Azarian’s Self-Modeling Strange Loop Theory

Cognitive neuroscientist Bobby Azarian’s Self-Modeling Strange Loop Theory holds that phenomenal consciousness arises when a system that models its world in order to resist entropy comes to include itself in that model as a causal agent. Building on Hofstadter’s Strange Loops, Azarian argues that this level-crossing, self-referential structure is what generates a first-person perspective, and that integration and differentiation of information, as measured in Integrated Information Theory, are necessary for experience but not sufficient without it. Once a subject exists, the qualitative character of each experience is identified with the position of the occupied state within a relational information geometry, and its valence with that state’s distance from the equilibrium the organism must maintain. He presents the result as a dissolution of the hard problem (Chalmers, 1995) and classifies it as multilevel monism with strong emergence, neither reductive materialism nor dualism (Azarian, 2027).

In other words, Azarian’s theory has phenomenal consciousness emerging when an integrated model of the world recursively represents the system doing the modeling, establishing a first-person perspective. Within that perspective, particular experiences are constituted by their positions in a structured space of possible states, while their affective significance reflects the organism’s requirements for continued existence. Combining Hofstadter’s self-reference, Tononi’s integrated information, and Friston’s free-energy and active inference, Azarian identifies experience with physically realized self-modeling organization, yet describes its emergence as ontologically novel and causally efficacious. His “multilevel monism” therefore treats consciousness as a real, irreducible level of natural organization, and not an illusion, a separate substance, or a property ubiquitous throughout matter (Azarian, 2027).

From Biological Persistence to Subjectivity

The theory begins from survival. Azarian situates consciousness within the problem of maintaining organized life far from thermodynamic equilibrium. Organisms persist through exchanges of energy and matter and through adaptive sensitivity to their environments. His broader account links increasingly capable models to increasingly flexible control: information becomes subjectively experienced when this modeling turns recursively inward (Azarian, 2027).

Azarian places this starting point alongside Friston’s Free-Energy Principle, Solms’s Affect as the Hidden Spring of Consciousness, and Seth’s “Beast Machine” Theory: a self-organizing system must model its world in order to remain in existence. He takes the move to be naturalistic and mechanistic without being reductive, because a world model is an emergent global structure, and explanations framed in terms of world models and simulations “do not and cannot reduce the system to only its low level components” (Azarian, personal communication, 2026).

Living systems persist by capturing free energy, which requires predicting an uncertain environment; Azarian adopts Friston’s free-energy formulation, supplying part of the proposed connection between persistence and cognition in which persistence is the minimization of surprise by a generative model, and the cybernetic law of requisite variety, under which a controller’s repertoire of responses must match the variety of disturbances it faces (Ashby, 1956). Competition then ratchets model complexity upward. This is the framework where Azarian treats life and mind as stages in a universe that organizes itself toward greater complexity (Azarian, 2022). For consciousness, the relevant consequence is that the phenomenal world is the working format of an organism’s own persistence.

Here, variational free energy is a mathematical bound on surprisal under a generative model, not simply usable thermodynamic energy (Friston, 2010). Azarian interprets conscious experience as the organism’s internally accessible means of navigating conditions relevant to survival. However, self-maintenance and adaptive regulation alone do not establish consciousness; his distinctive explanatory step is the emergence of recursive self-modeling.

Repertoire Entropy

Azarian introduces repertoire entropy: the Shannon entropy of the viable cognitive and behavioral states a system can occupy, excluding disorganized or nonfunctional states. Because an observation is informative in proportion to the alternatives it excludes, a larger repertoire means that each conscious state carries more information, and repertoire entropy is offered as a proxy for experiential richness. He relates it to the cause-effect repertoires of Integrated Information Theory and to Carhart-Harris’s Entropic Brain Hypothesis, and distinguishes it from phi: repertoire entropy measures differentiation, phi integration. He claims that a large set of noninteracting switches has the first without the second and is not conscious.

