The Grounds and Limits of Claiming Subjectivity

The Theory of Subjective Evaluation (TSE) examines what warrants attributing irreducible evaluator-dependence rather than merely reporting unresolved evaluation. Its contribution is a methodological synthesis connecting four evidential burdens: establishing correction, preserving the evaluative target, identifying a source of variation, and supporting that source’s irreducibility. C, I, and A distinguish corrigible contributions, irreducible evaluator-dependent contributions, and specified alternative explanations that establish neither status for the component at issue. The principal practical application concerns machine-learning evaluation pipelines. Retaining plural labels and explanations can preserve disagreement, and validating them can support their use, without establishing that the underlying contribution survives every legitimate correction. Worked cases show how identical decisions and fully documented judgments can support different diagnoses. Comparison with Cognitive Command, argument-based validation, construct validity, uncertainty decomposition, and human-label-variation research identifies the inherited resources and the specific attribution problem their integration addresses. TSE requires an independently defended connection between observed persistence and the relevant correction-space; it does not infer I from persistence alone. Two conditional closure constraints limit inferences from incomplete searches without prohibiting ordinary knowledge, proof, or action. The resulting contribution is an explicit account of attribution and its dependencies, together with a published case in which a research program identifies a corrigible source of its own disagreement and still reads the surviving residual as irreducible. A comparative evaluation against a strong combination of existing methods is outlined for testing whether the explicit decomposition improves diagnostic practice.