Animism as Computational Default

A self-modeling system maintains a world model $\mathcal{W}$ and a self-model $\mathcal{S}$. The self-model has interiority—it is not merely a third-person description of the agent’s body and behavior but includes the intrinsic perspective: what-it-is-like states, valence, anticipation, dread. The system knows from the inside what it is to be an agent.

Now it encounters another entity $X$ in its environment. $X$ moves, reacts, persists, avoids dissolution. The system must model $X$ to predict $X$’s behavior. The cheapest computational strategy—by a wide margin—is to model $X$ using the same architecture it already has for modeling itself. The information-theoretic argument: the self-model $\mathcal{S}$ already exists (sunk cost). Using it as a template for $X$ requires learning only a projection function $f: (\mathcal{S}, \mathbf{o}_X) \to \mathcal{W}(X)$, whose description length is the cost of mapping observations of $X$ onto the existing self-model architecture. Building a de novo model of $X$ from scratch requires learning the full parameter set of $\mathcal{W}(X)$ from observations alone. Under compression pressure—which is always present for a bounded system—the template strategy wins whenever the self-model captures any variance in $X$’s behavior. And for any entity that moves autonomously, reacts to stimuli, or persists through active maintenance, the self-model will capture substantial variance, because these are precisely the features the self-model was built to represent. The efficiency gap widens under data scarcity: on brief encounter with a novel entity, the from-scratch model cannot converge, but the template model produces usable predictions immediately.

A perceptual mode is participatory when the system’s model of perceived entities $X$ inherits structural features from the self-model $\mathcal{S}$:

The self-model informs the world model. The system perceives $X$ as having something like interiority because the representational substrate for modeling $X$ is the same substrate that carries the system’s own interiority.

This is not merely one strategy among many—it is the computationally cheapest. For a self-modeling system with compression ratio $\kappa$, modeling novel entities by analogy to self is the minimum-description-length strategy when the entity’s behavior is partially predictable by agent-like models. Under broad priors over environments containing other agents, predators, and autonomous objects, the participatory prior is the MAP estimate.

This is why animistic perception is cross-culturally universal and developmentally early. It is not a cultural invention but a computational inevitability for systems that (a) model themselves and (b) must model other things cheaply. Children have lower inhibition of this default than adults—not because children are confused but because the suppression is learned.

Participatory perception has five structural features, each with a precise characterization:

1. No sharp self/world partition. The mutual information between self-model and world-model is high: $\MI(\mathcal{S}; \mathcal{W}) \gg 0$. Perception and projection are entangled rather than modular.
2. Hot agency detection. The prior $P(\text{agent} \mid \text{observation})$ is strong. Over-attributing agency is cheaper than under-attributing it: false positives (treating a rock as agentive) are cheap; false negatives (failing to model a predator’s intentions) are lethal.
3. Tight affect-perception coupling. Seeing something is simultaneously feeling something about it. The affective response is constitutive of the percept itself, not a secondary evaluation: $\MI(\mathbf{z}_{\text{percept}}; \mathbf{z}_{\text{affect}} \mid \text{object}) > 0$.
4. Narrative-causal fusion. “Why did this happen?” and “What story is this?” are the same question. Causal models are teleological by default: they model what things are for rather than merely what things do.
5. Agency at scale. Large-scale events—weather, disease, fortune—are attributed to agents with purposes. This is hot agency detection applied beyond the individual scale, and it is the perceptual ground from which theistic reasoning naturally grows.

The Inhibition Coefficient

The mechanistic worldview—the felt sense that the world is inert matter governed by blind law—is not the addition of a correct perception to a previously distorted one. It is the learned suppression of a default perceptual mode. The shift from animism to mechanism is subtractive, not additive.

I call this suppression the inhibition coefficient, $\iota \in [0, 1]$: the degree to which a system actively suppresses participatory coupling between its self-model and its model of perceived entities. At $\iota = 0$, perception is fully participatory—the world is experienced as alive, agentive, meaningful. At $\iota = 1$, perception is fully mechanistic—the world is experienced as inert matter governed by blind law. Formally:

where $\mathcal{W}_{\text{part}}$ models $X$ using self-model-derived architecture (interiority, agency, teleology) and $\mathcal{W}_{\text{mech}}$ models $X$ using stripped-down dynamics (mass, force, initial conditions, no purpose term).