The Loop That Generates a Subject and a Self

Differentiation and integration together still leave, on this account, a display with no one viewing it. The decisive step is recursive: the generative model models itself modeling the world, representing the organism as the perceiving and acting agent at the center of its own simulation. The observer is therefore a model produced by the system, which Azarian takes to avoid the homunculus regress while preserving the reality of the inner theater. Nervous systems make this possible by recording the consequences of action in real time, so that the system accumulates a model of its own behavior. Further iterations, modeling oneself modeling oneself, yield meta-awareness, and consciousness is said to deepen in discrete jumps with each level, not by gradual accumulation.

Following Hofstadter (2007), Azarian locates the experiential subject in a feedback organization that crosses levels of representation. The world model incorporates the organism as perceiver and agent, including its own modeling activity. As noted, the resulting perspective is not inspected by an additional internal observer: the represented observer belongs to the same process that generates the represented world. This is intended to terminate the homuncular regress without eliminating subjectivity.

In correspondence, Azarian states the transition compactly. Integrated world modeling is not by itself sufficient for subjective experience; as the world model becomes sufficiently integrated and temporally deep, it must also incorporate the agent doing the modeling. When the model “recursively contains the modeler as an embodied center of perception and action, perspective appears.” There is then no longer merely a representation of a world but “a world represented for a subject” (Azarian, personal communication, 2026).

The self so generated is credited with causal power, a new layer of top-down cybernetic control that supports counterfactual simulation and, when sufficiently rich, what is called free will. Azarian cites the absence of consciousness in the largely feed-forward cerebellum, and the loss of consciousness when thalamocortical and frontoparietal recurrence is disrupted, as consistent with the requirement for recurrent architecture, and he draws on global workspace ignition as a description of the same dynamics.

Memory and counterfactual simulation enable the system to represent its actions, anticipate their consequences, and regulate itself accordingly. Azarian thus distinguishes information integration, which organizes content, from self-modeling, which, on his account, makes that content present to a subject (Azarian, 2027).

The crucial qualification is that neural recurrence, self-representation, and reflective self-awareness are different properties. Feedback does not automatically constitute a model of the modeler. The theory must specify which form of self-modeling is required for phenomenal consciousness, particularly where experience appears possible without explicit reflection.

Qualia as Occupied Positions

Azarian’s account of qualitative character is relational. A sensation is determined by its position within an organized space of discriminable alternatives: experienced red depends on its relations to other colors, rather than possessing an isolated intrinsic character. More generally, each state is defined by its contrasts with every state the system could occupy but does not. He calls this “dialectical information geometry,” emphasizing that unrealized alternatives help determine the significance of the state currently occupied (Azarian, 2027). He then links this “geometry” to the conceptual structures of Integrated Information Theory. A second description is laid over the first: following Friston, qualia function as precision-weighted signals indicating direction and distance from the organism’s attracting set, which is what gives a position its valence.

Azarian’s later summary assigns each component of the theory its own explanatory task. Information geometry accounts for the relational structure of the conscious field. Self-modeling accounts for the first-person, what-it-is-like perspective. And “something close to Solms’s account of affect” accounts for qualia as intrinsically felt, action-guiding signals that tell the organism how its current state relates to its existential needs and attractors—what to approach, avoid, preserve, or change. Qualia are, in his phrase, “automatic instructions for maintaining dynamic equilibrium against entropy’s tide,” and consciousness is “what it is like to be this recursively self-modeling control architecture” (Azarian, personal communication, 2026). The appeal to Solms adds to his deriving valence from Friston’s attracting set. The two are compatible, since Solms’s account is itself framed in free-energy terms, but the later statement places the weight on felt affect and not only on precision-weighted inference.

Three classical arguments are answered from this base. Zombies are impossible because any physical duplicate would perform the same integrated, recurrent self-modeling. An undetectable inverted spectrum is impossible because human color space is asymmetric, so no inversion preserves every similarity relation. Mary learns nothing from her first experience of color that falsifies physicalism (Jackson, 1982). The geometry was describable in advance, but a quale is a position occupied, and description does not occupy it. In other words, complete third-person knowledge describes the relevant structure; seeing red instantiates a position within it. The difference is between describing a possible experience and undergoing it, rather than necessarily discovering an additional nonphysical property.

