Grounded Uncertainty: Graded Truth for the Indeterminate Status of Synthetic Subjects

Claims about the ontological status of a synthetic subject (whether it is a subject, whether it is conscious) are today reported either as a definite yes/no or as a single probability p. Both are dishonest in the face of an irreducible undecidability we denote diamond-C. We propose to report such claims as graded, revisable truth, a pair (frequency, confidence), borrowed from the Non-Axiomatic Reasoning System (NARS, Pei Wang). The key distinction, unavailable to a single probability, is between '0.5 from an absence of evidence' (confidence low) and '0.5 from balanced strong evidence' (confidence high). The second we call grounded uncertainty, and we argue that it is precisely the honest form in which the indeterminate status should be reported. The method is accompanied by a runnable artifact, a tiny language, nolang, over NARS, in which an agent's assertions and queries execute as reasoning under uncertainty. A companion to 'Indeterminate Ontologies of Synthetic Subjects' (DOI 10.5281/zenodo.21288590).