Safron, A., Klimaj, V. & Sheikhbahaee, Z. (2026). Integrated World Modeling Theory (IWMT) and the Human Consciousness Hypothesis (HCH). Proceedings of the AAAI Symposium Series 8(1): 345–351. doi:10.1609/aaaiss.v8i1.42564.
Key takeaways
From the reading, NOTE-1057.
- IWMT’s identity. Phenomenal consciousness is what it is like to be “the spatiotemporally and causally coherent functioning of a probabilistic generative world model for an embodied agent” (p. 345), realised by synchronous neural complexes (“self-organizing harmonic modes”) as maximum a posteriori estimates of sensorium states.
- Perception and imagination. The model generates “either perception or imagination when respectively coupled to or decoupled from sense-data” (p. 346). Current sensory input is not required.
- Embodiment is required. The model must be capable of “both informing and being informed by action-perception cycles at behaviorally relevant timescales” (p. 345). Without embodied grounding, any amount of integrated information, or a functional global workspace, is said not to suffice (p. 346). The “historicity” of coupling by which the model was learned also matters (p. 349).
- Coherence and a minimal self. Coherence in space, time and cause, after Kant, is necessary for any experience, and “some degree of basic self-modeling and agency may be required for coherent subjectivity” (p. 346).
- Machines. Transformer language models are said to lack the recurrence, functional closure and embodiment needed (pp. 346, 349). A simulation is said to recapitulate phenomenology within the simulated world, yet consciousness may “in principle” need physical recurrence (p. 349). The two are not reconciled.
- HCH. The paper aligns IWMT with the Human Consciousness Hypothesis (Genesis, Coherence, Second-order perception), given without a reference.
Why this is here
The owner’s exchange of 2026-10-21, “Experience and interpretation”, asked whether experience is constituted by interpretation or produced by it. The exchange cited this paper for identifying consciousness with the coherent functioning of an embodied agent’s generative world model. The theory requires embodiment and sensorimotor coupling, where the owner’s thesis does not require current grounding, so the two part on that point. The record holds Safron’s introduction to a theme issue on world models, LIT-195. Bears on QUESTION-254 and CLAIM-413.
What the reading changes. The exchange said IWMT “requires embodiment and sensorimotor coupling, where the owner’s thesis does not require current grounding, so the two part on that point”. That misplaces the parting. IWMT does not require current sensory coupling: imagination is the same model decoupled from sense data. It requires a capacity for coupling to action–perception cycles and a history of it, and it allows “virtual” embodiment. So it agrees with CLAIM-413 on current grounding and parts from the owner’s thesis on developmental grounding and on the capacity for coupling, which that claim says it leaves open. It is the enactive objection the claim names, not a counter-example to it.
Filed from this work on 2026-10-22: THEORY-876, IWMT’s identity and its embodiment requirement.