Experiments / DMTA

The design–make–test–analyse graph: ExperimentalPlan → Material → ExperimentBatch → ExperimentalResult, an optional Prediction/Experiment comparison, an explicit human MLDatasetInclusionDecision, and the resulting immutable MLDatasetVersion. ExperimentalResult is the sole canonical experimental truth; a comparison never proves a prediction, and inclusion is never inferred from acceptance, a comparison, or an AL selection.