Examination as Bounded Inference: An Ontology of the Ideal Examination and Its Deviations

I offer a conceptual account of what an examination is, as distinct from how examinations are validated or used. I argue that an examination is constitutively an act of bounded inference: it exists only in the gap between a person’s true competence and an observer’s capacity to read that competence from a finite, interference-laden sample of enacted performance. From this single premise I derive a theoretical ideal—the derived perfect examination (which I term the Exetas)— defined as the examination in which the currently active sources of inferential uncertainty are driven to zero. I show that this ideal coincides with the limit of exhaustive coverage, at which the activity ceases to be inference and becomes direct enumeration: at total coverage there is nothing left to infer. This yields a three-way distinction between inspection (reading a specification without demonstration), the Exetas (full demonstration with zero inference), and the real examination (partial demonstration requiring inference). I argue that the ideal is derived not from any fixed list of limitations but from whatever set of limitations is active in a given era, so that the domain over which it is realizable expands as technology removes limitations, while the ideal itself is invariant. A consequence is that no large-scale examination can be perfect under present constraints. I situate the account within argument-based and consequential theories of validity (Kane; Messick) and within generalizability theory (Cronbach; Brennan), arguing that the contribution is ontological and organizational rather than a new theory of measurement. I close by distinguishing measurement questions, which concern the accuracy of inference, from ethical questions, which concern the distribution and use of accurate inferences, and propose a single boundary criterion separating the two.