The Self-Reporting Boundary: On Criteria That Report Their Own Failures
Year: 2026
Type: preprint
Abstract: A criterion that cannot say where it fails cannot be relied on outside the domain in which it was checked. Three literatures already implement the requirement in part: a declared domain of applicability in regulatory modelling, a declared strength of evidence in evidence grading, and the option to abstain in selective prediction. Two of them are institutional. The requirement's third clause — a counted, signature-separated account of failure — has no precedent in any of them, and that is where this note's increment begins. This note argues that the three cannot be merged. The separation is not one of degree: a declared domain is a property of a model and can be redrawn after the model exists; a grade of evidence is assigned after the studies are collected; but the property required here belongs to a verdict, is issued with it, and is not a number — it cannot be averaged, pooled, or carried to a corpus. What the refusal to merge yields is stated as two things the neighbouring systems do not have: an out-of-bounds register that separates two directions with opposite signatures instead of one distance, and a vacuous verdict that is distinguished from a thin one when both arrive as the same formal sign. The contribution is a definition and a short argument for its necessity, not a theorem; three criteria of one programme illustrate the register and share none of its premises.
Author: Li, GenMin
Language: English
Subfield: Philosophy
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7236753562