Artificial Systems

Azarian holds that consciousness is in principle multiply realizable, yet follows Integrated Information Theory in denying it to large language models running on conventional hardware, which could simulate self-modeling without being a strange loop at the level of the physical substrate (Findlay et al., 2024). Candidate substrates are neuromorphic and self-organizing hardware, possibly requiring the electromagnetic integration proposed in McFadden’s Conscious Electromagnetic Information Theory, or cultured neural tissue.

Ontology

To the objection that the mechanism might operate without experience, Azarian replies that no further bridging fact exists: the structured self-relation is the experience, mechanism and phenomenology being one process described from outside and lived from within. He nonetheless calls the result strong emergence, an ontologically novel level that, he says, extends physics without violating it. Reality has mechanical, informational, and experiential levels, each real because it does causal work the level below cannot. Gödelian incompleteness is accepted, in response to Hoel (2024), but read as generative: understanding consciousness alters the system that understands.

Azarian has since addressed directly how identity and strong emergence can be held together. The proposal is “technically” an identity proposal, he writes, but not physicalism in the usual sense, because “the identity itself is emergent.” When the relevant architecture comes into existence, phenomenal subjectivity comes into existence with it, and in sufficiently self-reflective organisms that subject becomes a locus of control, exerting top-down causal influence on the system from which it emerged. Experience remains dependent on lower-level physical processes without being an illusion, an epiphenomenon, or “merely another description of what was already there.” Strong emergence is meant “in the sense of genuine ontological novelty and new causal powers, not something forever beyond scientific explanation”: it lies beyond the framework of earlier paradigms, not beyond science. Emergence brings new phenomena into existence, and science must then expand to include what had been considered magical or supernatural, which thereby becomes natural. For such phenomena, he proposes the less loaded term “meta-natural,” and he develops a corresponding onto-epistemology (Azarian, personal communication, 2026; 2027).

Emergence and Explanatory Status

Azarian calls consciousness strongly emergent because recursive organization establishes a novel subject and a new level of cybernetic control. Mental causation remains physically embodied, yet higher-level models guide processes that cannot be adequately characterized as purposive at the component level alone. A tension remains between explaining experience entirely through organization and asserting strong emergence: explanatory autonomy and top-down control do not by themselves demonstrate fundamentally new causal powers (Azarian, 2027). The notion of an emergent identity is Azarian’s answer to this tension. Whether an identity can itself be emergent, and how that differs from a psychophysical law that holds once the architecture exists, is left to the fuller treatment he promises.

The synthesis has substantial precedents, especially Safron’s integration of world modeling, IIT, global workspace theory, and active inference (Safron, 2020). Azarian’s emphasis falls on the conjunction of recursive subject formation, relational qualia, repertoire entropy, and regulatory valence. Azarian’s central claim—that appropriately organized self-relation constitutes phenomenality—offers a proposed dissolution of Chalmers’s explanatory gap, while leaving that identity and its discriminating empirical consequences to be established (Chalmers, 1995).

Multilevel Monism and Multilevelism

In correspondence, Azarian says that after reading the Landscape taxonomy (Kuhn, 2024), he extends his ontology (which he presents as provisional). He accepts that his view falls within non-reductive physicalism, and he also accepts the observation that non-reductive physicalism, once mental states are treated as ontologically distinct and causally efficacious, is hard to tell apart from property dualism. He suspects the point generalizes. Whenever a new level of organization emerges with new causal powers, he asks what additional fact about nature determines whether it is a new property of the one underlying reality or a new substance that has come into existence. Life is his first case: information acquires a new kind of causal efficacy, and an agent appears. Consciousness is the second: qualitative subjectivity appears, and eventually a subjective controller. In each, what emerges is wholly continuous with and dependent on what produced it, and is also new.

He therefore sets his multilevel monism—one underlying reality that repeatedly develops new emergent properties and causal regimes, in the manner of Ellis and Murphy—beside what he calls “multilevelism,” dualism generalized beyond two levels, in which new kinds of properties or substances emerge at successive levels. If the two descriptions specify exactly the same nested causal architecture, he doubts that any further ontological fact decides between them: “One description emphasizes continuity; the other emphasizes novelty. Nature may simply be both.” The distinction between a new causally efficacious property and a new substance may be “largely a language game” (Azarian, personal communication, 2026).

