article: id: CA-UI-006 title: Evidence and Provenance Schema version: 1.0 status: canonical type: specification parent_protocol: CA-UI-001 depends_on: - CA-UI-002 - CA-UI-003 - CA-UI-004 - CA-UI-005
Purpose
purpose: > Ensure that every Atlas object, coordinate, relationship, dependency, crosswalk and diagnosis remains traceable to its evidence, source, interpretation and uncertainty.
NO CLAIM WITHOUT PROVENANCE.NO COORDINATE WITHOUT EVIDENCE.NO SYNTHESIS WITHOUT TRACEABILITY.NO UNCERTAINTY HIDDEN FROM THE USER.
Evidence architecture
Source ↓Evidence Record ↓Observation ↓Interpretation ↓Claim ↓Atlas Mapping ↓Synthesis
evidence_architecture: source_layer: where_information_originated evidence_layer: what_information_was_obtained interpretation_layer: what_the_information_may_mean claim_layer: what_is_being_asserted atlas_layer: where_the_claim_is_mapped synthesis_layer: how_multiple_claims_are_combined
Core separation rule
SOURCE ≠ EVIDENCEEVIDENCE ≠ INTERPRETATIONINTERPRETATION ≠ FACTCONSENSUS ≠ CERTAINTYABSENCE OF EVIDENCE ≠ EVIDENCE OF ABSENCE
separation_rule: source_and_evidence_separate: true evidence_and_interpretation_separate: true claim_and_coordinate_separate: true uncertainty_required: true disagreement_preserved: true
Source record
source_record: source_id: required title: required source_type: required authors: [] organisation: optional publication_date: optional retrieval_date: required language: required identifiers: doi: optional isbn: optional url: optional archive_id: optional dataset_id: optional publication_status: - published - peer_reviewed - preprint - archival - primary_record - unpublished - oral_history - machine_generated - unknown scope: spatial: optional temporal: optional disciplinary: optional access: - open - restricted - private - unavailable version: required
Source types
source_types: primary: - artefact - inscription - legal_record - administrative_record - census - direct_observation - sensor_reading - interview - experiment - original_dataset secondary: - research_article - scholarly_book - systematic_review - historical_analysis - technical_report tertiary: - encyclopedia - textbook - reference_database - summary_article computational: - simulation - model_output - algorithmic_classification - machine_generated_summary interpretive: - expert_commentary - theoretical_argument - comparative_analysis - scenario
Evidence record
evidence_record: evidence_id: required source_id: required evidence_type: required extracted_content: description: required exact_location: required quoted_text: optional table_reference: optional figure_reference: optional dataset_fields: optional supports: object_ids: [] claim_ids: [] relationship_ids: [] coordinate_ids: [] crosswalk_ids: [] temporal_scope: required spatial_scope: optional system_level: required method: collection_method: required analytical_method: optional sample_size: optional measurement_unit: optional quality: source_quality: required directness: required methodological_strength: required reproducibility: optional independence: required uncertainty: required confidence: required version: required
Evidence types
evidence_types: - archaeological - documentary - statistical - observational - experimental - comparative - computational - ecological - infrastructural - demographic - economic - institutional - ethnographic - theoretical - expert_interpretation
Evidence directness
evidence_directness: DIRECT: definition: > The evidence directly observes or measures the claim. PROXY: definition: > The evidence measures a related variable used as an indicator for the claim. INFERRED: definition: > The evidence supports the claim only through an explicit reasoning chain. ANALOGICAL: definition: > The evidence comes from a comparable object or system. SPECULATIVE: definition: > The claim extends beyond currently available evidence.
Claim contract
claim: claim_id: required statement: required claim_type: required subject_object: required predicate: required object_or_value: required scope: spatial: required temporal: required system_level: required evidence_ids: required interpretation_ids: optional assumptions: [] counterevidence_ids: [] confidence: required status: required version: required
Claim types
claim_types: - identity - classification - measurement - relationship - dependency - causal - comparative - coordinate_assignment - transition - historical - diagnostic - predictive - normative
Claim status
claim_status: observed: definition: directly_supported strongly_supported: definition: supported_by_multiple_independent_sources provisional: definition: plausible_but_incomplete disputed: definition: credible_competing_positions_exist speculative: definition: weak_or_future_facing_evidence rejected: definition: contradicted_or_invalidated superseded: definition: replaced_by_a_newer_claim
Interpretation contract
interpretation: interpretation_id: required claim_id: required interpretation_text: required interpreter: person: optional organisation: optional theory_id: optional atlas_runtime: optional reasoning_type: - deductive - inductive - abductive - comparative - causal - analogical - model_based evidence_ids: required assumptions: [] alternative_interpretations: [] limitations: [] confidence: required status: required
ONE EVIDENCE RECORD MAY SUPPORTMULTIPLE INTERPRETATIONS.
