nucleation

THEORY-883: Under active inference, the contents of consciousness are posterior beliefs about the hidden states of world, body or brain, and a posterior is conscious when it has the capacity to inform discrete, counterfactual policy selection, at a timescale slow enough to abstract from sensory flux and with enough precision to tell alternatives apart; on this account no conscious content changes without a change in the posterior, and a stimulus given precision by an attention or saccade policy should reach awareness sooner than a retinally matched one

4 min read

Properties4
StatusProposed
Tagsconsciousness, neuroscience, probabilistic-modeling
SourceLIT-1364 — On the minimal theory of consciousness implicit in active inference
Bears onQUESTION-254 — What separates an interpretive organization that constitutes an experiential perspective from one that only processes information, or only represents a perspective, when the separation must not require an episode's current contents to be veridically constrained by their referents, must not be drawn from the verdict it gives on groups and machines, and must justify the queries, interventions and timescales by which it individuates a perspective without appeal to the phenomenal unity it is meant to explain?

Graph View

Source

Whyte et al. (2026), LIT-1364, read in full in NOTE-1062 from arXiv:2410.06633v2: section 5 (pp. 24–31), with the models of section 4 and Appendices 1–2. Page numbers are the manuscript’s. The modelling studies it reviews were not read.

What was actually shown

The paper is a review with a theoretical conclusion, and three small simulations of its own.

  • The core, by inspection of the models. Every reviewed model minimises variational and expected free energy, so “all behaviour, both conscious and unconscious, must, in some minimal sense, be aimed at minimising these quantities” (p. 25). Each model maps conscious content to the posterior over states: “The most minimal such link implicit in all the models reviewed above is between the contents of consciousness and the inferred hidden state” (p. 25). Each model reads out awareness where a discrete level drives report.
  • The criterion. “the conscious/unconscious distinction is tied to the mechanism(s) of conscious access, and that it occurs at the discrete-continuous interface” (p. 26). It is dispositional: “it is the capacity (in a counterfactual sense) for such a posterior belief to inform policy selection processes at the right timescale and with sufficient precision that makes it conscious” (p. 27). The authors hold it “a necessary and sufficient condition for content to become conscious in humans (and non-human animals with similar neuronal architectures)” (p. 27).
  • What could have come out otherwise. The reviewed no-report model lost its prefrontal and P3b-like responses without report while the second-level posterior stayed reportable (p. 18); it might have lost both. The paper’s re-run rivalry model fits Levelt’s second and fourth propositions (Appendix 1); it might not have. Two predictions are named as close to the core (pp. 29–30): reward should trade off against epistemic value in rivalry (Appendix 2), and “To See is to Look”, so a precision-weighted stimulus “will enter the contents of consciousness sooner than a retinally-matched stimulus that is not the target of a saccade” (p. 30). Neither is tested in the paper.

What this does not say

  • It is not a criterion for experience in general. Its condition is the mechanism of access, read off models that had to report. The authors “remain agnostic as to whether this condition is both necessary and sufficient in systems vastly different to ourselves” (p. 27), and say the link between content and posterior “is consistent with non-conscious creatures exhibiting behaviour governed by active inference” (p. 26). It does not meet CLAIM-411’s defeater, which asks for a condition independent of report.
  • The third part is not derived. The first two parts are “in some sense logically entailed” by the framework; the third is “an inference to the best explanation” (p. 27).
  • The content link is assumed in every model, not tested. “Assuming a correspondence between the posterior over states and the content of awareness” (p. 14) is the step each model of content takes.
  • The empirical confirmations are second-hand. One, of the no-report model’s first prediction, is cited without a reference in v2 (p. 18).
  • It is not identity. The paper speaks of correspondence and link. It takes no side between constitution and production (CLAIM-412).
  • The precision prediction is not yet a test of experience. A stimulus that reaches report sooner is a fact about access. Whether that is a fact about experience is QUESTION-048’s question, and THEORY-094’s point that post-perceptual cognition survives without report applies.

Connections

  • QUESTION-254. Answers it in part, on the organizational side only: a specified condition (discrete format, timescale, precision, capacity to inform policy) that separates conscious from unconscious posteriors in humans. It does not address the phenomenal side.
  • CLAIM-411. The paper grants its premise in so many words, and its condition is the kind of restriction that claim asks for, but tied to report.
  • THEORY-882 (Laukkonen et al.) calls this view complementary and adds a coherence requirement and a recursive one. The two are compatible.
  • THEORY-094, THEORY-577. The no-report model is a mechanism for THEORY-094’s persistence of post-perceptual processing. The criterion builds report into the mechanism, so it answers THEORY-577’s demand by construction and gains nothing on experience from doing so.
  • LIT-1191 (Sandved-Smith et al.), read in NOTE-958, is one of the reviewed models, and the paper calls it conceptual analysis rather than a fit to data.

Backlinks