From this follows a conciliatory reading of the standing positions that each captures a partial truth. Reductionism sees continuity but becomes inadequate if it denies the causal reality of emergent levels. Dualism recognizes that consciousness introduces ontological novelty, though the novelty need not be a separate substance inserted into nature from outside. Non-reductive physicalism recognizes dependence without reduction; dual-aspect monism, the unity underlying different modes of reality; emergentism, the appearance of new causal regimes. They are not all simply right, Azarian says, but each may isolate a real feature of a nested architecture richer than any one category allows, in which nature changes regimes through transitions that are distinct yet fuzzy (Azarian, personal communication, 2026). The claim is offered as a suspicion, not as a result.

Distinguishing the Self

The theory departs from its source in holding the self to be real and causally potent, where Hofstadter treats the “I” as a mirage, and in requiring a physical recurrent substrate. Its use of Integrated Information Theory is revisionary, since that theory identifies experience with integrated cause-effect structure and has no separate condition of self-modeling. The claims of identity and of strong emergence are asserted together, a conjunction Azarian defends as emergent identity, and substrate-independence sits beside substrate requirements. Repertoire entropy and qualia geometry are not yet formalized; Azarian describes rigorous formulation and experimental demonstration as the remaining task.

Taxonomic Placement

Azarian’s Self-Modeling Strange Loop Theory is not classified with Hofstadter’s Strange Loops in Materialism/Relational for ontological reasons: While Azarian explicitly maintains one physical reality, he affirms an irreducible experiential level with genuine top-down causal efficacy and emergent mental causation. Moreover, Azarian treats “materialism” as a rival position that dodges the hard problem, and calls his own view multilevel monism with strong emergence.

By ontology, the best fit is Non-Reductive Physicalism, beside Ellis’s Strong Emergence and Top-Down Causation. Azarian concurs: he places the theory in “the nonreductive physicalist camp” and names Ellis and Murphy as its nearest kin, while holding that the distinction between that camp and property dualism may be “actually something of a language game" (Azarian, personal communication, 2026)—for a comparison, see Landscape’s Non-Reductive Physicalism-Overview and Property Dualism.

Cross-references should connect Azarian’s Self-Modeling Strange Loop Theory to Hofstadter’s Strange Loops, Safron’s Integrated World Modeling Theory, Friston’s Free-Energy Principle and Active Inference, Solms’s Affect as the Hidden Spring of Consciousness, Seth’s “Beast Machine” Theory, Ellis’s Strong Emergence and Top-Down Causation, and Murphy’s Non-Reductive Physicalism.

References

Ashby, W. R. (1956). An Introduction to Cybernetics. London: Chapman & Hall.

Azarian, B. (2022). The Romance of Reality: How the Universe Organizes Itself to Create Life, Consciousness, and Cosmic Complexity. Dallas: BenBella Books.

Azarian, B. (2026). Personal communication to R. L. Kuhn.

Azarian, B. (2027). Our Cosmic Purpose: A Practical Guide to Waking Up the Universe. Dallas: BenBella Books. [2027; chapter 5, “Strange Loop Theory,” consulted in uncorrected advance excerpt.]

Chalmers, D. J. (1995). Facing up to the problem of consciousness. Journal of Consciousness Studies, 2(3), 200–219.

Findlay, G., et al. (2024). Dissociating artificial intelligence from artificial consciousness. arXiv preprint.

Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138.

Hoel, E. (2024). Consciousness as a Gödel sentence in the language of science. The Intrinsic Perspective. https://www.theintrinsicperspective.com/p/consciousness-as-a-godel-sentence

Hofstadter, D. R. (2007). I Am a Strange Loop. New York: Basic Books.

Jackson, F. (1982). Epiphenomenal qualia. The Philosophical Quarterly, 32(127), 127–136.

Kuhn, R. L. (2024). A landscape of consciousness: Toward a taxonomy of explanations and implications. Progress in Biophysics and Molecular Biology, 190, 28–169.

Safron, A. (2020). An Integrated World Modeling Theory (IWMT) of consciousness: Combining integrated information and global neuronal workspace theories with the free energy principle and active inference framework; toward solving the hard problem and characterizing agentic causation. Frontiers in Artificial Intelligence, 3, 30.