Provenance chain
provenance_chain: provenance_id: required source_id: required evidence_id: required interpretation_id: optional claim_id: required atlas_target_id: required transformations: - transformation_id: required operation: required performed_by: required timestamp: required assumptions: [] information_loss: required final_output: object_id: optional coordinate_id: optional relationship_id: optional diagnosis_id: optional answer_id: optional
SOURCE→ EXTRACTION→ NORMALISATION→ INTERPRETATION→ CLAIM→ ATLAS MAPPING→ OUTPUT
Transformation operations
transformation_operations: - extract - translate - normalise - aggregate - disaggregate - classify - infer - compare - crosswalk - calculate - summarise - synthesise
Information-loss contract
information_loss: level: - none - minor - moderate - substantial - unknown lost_elements: [] retained_elements: [] consequence: required
EVERY SUMMARY LOSES DETAIL.THE SCHEMA MUST RECORD WHAT WAS LOST.
Confidence model
confidence_record: confidence_id: required target_id: required components: source_quality: 0.0_to_1.0 evidence_directness: 0.0_to_1.0 methodological_strength: 0.0_to_1.0 source_independence: 0.0_to_1.0 cross_source_agreement: 0.0_to_1.0 temporal_relevance: 0.0_to_1.0 scale_match: 0.0_to_1.0 penalties: interpretive_distance: 0.0_to_1.0 unresolved_conflict: 0.0_to_1.0 missing_data: 0.0_to_1.0 overall_score: derived confidence_class: required
overall_confidence = weighted_mean( source_quality, evidence_directness, methodological_strength, source_independence, cross_source_agreement, temporal_relevance, scale_match ) - penalties
Confidence classes
confidence_classes: VERY_HIGH: range: 0.90_to_1.00 HIGH: range: 0.75_to_0.89 MEDIUM: range: 0.50_to_0.74 LOW: range: 0.25_to_0.49 VERY_LOW: range: 0.01_to_0.24 UNKNOWN: range: unavailable
Source-quality dimensions
source_quality_dimensions: authorship: - known - credible - relevant_expertise - unknown methodology: - transparent - reproducible - partially_transparent - opaque evidence_access: - full - partial - summary_only - unavailable review_status: - independently_reviewed - editorially_reviewed - unreviewed - unknown conflict_of_interest: - absent - declared - possible - significant - unknown
Independence test
source_independence: independent: definition: separately_collected_or_derived partially_independent: definition: shares_some_data_or_assumptions dependent: definition: repeats_or_derives_from_same_origin unknown: definition: lineage_not_clear
TEN SOURCES COPYING ONE SOURCEDO NOT EQUAL TEN INDEPENDENT SOURCES.
Corroboration contract
corroboration: corroboration_id: required claim_id: required supporting_sources: [] contradicting_sources: [] dependent_sources: [] independent_sources: [] agreement: - strong - moderate - weak - divided - unknown unresolved_questions: [] resulting_confidence: required
Conflict contract
evidence_conflict: conflict_id: required target_claim: required positions: - position_id: required statement: required evidence_ids: required confidence: required conflict_type: - factual - definitional - methodological - chronological - causal - scale - interpretation - measurement likely_reason: - different_data - different_scale - different_period - different_definition - measurement_error - political_interest - incomplete_evidence - unknown resolution_status: - unresolved - partly_resolved - context_dependent - scale_dependent - superseded
Contradiction handling
function handle_conflict(claim): positions = retrieve_competing_positions(claim) for position in positions: preserve(position.statement) preserve(position.evidence) calculate(position.confidence) determine: shared_facts disputed_facts differing_definitions scale_differences temporal_differences return: consensus disagreement uncertainty unresolved_questions
THE ATLAS DOES NOT VOTE AWAY DISAGREEMENT.
