The Dark Sector as a Classification Error

Dark matter and dark energy are usually introduced through empirically successful metric cosmological models, especially ΛCDM. The usual inference is then strengthened into an ontological claim: because additional gravitating or accelerating terms are indispensable to the model, its sector terms must be invariantly extracted as primitive constituents of the universe. This paper argues that the strengthening is not admissible. The target is narrow. We do not deny galactic dynamics, gravitational lensing, the cosmic microwave background, baryon acoustic structure, supernova acceleration, or the predictive success of ΛCDM. We ask only whether the role terms “dark matter” and “dark energy”, as they function in metric surrogate inference, achieve representation-independent extraction sufficient for primitive ontological status. The answer is negative. A dark-sector term is introduced as a stabilizing residual inside a non-invariant decomposition of witness constraints into metric geometry, known stress-energy, auxiliary source terms, boundary data, calibration records, and error discipline. Over the restricted observational regimes motivating the corresponding dark-sector inference, and for fixed comparison targets, admissible redescriptions can preserve the relevant witness content while redistributing the stabilizing load among these coordinates: an added stress-energy component, a modified gravitational law, an effective curvature term, an emergent variable, a calibration or transport correction, or a changed boundary ledger. This is not a claim of global empirical equivalence among complete cosmological theories. It is a local non-extraction result: a term whose identity is fixed only by this load-bearing role is not standing-preserving. It is a surrogatedependent stabilization structure, not an entity-class thereby invariantly extracted. The resulting conclusion is conditional but sharp. If a future particle, field, or invariant structure is extracted by a standing-preserving witness protocol, that object may receive its own ontological assessment. The presently inferred dark-sector structure does not receive that status from current metric-surrogate indispensability. Dark-sector terms have predictive necessity inside successful representations; they do not thereby achieve invariant primitive ontological extraction. Treating the presently inferred dark-sector structure as already invariantly extracted primitive structure is a classification error: it mistakes a surrogate stabilizer for an invariantly extracted entity-class. The result is stronger than ordinary underdetermination because it identifies a structural failure of invariant extraction, not merely a plurality of empirically adequate stories.