AIM Framework glossary
Referent Absence Property
Canonical definition
The Referent Absence Property is the result that the Self-Model lacks the World-Model's structural guarantee of decoherence against reality. Even where exogenous evidence participates in Self-Model formation, three features attenuate its corrective force: (1) mixing, where Source Opacity ensures exogenous evidence is processed identically to endogenous evidence upon entry; (2) contingent decoherence, where whether a decisive reality-contact event will ever arrive for any given identity-element cannot be known in advance; and (3) the evaluative layer, where most identity-relevant content contains evaluative components for which no mind-independent Referent exists. The result is a gradient from reliably-corrected physical Self-Model elements (Referent-Present Content) to effectively uncorrectable evaluative identity-elements (Referent-Absent Content), with the majority of identity-relevant content sitting toward the uncorrectable end. The normative criterion for such content is therefore not accuracy but subcriticality. The Property is what makes the Self-Model's identity-Content constitutive rather than confabulated: in the source-attribution domain a Referent exists and is merely hidden by Source Opacity, whereas for evaluative identity-Content there is no Referent to reach, so the mechanism's phenomenologically certain outputs are constructions rather than descriptions. Distinguishing the two senses of "Referent." This Property must not be run together with Referent Abandonment or the Diversity Referent, because "Referent" does different work in each, and keeping them apart is exactly what lets a Belief be placed in the right regime. The Referent Absence Property concerns whether an external ground-truth exists at all against which a given item of self-content could be corrected; for evaluative identity-content, none does — the Referent is genuinely absent. Referent Abandonment and the Diversity Referent concern the opposite condition: a determinate fact about the world that is available and has already returned its verdict — such as the non-modality of human variation — which a Belief nonetheless defies. So a self-Model element can be Referent-absent in the first sense (there is no external corrector for "I am the kind of person who belongs here"), while a kind-claim about a population is Referent-abandoned in the second sense (the non-modal distribution is available and is being defied). The first is the absence of a corrector; the second is the defiance of an available verdict. The Hardness–Referent Matrix classifies Beliefs on exactly this second axis (whether a live Referent governs, a decohered Referent is defied, no Referent was ever identified, or no Referent is constitutively available); the Referent Absence Property explains why one whole domain — evaluative self-content — sits in the last of those cells by its nature rather than by neglect.
Explanation
Some of what you believe about yourself can be checked against physical reality, and some cannot. "I can lift this" or "I am afraid of heights" has a reality that pushes back: you try, and the world corrects you. But most of what makes up a sense of self — "I am a generous person," "I am the sort of person who belongs here," "my way is the right way" — has no such external check, because it is evaluative, and there is no fact out in the world that settles it. Worse, even the parts that could in principle be checked are handled the same way as the parts that can't, so the mind cannot tell which is which, and a decisive correction may simply never arrive. The upshot is that the self-model runs mostly on material that reality never grades. This is not a flaw to be fixed by trying harder to be accurate; for most identity-content there is nothing to be accurate about. The sensible standard for such content is not "is it true?" but "is it stable, and does it do no harm?" — which is what subcriticality means. This is a different situation from holding a belief the world has already answered and ignoring the answer (that is Referent Abandonment): here there is no answer to ignore.
Epistemic status
Framework-Specific Named Result. Derived in Chapter 15 (The Bayesian Self-Model), §15.6.