Temporal relevance
temporal_relevance: observation_date: required target_date: required relevance: - current - recent - historically_relevant - outdated - temporally_incompatible decay_rate: - none - slow - moderate - rapid
Example:
temporal_relevance: observation_date: 1960 target_date: 2026 relevance: historically_relevant decay_rate: rapid
OLD EVIDENCE MAY REMAIN VALID FOR HISTORY.OLD EVIDENCE MAY BE INVALID FOR PRESENT STATE.
Scale-match rule
scale_match: evidence_level: required claim_level: required match: - exact - near - extrapolated_upward - extrapolated_downward - incompatible justification: required
INDIVIDUAL DATADOES NOT AUTOMATICALLY DESCRIBE A CIVILISATION.PLANETARY DATADOES NOT AUTOMATICALLY DESCRIBE EVERY COMMUNITY.
Coordinate evidence contract
coordinate_evidence: coordinate_id: required target_object: required capability: claim_ids: required evidence_ids: required confidence: required lifecycle: claim_ids: required evidence_ids: required confidence: required habitat: claim_ids: required evidence_ids: required confidence: required observation_date: required competing_coordinates: []
Relationship evidence contract
relationship_evidence: relationship_id: required source_object_evidence: required predicate_evidence: required target_object_evidence: required direction_evidence: required strength_evidence: optional temporal_evidence: required causal_mechanism_evidence: optional counterevidence_ids: [] confidence: required
Inference contract
inference: inference_id: required input_claims: required output_claim: required inference_type: - deductive - probabilistic - causal - comparative - analogical - model_generated reasoning_summary: required assumptions: required uncertainty: required confidence: required
function infer(inputs): require(valid_claims(inputs)) require(explicit_assumptions) require(reasoning_summary) output = apply_inference_rule(inputs) output.status = "inferred" output.confidence = minimum(input_confidence) - inference_penalty return output
Machine-generated evidence rule
machine_generated_content: may_be_source: true may_be_evidence: conditional automatically_verified: false required: - originating_model - generation_date - input_provenance - validation_method - human_or_external_check
AI OUTPUT IS NOT SELF-VALIDATING.
Citation object
citation: citation_id: required source_id: required evidence_id: optional locator: page: optional section: optional paragraph: optional line: optional timestamp: optional table: optional figure: optional citation_style: optional access_date: required
Answer provenance
answer_provenance: answer_id: required query_id: required generated_at: required used_objects: [] used_coordinates: [] used_relationships: [] used_crosswalks: [] used_claims: [] used_evidence: [] used_sources: [] inferences: [] excluded_sources: [] unresolved_conflicts: [] uncertainty_summary: required
Provenance query runtime
query: "Why did the Atlas assign this coordinate?"return: coordinate_claims evidence_records source_records transformations interpretations confidence competing_assignments
query: "Where did this statement come from?"return: claim provenance_chain exact_source_location transformation_history
query: "How certain is this?"return: confidence_components evidence_gaps disagreements temporal_relevance scale_match
Validation runtime
function validate_evidence_record(record): require(record.evidence_id) require(valid_source(record.source_id)) require(record.extracted_content) require(record.exact_location) require(record.temporal_scope) require(record.system_level) require(record.quality) require(record.uncertainty) require(record.confidence) reject(untraceable_claim) reject(source_without_retrieval_date) reject(interpretation_marked_as_observation) reject(causal_claim_without_mechanism) reject(summary_without_information_loss_record) reject(machine_generated_claim_without_validation) reject(scale_extrapolation_without_justification) return VALID
Minimum viable provenance
minimum_viable_provenance: required: - source_id - evidence_id - exact_source_location - claim_id - atlas_target_id - confidence - observation_date - transformation_history
Canonical rule
canonical_rule: > Every active Atlas claim must remain traceable from output to claim, interpretation, evidence and originating source. Evidence quality, scale, date, independence, disagreement, transformation and uncertainty must remain visible.
Runtime summary
THE SOURCE SHOWS WHERE INFORMATION BEGAN.THE EVIDENCE SHOWS WHAT WAS OBSERVED.THE INTERPRETATION SHOWS WHAT IT MAY MEAN.THE CLAIM SHOWS WHAT IS ASSERTED.THE PROVENANCE CHAIN SHOWS HOW IT TRAVELLED.CONFIDENCE SHOWS HOW MUCH TRUST IT DESERVES.DISAGREEMENT REMAINS VISIBLE.THE ATLAS CONNECTS KNOWLEDGEWITHOUT BREAKING ITS LINEAGE.