The Theory of Data

Abstract

A value is not yet data. The number 42 becomes data only when it is bound to something it is about: age = 42, and more completely, John's age = 42. This simple observation leads to a larger problem. Analytical systems manipulate values easily, but they often leave implicit the analytical objects those values are supposed to represent: what points exist, how those points are partitioned, what quantity exists at a partition, which transformations preserve that quantity, and when two derivations should count as derivations of the same thing. The Theory of Data (ToD) develops a foundation for governing those questions. A universe establishes a governed root-point domain under an explicit existence law, together with its primitive analytical core. An anchor is a governed partition of the universe. A measure family is a uniquely identified and uniquely named coherent analytical family. A measure is that family at an anchor, written F@A. Two primary value operators act on measures. A mapper preserves anchor. A reducer contracts a strictly finer anchor to a coarser anchor. Reducers are governed through sufficient state rather than displayed values alone. Inside a measure family, lawful reductions share one coherent commutative-monoid state law. This makes staged reduction path-independent and allows family-internal derivation paths to be compressed without changing analytical identity. The family is therefore the basic algebraic unit of analytical coherence. At the universe level, well-formed constitutive lineage organizes measure families as a directed acyclic lineage graph. Its nodes are uniquely named measure-family identities; its directed edges record family-establishing analytical ancestry. The lineage graph is not outside analytical governance. It is the governed structure that determines where measure families come from and how different analytical identities are related. Two family-changing derivations that produce analytically different quantities do not constitute inconsistent paths to one measure: they establish different measure-family identities and therefore must resolve to different governed family names. This yields a compact account of analytical identity. A family has a unique canonical identity and unique governed name; a measure is that family at an anchor. Consistency is therefore identity-relative: if two lawful derivations canonicalize to the same family @ anchor, they claim the same analytical object and must agree; if they canonicalize to different identities, they are different measures and equality is not required. Expanded canonical ancestry may be shown when explanation requires it, but ordinary analytical reference needs only family @ anchor. The paper develops this foundation, its coherence result, the lineage model, and the principal boundaries of analytical governance. It is intentionally a foundation paper rather than a manual for every analytical operation or implementation case.