Conscious experience is encountered as an intrinsic fact: there is something it is like for a subject, and the unity and differentiation of that experience are not first given as measurements from outside. Physical science, by contrast, earns objectivity through descriptions that remain stable under admissible changes of representation. The tension between these standpoints is often formulated as an explanatory gap between phenomenal existence and physical structure. Integrated Information Theory (IIT) is important because it attempts an experience-first reconstruction of physical requirements, while ontic structural realism and related forms of structural monism challenge the assumption that the physical must ultimately be described as a collection of self-subsistent substances. Yet neither move by itself yields an identity between phenomenal existence and physical organization. The central proposal developed here is therefore narrower and more demanding: if a physical structure is to be a serious candidate for the identity of a subject, it should not be a state vector, an entanglement value, a graph drawn in one basis, or a single scalar such as a surrogate Φ. It should be an equivalence class of causally irreducible processes under transformations that preserve the system’s compositional ontology, and it should remain identifiable across a nonzero interval of time and across physically justified changes of grain.
This proposal is called invariant causal process structure (ICPS). A process is represented by a quadruple 𝔖 = (𝒜, ω, {ℰt:s}, Γ): an algebra of observables, a state, a family of dynamical maps, and an explicit compositional structure Γ that determines which partitions count as decompositions of the same physical system. Γ is indispensable: without a declared factorization, claims about integration are formally underdetermined. The intrinsic candidate is not 𝔖 in one notation but its equivalence class [𝔖]GΓ under admissible automorphisms that preserve Γ. Classical stochastic kernels and finite-dimensional quantum channels are treated as two realizations of this schema. Irreducibility is measured by explicit cut interventions; closure is defined from internal effective information relative to cross-boundary contextual sensitivity; temporal persistence is measured from invariant cut fingerprints across a time window; and grain is handled by induced macro-dynamics rather than by averaging coupling magnitudes. These quantities are deliberately not collapsed into one consciousness score. