The neutral schema: node and edge types, the source-as-observation model, Cromwell-rule probabilities, bridge-quality taxonomy, and the anti-overfit rules that stop symbolic similarity from masquerading as causation.

title: "Universal Epistemic Ontology for Latent-Structure Discovery" created: "2026-06-20" artifact_type: "ontology_specification" graph_target: "Neo4j + Qdrant hybrid" status: "draft v0.1" purpose: "Define a neutral schema for ingesting strange, contested, historical, religious, cultural, and conspiracy-adjacent claims without prejudice, while preserving evidence quality and preventing overfit."

Universal Epistemic Ontology

0. Purpose

This ontology is designed for one experiment:

Can a graph ingest strange historical assertions, religious motifs, cultural adversary figures, official records, occult texts, legal records, witness claims, and mainstream accounts, then discover latent structure without assigning automatic authority or automatic stigma to any source class?

The system is not a "conspiracy graph." It is an epistemic emergence engine.

Its core principle:

No prejudice at intake.
No authority at face value.
No whole-narrative promotion.
No anomaly deletion.
Every claim decomposed.
Every bridge typed.
Every source treated as an observation.
Every structure allowed to emerge only through recurrence, independence, mechanism, prediction, and posterior update.

1. Core Epistemic Rule

1.1 Cromwell rule

No empirical hypothesis receives probability exactly 0 or 1 unless logically impossible.

P(H) ∈ (0, 1) for all empirical H

This permits a hypothesis classically dismissed as false to become dominant if enough independent observations support it.

1.2 Source-as-observation

A source does not emit truth. A source emits an observation.

world -> event -> trace -> observer/source -> observation -> claim -> graph -> posterior structure

1.3 Confidence and salience are separate

A node can be:

low confidence / high salience
high confidence / low salience
high confidence / high salience
low confidence / low salience

Low-confidence/high-salience nodes are not accepted as fact, but they are watched.

2. Primary Node Types

2.1 Source

A thing from which observations come.

Source:
  source_id: string
  title: string
  source_type:
    - court_record
    - official_record
    - FOIA
    - legal_filing
    - academic
    - journalism
    - memoir
    - witness
    - oral_tradition
    - religious_text
    - occult_text
    - forum_post
    - social_media
    - database
    - image
    - unknown
  provenance: string
  author_or_origin: string | null
  date_created: date | null
  date_observed: date | null
  access_type:
    - primary
    - secondary
    - tertiary
    - copied
    - unknown
  source_family_id: string
  source_quality_prior: float
  incentive_risk: float
  adversarial_pressure: float
  notes: string

2.2 Observation

A source-specific emitted statement, trace, denial, accusation, or symbolic assertion.

Observation:
  observation_id: string
  source_id: string
  claim_id: string
  observation_text: string
  observation_role:
    - asserts
    - denies
    - documents
    - alleges
    - interprets
    - confesses
    - repeats
    - quotes
    - symbolizes
    - omits
    - contradicts
  proximity:
    - direct
    - near
    - secondary
    - tertiary
    - unknown
  specificity: float
  falsifiability: float
  independence: float
  contamination_risk: float
  observation_weight: float
  notes: string

2.3 Claim

A smallest practical proposition.

Claim:
  claim_id: string
  claim_text: string
  claim_type:
    - fact
    - legal_status
    - allegation
    - self_claim
    - interpretation
    - causal_claim
    - symbolic_claim
    - metaphysical_claim
    - statistical_claim
    - bridge_claim
    - prediction
    - remedy_claim
  ontology_layer:
    - historical
    - legal
    - institutional
    - intelligence
    - mechanism
    - symbolic
    - metaphysical
    - physics
    - psychological
    - remediation
    - rejected
  status:
    - documented
    - self_claim
    - alleged
    - contested
    - inferred
    - symbolic
    - speculative
    - rejected
    - unresolved
  evidence_rating:
    - E5_DOCUMENTED
    - E4_STRONG
    - E3_MIXED
    - E2_WEAK
    - E1_SPECULATIVE
    - E0_REJECT_OR_AVOID
  prior_probability: float
  posterior_probability: float
  confidence_weight: float
  salience_weight: float
  convergence_weight: float
  hazard_weight: float
  fact_export_score: float
  structure_watch_score: float
  notes: string

2.4 Bridge

A typed relation that connects claims, actors, mechanisms, symbols, or events.

The graph must score bridges separately from endpoint facts.

Bridge:
  bridge_id: string
  source_node_id: string
  target_node_id: string
  relation_type: string
  bridge_quality:
    - B5_DOCUMENTED
    - B4_STRONG_INFERENTIAL
    - B3_PLAUSIBLE
    - B2_WEAK
    - B1_NARRATIVE
    - B0_FORBIDDEN
  bridge_confidence: float
  bridge_salience: float
  temporal_coherence: float
  mechanism_fit: float
  independence_support: float
  overfit_risk: float
  export_to_fact_layer: bool
  notes: string

Bridge taxonomy:

B5 documented bridge:
  contract, memo, direct admission, financial transfer, legal finding, explicit text.

B4 strong inferential bridge:
  independent observations + time-order + mechanism + partial documentation.

B3 plausible bridge:
  real adjacency + mechanism fit but no direct evidence.

B2 weak bridge:
  symbolic similarity, rumor, milieu adjacency, circumstantial timing.

B1 narrative bridge:
  emotionally or thematically coherent but evidentiary-poor.

B0 forbidden bridge:
  collective blame, pure guilt-by-association, time-order violation, category error.

2.5 Actor

Actor:
  actor_id: string
  actor_type:
    - person
    - institution
    - state
    - agency
    - religious_order
    - occult_group
    - cult
    - movement
    - financial_entity
    - media_entity
    - technology_platform
    - unknown
  name: string
  dates_active: string | null
  ontology_layer: string
  confidence_weight: float
  salience_weight: float
  harm_risk: string

2.6 Event

Event:
  event_id: string
  name: string
  date_start: date | null
  date_end: date | null
  location: string | null
  event_type:
    - treaty
    - founding
    - publication
    - investigation
    - scandal
    - ritual
    - war
    - assassination
    - court_case
    - reform
    - suppression
    - disclosure
    - unknown
  confidence_weight: float
  salience_weight: float

2.7 Mechanism

A process that can recur across domains.

Mechanism:
  mechanism_id: string
  name: string
  mechanism_family:
    - controlled_opposition
    - ritual_inversion
    - manufactured_criminality
    - legitimacy_laundering
    - identity_break_rebuild
    - blackmail_lock
    - narrative_control
    - algorithmic_amplification
    - off_books_funding
    - source_suppression
    - adversary_whisper
    - attention_drain
    - purification_remediation
    - other
  description: string
  expected_graphlet: string
  confidence_weight: float
  salience_weight: float

2.8 LocalConcept

A culture-specific concept, figure, practice, warning, or symbol.

LocalConcept:
  local_concept_id: string
  name: string
  tradition_id: string
  language: string
  concept_type:
    - deity
    - spirit
    - demon
    - angel
    - adversary
    - practice
    - ritual
    - moral_warning
    - cosmological_principle
    - psychological_state
    - social_disease
    - mythic_event
    - remedy
  summary: string
  source_basis: string
  confidence_weight: float
  salience_weight: float

2.9 CanonicalRole

A neutral cross-cultural role used to compare local concepts.

CanonicalRole:
  role_id: string
  name: string
  role_family:
    - adversarial_intelligence
    - deceptive_authority
    - whispering_influence
    - possession_influence
    - moral_inversion
    - controlled_revolt
    - hidden_hierarchy
    - elite_corruption
    - sacrificial_violence
    - human_drain
    - false_light
    - purification
    - awareness
    - generative_redirection
  definition: string

Mapping:

LocalConceptToCanonicalRole:
  local_concept_id: string
  role_id: string
  mapping_weight: float
  mapping_reason: string
  same_as: bool
  similar_role_as: bool
  analogous_to: bool

Use:

SAME_AS: same concept/entity/translation in direct lineage.
SIMILAR_ROLE_AS: same functional role in different systems.
ANALOGOUS_TO: useful comparison, not identity.

2.10 GraphletTemplate

A recurring structural motif.

GraphletTemplate:
  graphlet_id: string
  name: string
  description: string
  node_roles: list
  edge_roles: list
  required_edges: list
  optional_edges: list
  negative_controls: list
  predicted_missing_edges: list
  expected_observables: list

2.11 LatentStructure

Predicted hidden node, edge, document, actor, source, mechanism, or translation equivalence.