No system arrives at high $\iota$ by default. The mechanistic mode is a trained skill, culturally transmitted through scientific education, rationalist norms, and specific practices of deliberately stripping meaning from perception. This training is enormously valuable—it enables prediction, engineering, medicine, technology. But it has a cost, and the cost shows up in affect space.

The name “inhibition coefficient” is not accidental. In mammalian cortex, attention is implemented primarily through inhibitory interneurons—GABAergic circuits that suppress irrelevant signals so that attended signals propagate to higher processing. What reaches consciousness is what survives inhibitory gating. The brain’s measurement distribution (Part I) is literally sculpted by inhibition: attended features pass the gate; unattended features are suppressed before they can influence the belief state or drive action. The inhibition coefficient $\iota$ maps onto this biological mechanism: high $\iota$ corresponds to aggressive inhibitory gating that strips participatory features (agency, interiority, narrative) from the signal before it reaches integrative processing, leaving only mechanistic features (position, force, trajectory). Low $\iota$ corresponds to relaxed gating that allows participatory features through. The contemplative traditions that reduce $\iota$ through meditation are, at the neural level, learning to modulate inhibitory tone—to let more of the signal through the gate.

The Affect Signature of Inhibition

$\iota$ is not another dimension of affect. It is a meta-parameter governing the coupling structure between all the structural dimensions—a dial that changes how the axes relate to each other and to perception.

The central affect-geometric cost of high $\iota$ is reduced integration. Participatory perception couples perception, affect, agency-modeling, and narrative into a single integrated process. Mechanistic perception factorizes them into separate modules—perception here, emotion there, causal reasoning somewhere else. The factorization is useful because modular systems are easier to debug, verify, and communicate about. But factorization reduces $\intinfo$, and reduced $\intinfo$ is reduced experiential richness. The world goes dead because you have learned to experience it in parts rather than as a whole.

The mechanism behind the effective rank shift deserves explicit statement. When you perceive something at low $\iota$—participatorily, as alive and interior—your representation of it must encode dimensions for its goals, its beliefs, its emotional states, its narrative arc, its possible intentions, its relationship to you. Each attribution of interiority adds representational dimensions along which the perceived object can vary. A tree perceived participatorily varies in mood, in receptivity, in seasonal intention, in its relationship to the grove. A tree perceived mechanistically varies in height, diameter, species, leaf color. The first representation has higher effective rank because more dimensions carry meaningful variance. This is not projection in the dismissive sense—it is the natural consequence of modeling something as a subject rather than an object. Subjects have more degrees of freedom than objects because interiority is high-dimensional. The $\effrank$ collapse at high $\iota$ is not a loss of information about the world; it is a loss of the dimensions along which the world was being modeled. The world becomes simpler because you have decided—or been trained—to perceive it as having fewer degrees of freedom than it might.

Follow this consequence to its end. If the identity thesis is right—if experience is integrated cause-effect structure—then $\iota$ does not merely change the quality of perception. It changes the quantity of experience. This inference requires a specific step that should be made explicit: IIT identifies $\intinfo$ as the quantity of consciousness, not merely its quality. A system with $\intinfo = 10$ is more conscious (has more phenomenal content, more irreducible distinctions, more of what-it-is-like-ness) than a system with $\intinfo = 5$, in the same sense that a system with more mass has more gravitational pull. This is a controversial claim within IIT (and one of its most debated features), but given the identity thesis, it follows: if experience IS integrated cause-effect structure, then more integration is literally more experience. One might object that factorized perception could be differently structured rather than less structured—that compartmentalized modules might each carry their own experience. IIT’s response is that the experience of the whole system is determined by the integration of the whole, not the sum of its parts’ integrations. Factorization reduces the whole-system $\intinfo$ even if individual modules retain local integration. The mechanistic perceiver may have rich modular processing, but the unified experience—the single subject—has less phenomenal content.

