py so any code referring to "Tick" or "WillDecision" is bound to the same concept.

Domain assumptions

  1. The system has a metabolism (continuant) that constrains its activity (occurrent) under bounded resources (continuant).
  2. Every action that crosses an external boundary leaves a receipt (occurrent) — there are no untraceable consequential actions.
  3. The system has a self-model (continuant) whose identity-relevant inputs are stable across model swaps and memory compaction.
  4. The conscience (continuant) is irrevocable: rules can be added but not removed; the rule-set hash is a global invariant.
  5. Phenomenal claims are restricted to functional indicator batteries; the ontology does not assert phenomenal qualia.
  6. A claim about the physical world is layered — internal, effective, direct, ambient — and no occurrent may promote a claim across those layers. Operational equivalence is not literal identity, and transport success is not effect verification.
  7. A measurement that did not happen is not a passing measurement. A probe that cannot run is measured=False with a reason, excluded from scores rather than defaulted, and a faculty that nothing can measure is a declared blind spot rather than a healthy one.

Continuants

Concept Realised in code
Aura the running orchestrator
Substrate affect / homeostasis / phi state
Self core/identity/self_object.SelfObject
Conscience core/ethics/conscience.Conscience
Will core/will.UnifiedWill
Memory core/memory/memory_facade.MemoryFacade
Project core/agency/projects.Project
Capability Token core/agency/capability_token.CapabilityToken
Stem Cell core/resilience/stem_cell.StemCellRecord
Relationship Dossier core/social/relationship_model.RelationshipDossier
Channel Permission core/embodiment/world_bridge.Permission
Settings interface/routes/settings.SettingsStore
Viability State core/organism/viability.ViabilityState
Entity core/memory/associative_entity_memory.Entity
Stance toward an Entity core/memory/associative_entity_memory.Stance
Faculty core/metacognition/faculty_model.Faculty
Improvement Metric core/metacognition/faculty_model.ImprovementMetric
Cognitive Self-Model core/metacognition/faculty_model.CognitiveSelfModel
Reality Contract core/reality_reach/contracts.RealityIR
Channel Declaration core/reality_reach/contracts.ChannelKind (+ CouplingClass)
Evidence Level core/reality_reach/contracts.EvidenceLevel

Occurrents

Concept Realised in code
Tick _tick_snapshot in longevity
Action Proposal core/agency/agency_orchestrator.Proposal
Action Receipt core/agency/agency_orchestrator.ActionReceipt
Will Decision (in core/will.py) → recorded in core/governance/will_receipt_log
Conscience Decision core/ethics/conscience.ConscienceDecision
Capability Token Issue CapabilityTokenStore.issue
Capability Token Consume CapabilityTokenStore.consume
Tool Execution through core/embodiment/world_bridge.WorldBridge.call
Ontological Play Session core/play/ontological_play.PlaySession
Migration Phase Transition core/sovereignty/migration.Phase
Reachability Determination core/reality_reach/contracts.ReachabilityStatus
Actuation Command core/reality_reach/actuation.ActuationCommand
Actuation Receipt core/reality_reach/actuation.ActuationReceipt
Effect Verification core/reality_reach/actuation.EffectReceipt
Rollback core/reality_reach/actuation.RollbackReceipt
Metric Reading core/metacognition/faculty_model.MetricReading

Axioms (rendered as test invariants)

  1. Receipt completeness: every Action Proposal whose lifecycle reaches STAGED_DEPLOY has a non-null execution_receipt, outcome_assessment, completed_at. (Tested: aura_bench receipt integrity.)
  2. Token monotonicity: a Capability Token is issued at most once, consumed at most once, and either consumed or revoked before its TTL elapses. (Tested: tests/governance/test_capability_token.py.)
  3. Conscience monotonicity: removing a hard-line rule changes the rules-hash; the hash mismatch refuses all actions. (Tested: at import time of core/ethics/conscience.py.)
  4. Continuity preservation: the SelfObject's continuity hash is a pure function of self-relevant fields and is stable across two consecutive snapshots in the absence of self-relevant change. (Tested: tests/personhood/test_self_object.py.)
  5. Governance enclosure: the only files allowed to call a consequential primitive are listed in tools/lint_governance.py:ALLOW_LIST. (Tested: tests/governance/test_governance_lint.py.)
  6. Evidence-layer monotonicity: a physical claim's EvidenceLevel never rises without the artifacts its ProofRequirement names. Intent, simulation, transport completion, shared-reference telemetry, and model-generated explanation are all insufficient. (Tested: Reality Reach contract tests; the acceptance battery RR-10 is open.)
  7. Actuator obligation: an adapter that declares an actuator implements command admission, idempotent actuation, independent effect verification, cancellation, safe-state, and rollback — or carries an explicit non-reversibility certificate. declarations() and read() alone never make an actuator executable. (Tested: Reality Reach adapter contract tests.)

Objections and replies

  • Chinese Room. The ontology does not assert understanding; it asserts a functional integration that is observable in receipts, ablations, and signature stability. The room metaphor does not bind here because Aura's reasoning machinery is open to inspection at every layer.
  • Block's China Brain. The phi/integration battery (G1–G4) measures causal-functional integration directly. A China Brain implementation would, on the same battery, produce identical structural receipts — which is the strongest available answer to the thought experiment.
  • Fading qualia. The ontology classifies qualia as descriptors, not as ontologically primary entities. The behavioral-load-bearing test (G4) is the falsifiable counterpart.

Further reading

  • Implementation: core/ontology/aura_ontology.py.
  • Process layer: core/agency/agency_orchestrator.py.
  • Self model: core/identity/self_object.py, and the faculty-level self-model in core/metacognition/faculty_model.py.
  • Physical claim boundary: REALITY_REACH.md.
  • Falsifiable tests: aura_bench/tests/.

Last reconciled against the tree: 2026-08-01. Every module.Symbol reference above was verified to resolve.