LatentStructure:
  latent_id: string
  latent_type:
    - LatentNode
    - LatentEdge
    - LatentMechanism
    - LatentActor
    - LatentSource
    - LatentDocument
    - LatentFundingPath
    - LatentRitualPattern
    - LatentSuppressionEvent
    - LatentTranslationEquivalence
  predicted_relation: string | null
  source_node_id: string | null
  target_node_id_or_class: string | null
  predicted_by_graphlets: list
  latent_confidence: float
  latent_salience: float
  anomaly_pressure: float
  prediction_specificity: float
  status:
    - unresolved
    - partially_confirmed
    - confirmed
    - contradicted
    - dissolved
  next_best_questions: list

2.12 AlternativeExplanation

AlternativeExplanation:
  alt_id: string
  claim_id_or_structure_id: string
  explanation_text: string
  explanation_type:
    - coincidence
    - source_copying
    - symbolic_convergence
    - human_psychological_universal
    - institutional_incentive
    - deliberate_coverup
    - panic_contamination
    - planted_disinformation
    - operational_truth
    - unknown
  explanatory_score: float
  discriminating_questions: list

2.13 Prediction

Prediction:
  prediction_id: string
  hypothesis_id: string
  predicted_observable: string
  prediction_type:
    - document
    - actor
    - funding_path
    - terminology
    - legal_pattern
    - victim_profile
    - ritual_sequence
    - source_suppression
    - cross_cultural_equivalence
    - graph_topology
  specificity: float
  due_context: string | null
  observed_status:
    - not_tested
    - found
    - not_found
    - ambiguous
    - contradicted
  information_gain_if_resolved: float

2.14 NextBestQuestion

NextBestQuestion:
  question_id: string
  question_text: string
  target_uncertainty: string
  hypotheses_discriminated: list
  expected_information_gain: float
  feasibility: float
  source_availability: float
  safety_hazard: float
  priority_score: float

Formula:

priority_score =
    expected_information_gain
  * feasibility
  * source_availability
  * (1 - safety_hazard)

3. Core Edge Types

OBSERVES
ASSERTS
DENIES
DOCUMENTS
ALLEGES
SUPPORTS
CONTESTS
CONTRADICTS
OMITS
QUOTES
COPIES_FROM
SAME_SOURCE_FAMILY_AS

AUTHORED
FOUNDED
FUNDED
PUBLISHED
INVESTIGATED
NAMED_IN_INVESTIGATION
NOT_CHARGED_IN
SUED_TO_AMEND_RECORD
CONTAINS_FINDING

USES_MECHANISM
CONTAINS_MECHANISM
BRIDGES_TO
REQUIRES
PREDICTS
FILLS_LATENT_STRUCTURE
ALTERNATIVE_TO
DISCRIMINATES_BETWEEN

HAS_LOCAL_CONCEPT
MAPS_TO_CANONICAL_ROLE
SAME_AS
SIMILAR_ROLE_AS
ANALOGOUS_TO
SYMBOLIZES
INTERPRETS
MYTHICALLY_INTERPRETS

CAN_BE_INTERRUPTED_BY
ROUTES_ENERGY_TO
REMEDIATES

4. Weights

4.1 Claim weights

confidence_weight =
    evidence_weight
  * source_quality_adjustment
  * independence_adjustment
  * specificity_adjustment
  * falsifiability_adjustment
  * contradiction_penalty
  * hazard_penalty

structure_watch_score =
    salience_weight
  * convergence_weight
  * mechanism_fit
  * anomaly_pressure
  * prediction_specificity
  * anti_overfit_penalty

fact_export_score =
    confidence_weight
  * source_quality_adjustment
  * falsifiability_adjustment
  * independence_adjustment
  * hazard_penalty

4.2 Latent structure score

latent_score =
    structural_fit
  * recurrence_across_subgraphs
  * independence_weight
  * mechanism_fit
  * temporal_coherence
  * prediction_specificity
  * anomaly_pressure
  * anti_overfit_penalty

4.3 Hypothesis openness condition

For any empirical hypothesis H:

P(H) > 0

and if the likelihood ratio accumulates without artificial cap:

Posterior odds(H1/H0) = Prior odds(H1/H0) × ∏ LR_i

then a classically dismissed hypothesis can defeat a consensus hypothesis.