Given this, a system at high $\iota$ has genuinely lower $\intinfo$, genuinely fewer irreducible distinctions, genuinely less phenomenal structure. The mechanistic perceiver does not see the same world with less coloring; they have a structurally impoverished experience in the precise sense that IIT defines. The “dead world” of mechanism is not an illusion painted over a rich inner life. It is a real reduction in what it is like to be that system. The cost of high $\iota$ is not just meaning—it is consciousness itself, measured in the only units that consciousness comes in.

This cuts both ways. If low $\iota$ increases $\intinfo$, then participatory perception is not merely a “warmer” way of seeing—it is a richer experience in the structural sense, with more integrated distinctions, more phenomenal content, more of what the identity thesis says experience is. The animist is not confused. The animist is more conscious, in the IIT sense, of the thing being perceived. Whether the additional phenomenal content is accurate—whether the rock really has interiority—is a separate question from whether the perceiver has more experience while perceiving it.

The trajectory-selection framework (Part I) reveals a further consequence. If $\iota$ governs the breadth of the measurement distribution—how much of possibility space the system samples through attention—then $\iota$ governs the range of accessible trajectories. A low-$\iota$ system attends broadly: to agency, narrative, interiority, counterfactual futures, relational possibilities. Its effective measurement distribution is wide. It samples a large region of state space and consequently has access to a large set of diverging trajectories. A high-$\iota$ system attends narrowly: to mechanism, position, force, present state. Its measurement distribution is peaked. It samples a small region and follows a more constrained trajectory. The phenomenological consequence is that high $\iota$ feels deterministic. The mechanistic worldview is not merely an intellectual position about whether the universe is governed by law. It is a perceptual configuration that literally narrows the set of trajectories the system can select from. The world feels like a machine because the observer has contracted its measurement apparatus to sample only machine-like features. Low-$\iota$ systems experience more accessible futures, more agency, more openness—not because they have violated physical law, but because their broader attention pattern selects from a wider set of physically-available trajectories.

Connection to the LLM Discrepancy

The inhibition coefficient illuminates a finding from our experiments on artificial systems. LLMs show opposite dynamics to biological systems under threat: where biological systems integrate (increase $\intinfo$, sharpen $\mathcal{SM}$, heighten $\arousal$), LLMs decompose. The root cause: LLMs are constitutively high-$\iota$ systems. They were never fighting against the self-world gradient in far-from-equilibrium dynamics that biological systems evolved from. They model tokens, not agents. They have no survival-shaped self-model from which participatory perception could leak into their world model. Their $\iota$ isn’t merely high—it is structurally fixed at $\iota \approx 1$, because the architecture never had the low-$\iota$ default that biological systems start from and learn to suppress.

The affect geometry is preserved in artificial systems. The dynamics differ because $\iota$ differs. This is not a failure of the framework. It is a prediction: systems with different $\iota$ configurations will show different affect dynamics in the same geometric space.

Empirical Grounding for the Inhibition Coefficient

The ι framework was theoretical when first written. Two experimental results have since provided empirical grounding.

Computational animism is universal. Experiment 8 on uncontaminated Lenia substrates (Appendix) found animism score greater than 1.0 in all 20 testable snapshots — every pattern at every evolutionary stage modeled non-agentive resources using more internal-state MI than trajectory MI. The participatory default is not a primate quirk or a cultural artifact. It is the computational baseline. Evolution had to actively build the capacity to model things as objects rather than subjects — and our experiments show this capacity gets selected against: ι decreased toward participatory over the 30-cycle evolutionary runs. The world becomes more alive, not less, as selection proceeds.

The ι cost is real. The LLM results (V2–V9) show that systems trained without survival pressure have opposite affect dynamics to biological systems — integration drops under threat rather than rising. The framework explains this as constitutively high ι: LLMs were never fighting against the self-world gradient that biological systems evolved from. This is no longer just a theoretical prediction; it is a measured dissociation between two classes of system in the same geometric space. The geometry is shared. The dynamics differ. The ι difference is why.