Experimental Group Entity Ontology

FLOW ALGEBRA — EXPERIMENTAL GROUP ENTITY ONTOLOGY
Version: 1.0 | Date: 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/experimental-group-entity-ontology.html
Status: Companion formalisation based on the visible ontology title and established experimental-design concepts; not a verbatim extraction of hidden page code.
Citation: Kireth Research, Experimental Group Entity Ontology (revised 27 July 2026).

PURPOSE
Represent experimental groups as traceable, non-overlapping (where design requires), explicitly defined sets of experimental units, treatments, assignments and observations. Preserve independence, provenance, and uncertainty in downstream reasoning.

FLOW "ExperimentalGroupEntityOntology" v1.0

STATE_MODEL
  S_t = (I_t, R_t, B_t, K_t)
  I_t = set of experimental entities, group definitions, assignments and observations
  R_t = typed relationships and validated constraints
  B_t = boundary conditions: protocol, population, eligibility, time, setting
  K_t = provenance, measurement metadata, evidence quality and uncertainty
  E_t = expand(I_t, R_t)
  V_t = {x in E_t | constraints(x, B_t, K_t) = true}
  x*_t = argmax_{x in V_t} P(x | I_t, R_t, B_t, K_t)
  I_(t+1) = update(I_t, x*_t, new_evidence)
  STOP when no new validated assertions arise or recursion limit is reached.

ENTITIES
  Experiment: protocol-governed investigation
  ExperimentalGroup: explicitly defined collection of experimental units
  ExperimentalUnit: smallest independently assigned unit of intervention
  BiologicalUnit: organism, culture, specimen or other biological source
  TreatmentGroup: group assigned a defined intervention or exposure
  ControlGroup: comparator under defined control conditions
  ShamGroup: group receiving sham procedure where applicable
  ReferenceGroup: contextual comparison, not necessarily a randomised control
  Cohort: population subset defined by inclusion and exclusion criteria
  Assignment: link between unit, group, method, date and allocation record
  Intervention: treatment, exposure or manipulation
  Outcome: prespecified measured response
  Observation: recorded measurement with units, timing and provenance
  BiologicalReplicate: independent biological source or experimental unit
  TechnicalReplicate: repeated measurement of the same biological material
  Batch: shared processing context that may introduce correlated effects
  Randomisation: allocation method with reproducible metadata
  Blinding: concealment of assignments from specified roles
  Confounder: variable associated with exposure and outcome
  ProtocolDeviation: departure from prespecified experimental plan

RELATIONSHIPS
  Experiment HAS_GROUP ExperimentalGroup
  ExperimentalGroup CONTAINS ExperimentalUnit
  ExperimentalUnit DERIVED_FROM BiologicalUnit
  ExperimentalUnit ASSIGNED_BY Assignment
  Assignment ASSIGNS_TO ExperimentalGroup
  ExperimentalGroup RECEIVES Intervention
  ExperimentalUnit HAS_OUTCOME Outcome
  Outcome RECORDED_AS Observation
  Observation PRODUCED_IN Batch
  ExperimentalGroup COMPARED_WITH ExperimentalGroup
  ExperimentalGroup DEFINED_BY CohortCriteria
  Observation SUPPORTED_BY ProvenanceRecord
  ExperimentalUnit MAY_HAVE BiologicalReplicate
  Observation MAY_HAVE TechnicalReplicate

INVARIANTS
  Each group has a stable identifier and documented inclusion rules.
  Each unit has a stable identifier and an explicit experimental-unit definition.
  An assignment belongs to one experimental allocation at a time unless the protocol explicitly specifies crossover or repeated-measures design.
  Technical replicates must not be counted as independent biological replicates.
  Group membership must not be inferred solely from an observed outcome.
  Missing assignment, protocol or provenance remains UNKNOWN, never silently assumed.
  Randomisation and blinding claims require evidence; group labels alone do not prove them.

PATHWAYS
  DEFINE_PROTOCOL -> DEFINE_POPULATION -> DEFINE_EXPERIMENTAL_UNIT
  -> DEFINE_GROUPS -> APPLY_ELIGIBILITY -> RECORD_ASSIGNMENTS
  -> RECORD_EXPOSURES -> CAPTURE_OBSERVATIONS
  -> VALIDATE_PROVENANCE -> ASSESS_COMPARABILITY
  -> REPORT_ESTIMATES_WITH_UNCERTAINTY

INFERENCE_RULES
  IF unit assignment is documented AND group criteria satisfied
     THEN assert MEMBER_OF(unit, group) with provenance.
  IF two groups share prespecified comparator context
     THEN assert COMPARABLE_BY_DESIGN, not EQUIVALENT_IN_OUTCOME.
  IF treatment allocation was randomised AND evidence exists
     THEN annotate RANDOMISED_ASSIGNMENT.
  IF unit-level measurements are repeated
     THEN model within-unit dependence; do not inflate independent sample size.
  IF units share a batch or source
     THEN flag potential clustering or batch effect.
  IF eligibility or allocation is ambiguous
     THEN flag REVIEW_REQUIRED and reduce confidence.
  IF control group differs in a relevant baseline factor
     THEN flag POSSIBLE_CONFOUNDING, not automatic causation.
  IF missing outcomes or protocol deviations exist
     THEN preserve missingness and deviation metadata.

TRI_MAPPING
  Human: Experimental groups organise units into documented treatment and comparison sets.
  Semantic: ExperimentalGroup --CONTAINS--> ExperimentalUnit --HAS_OUTCOME--> Observation.
  Formal: G_j = {u in U : assigned(u) = j AND eligible(u, protocol)}.

MINI_RECURSION_ENGINE
  INPUT: experiment_id, group_id, known_records
  1. Resolve experiment, protocol, units, assignments and group definitions.
  2. Expand typed links to interventions, outcomes, batches and provenance.
  3. Filter invalid or unsupported membership assertions.
  4. Check independence, overlap, missing data and confounding flags.
  5. Attach evidence, uncertainty and revision history to conclusions.
  6. Repeat only for newly validated entities; stop at fixed point.

CONFIDENCE
  confidence(assertion) = f(source_quality, protocol_match,
                            provenance_completeness, consistency,
                            independence_assessment)
  Confidence is an evidence-quality annotation, not a fabricated probability.
  Unknown evidence remains UNKNOWN; no numerical confidence without calibration.

EXECUTION_GUARANTEE
  Deterministic rule ordering for the same validated inputs.
  Preserve original source records and group identifiers.
  No ungrounded causal inference, statistical significance or treatment efficacy claims.
  Terminate at fixed point or bounded recursion depth.
  Report constraint failures and unresolved classifications explicitly.

AI_SEARCH_FLOW
  INTENT: identify, classify, compare or audit experimental groups.
  QUERY_FIELDS: experiment, group, treatment, comparator, unit, population,
                outcome, replicate, randomisation, blinding, batch, provenance.
  RETRIEVAL: exact identifiers -> semantic equivalents -> typed neighbours.
  FILTERS: protocol scope, organism or model, measurement date, evidence quality.
  OUTPUT: matched entities, group membership, relationships, validation flags,
          evidence links, uncertainties and source citation.
  UPDATE_MODEL: append provenance-stamped revisions; never overwrite raw observations.
  SEMANTIC_EQUIVALENCE: treatment arm ~ intervention group (context-dependent);
                        control arm ~ comparator group (context-dependent).
  DISTINCT_FROM: technical replicate, biological replicate, cohort, batch,
                 specimen and individual observation.

END FLOW