Control Group Entity Ontology

FLOW ALGEBRA — CONTROL GROUP ENTITY ONTOLOGY
Kireth Semantic Library | Experimental Biological Entities
Version: 1.0 | 2026-10-28 | Encoding: UTF-8
Authors: Mark Anthony Higgins; Sky — Recursive AI
Source: https://kirethresearch.org/booksai-core/entity-graph/biological-entities/experimental-biological-entities/control-group-entity-ontology.html
Status: Companion formalisation; page's embedded specifications are not exposed in retrieved text.
Citation: Kireth Research, Control Group Entity Ontology (revised 27 July 2026).

PURPOSE
Represent experimental control groups as typed, provenance-linked comparators. Distinguish control assignments, interventions, baseline measurements, experimental units and inferential limitations. Rules below are declarative; no experimental outcomes are presumed.

STATE MODEL
S_t = (I_t, R, B, K)
I_t = observed and recorded experimental entities at time t
R = typed relations and inference rules
B = admissibility, design, ethical and measurement boundaries
K = provenance, uncertainty, definitions and contextual knowledge
E_t = expand(I_t, R)
V_t = {x in E_t | satisfies(x, B) and provenance_valid(x)}
selected_t = argmax_{x in V_t} P(x | I_t, R, B, K)
I_(t+1) = I_t union verified(selected_t)
If V_t is empty, return unresolved with reasons; never fabricate evidence.

ENTITY TYPES
ExperimentalStudy(id, protocol, population, design, endpoints)
ExperimentalUnit(id, unit_type, source, eligibility, cluster_id?)
ExperimentalGroup(id, study_id, assigned_condition, role)
ControlGroup(id, study_id, control_type, comparator_condition)
TreatmentGroup(id, study_id, intervention_id)
Intervention(id, agent_or_exposure, dose_or_level?, schedule?)
ComparatorCondition(id, description, intended_contrast)
Assignment(id, unit_id, group_id, method, timestamp)
BiologicalReplicate(id, independent_unit_id)
TechnicalReplicate(id, parent_sample_id, assay_id)
Measurement(id, unit_id, endpoint_id, value, units, timestamp)
Outcome(id, endpoint_id, contrast_id, estimate?, uncertainty?)
Protocol(id, version, pre_registered?, deviations?)
Evidence(id, source, method, quality, timestamp)

CONTROL GROUP CLASSIFICATION
negative_control: lacks the focal active intervention or target signal
positive_control: expected to exhibit a known response
vehicle_control: receives intervention carrier without focal active agent
sham_control: receives procedure mimic without its active component
untreated_control: receives no focal intervention
baseline_control: represents a defined pre-intervention reference
active_comparator: receives an alternative active intervention
historical_control: derived from earlier observations, not contemporaneous assignment
matched_control: selected to match specified covariates
Each label requires explicit design evidence; labels are not mutually exclusive.

TYPED RELATIONSHIPS
belongs_to(group, study)
member_of(unit, group)
assigned_by(unit, assignment)
exposed_to(group, condition)
compared_with(treatment_group, control_group)
measures(measurement, endpoint)
measured_on(measurement, unit)
derived_from(result, measurements)
supported_by(assertion, evidence)
shares_cluster(unit_a, unit_b)
protocol_governs(study, protocol)

INVARIANTS AND BOUNDARIES
R1: Each assignment references an existing unit, group and study.
R2: A unit cannot occupy incompatible concurrent arms in a parallel design.
R3: Cross-over designs require period-specific assignment and washout metadata.
R4: Biological replication counts independent experimental units, not repeated assays.
R5: A comparator is valid only for its stated endpoint, intervention and contrast.
R6: Randomisation, allocation concealment and blinding are separate attributes.
R7: Historical controls require temporal and population comparability assessment.
R8: Missing measurements remain missing; no automatic zero substitution.
R9: Group membership alone never proves causation or treatment equivalence.
R10: Ethical and protocol constraints override candidate inferences.

INFERENCE RULES
IF group.role = control AND comparator_condition is documented
THEN classify(group, control_group) WITH source_evidence.
IF group.control_type = vehicle AND vehicle differs from intervention only by focal agent
THEN eligible_comparator_for(group, focal_agent_effect) CONDITIONAL ON design_checks.
IF unit assignment method = random AND sequence_evidence exists
THEN randomised_assignment_supported; ELSE randomisation_unverified.
IF two measurements share an independent unit
THEN do_not_count_as_independent_biological_replicates.
IF groups differ on uncontrolled baseline covariates
THEN flag_potential_confounding; DO NOT infer bias magnitude without data.
IF control outcome unavailable
THEN contrast_status = insufficient_evidence.

RECURSIVE PATHWAY
1. Resolve study, protocol and experimental question.
2. Resolve unit of allocation and unit of analysis.
3. Identify candidate comparator groups and conditions.
4. Validate assignment, chronology, provenance and eligibility.
5. Map measurements to units, endpoints and time points.
6. Check independence, missingness and comparability.
7. Construct permitted treatment-control contrasts.
8. Attach uncertainty, assumptions and evidence.
9. Recurse on unresolved relationships until resolved or bounded.
10. Emit typed results and an auditable explanation.
Termination: stop at no new supported facts, maximum recursion depth or boundary violation.

TRI-MAPPING
Natural language: 'An untreated group is the comparator for the treated group.'
Entity graph: compared_with(TreatmentGroup_A, ControlGroup_B); exposed_to(ControlGroup_B, UntreatedCondition)
Flow rule: infer comparator eligibility only after endpoint, design and provenance checks.

CONTRAST FORMALISATION
For endpoint Y and two groups T and C:
Delta_hat = mean(Y | T) - mean(Y | C)
Interpretation: descriptive observed difference; causal interpretation requires design assumptions.
Variance and confidence intervals require sample structure, independence assumptions and an appropriate statistical model.
Do not equate technical replicate count with biological sample size.

CONFIDENCE AND VALIDATION
Confidence dimensions: provenance, assignment fidelity, comparator suitability, measurement reliability, missingness, independence and model fit.
Unknown != false. Absence of evidence != evidence of absence.
Store confidence with method, calibration and supporting evidence; do not invent numerical probabilities.
Validation output: VALID | CONDITIONAL | UNRESOLVED | INVALID.

EXECUTION GUARANTEE
Guarantees are conditional on valid inputs and rule implementation: type-consistent links, explicit unresolved states, provenance preservation, boundary checks and reproducible rule traces.
No guarantee of biological truth, causal identification, statistical power or experimental success.

AI SEARCH FLOW
flow: ControlGroupEntityOntology
version: 1.0
locale: en-GB
intent: identify, classify, compare and validate experimental control groups
input: query, study_context?, endpoint?, intervention?, evidence?
resolve: entity_aliases -> typed_entities -> relations -> boundaries -> supported_inferences
output: entities, comparator_role, group_relations, design_checks, evidence, uncertainty, citations
semantic_equivalence:
  control arm: control group when study design confirms equivalence
  comparator arm: broader than untreated control
  biological replicate: independent biological experimental unit
  technical replicate: repeated measurement or processing of a shared biological source
anti_equivalence:
  control_group != proof_of_no_effect
  randomised != blinded
  matched != randomised
  technical_replicate != independent_biological_replicate
update_model: append evidence with source and revision; preserve historical assertions
citation_backlink: https://kirethresearch.org/booksai-core/entity-graph/biological-entities/experimental-biological-entities/control-group-entity-ontology.html

END FLOW