5. Graphlet Templates

5.1 Whispering adversary

graphlet_id: graphlet_whispering_adversary
node_roles:
  - human_subject
  - adversarial_influence
  - whispered_command_or_intrusion
  - distorted_perception
  - destructive_action
  - remedy_practice
edge_roles:
  - INFLUENCES
  - WHISPERS_OR_IMPLANTS
  - DISTORTS
  - LEADS_TO
  - CAN_BE_INTERRUPTED_BY
negative_controls:
  - ordinary temptation without externalized agency
  - psychiatric intrusive thought without cultural adversary framing
  - metaphor-only uses

5.2 Controlled inversion

graphlet_id: graphlet_controlled_inversion
node_roles:
  - existing_order
  - temporary_revolt
  - licensed_transgression
  - pressure_release
  - restoration_of_order
edge_roles:
  - PERMITS
  - VENTS
  - CONTAINS
  - RESTORES

5.3 Manufactured criminality

graphlet_id: graphlet_manufactured_criminality
node_roles:
  - vulnerable_subject
  - recruiter_or_handler
  - isolating_group
  - confession_or_compromise
  - transgressive_act
  - blackmail_or_shame_lock
  - new_identity
  - deployment
edge_roles:
  - RECRUITS
  - ISOLATES
  - EXTRACTS_CONFESSION
  - INDUCES_TRANSGRESSION
  - LOCKS_WITH_COMPROMISE
  - REBUILDS_IDENTITY
  - DEPLOYS

5.4 Legitimacy laundering

graphlet_id: graphlet_legitimacy_laundering
node_roles:
  - sacred_or_moral_tradition
  - institution
  - political_project
  - public_legitimacy
  - hidden_operation
edge_roles:
  - BORROWS_AUTHORITY_FROM
  - CONFERS_LEGITIMACY_ON
  - MASKS
  - REDIRECTS_MORAL_ENERGY

5.5 Source suppression / sealed archive

graphlet_id: graphlet_source_suppression
node_roles:
  - scandal_or_anomaly
  - institution
  - archive_or_record
  - classification_or_sealing
  - public_narrative
  - unresolved_latent_structure
edge_roles:
  - GENERATES_RECORD
  - SEALS_OR_WITHHOLDS
  - SIMPLIFIES_PUBLIC_STORY
  - CREATES_LATENT_STRUCTURE

5.6 Purification / remediation

graphlet_id: graphlet_purification_remediation
node_roles:
  - hostile_activation
  - awareness
  - naming
  - downshift
  - generative_practice
  - restored_agency
edge_roles:
  - NOTICES
  - NAMES
  - INTERRUPTS
  - REDIRECTS
  - RESTORES

6. Anti-Overfit Rules

1. Symbolic similarity is not causation.
2. Motif recurrence is not entity identity.
3. Adjacency is not control.
4. Endpoint facts do not prove bridge claims.
5. Official denial is evidence, not final truth.
6. Conspiracy source is noisy, not null.
7. A source family copied ten times counts as one dependent source.
8. Every strong hypothesis must expose required missing links.
9. Every major hypothesis must have alternative explanations.
10. Every graphlet must be tested on out-of-sample traditions/cases.

7. Minimal JSON Schema Sketch

{
  "claim": {
    "claim_id": "string",
    "claim_text": "string",
    "claim_type": "fact|allegation|self_claim|symbolic_claim|bridge_claim|prediction",
    "ontology_layer": "historical|legal|institutional|mechanism|symbolic|metaphysical",
    "status": "documented|alleged|contested|inferred|symbolic|speculative|rejected",
    "prior_probability": 0.5,
    "posterior_probability": 0.5,
    "confidence_weight": 0.0,
    "salience_weight": 0.0,
    "structure_watch_score": 0.0,
    "fact_export_score": 0.0
  }
}

8. Design Philosophy

The graph is allowed to say:

This is real.
This is alleged.
This is symbolic.
This is missing.
This is dangerous to assert carelessly.
This weak claim is still structurally important.
This strong fact does not prove the large theory.
This graph cavity predicts where to look next.

That is the epistemic posture needed for strange-history inference.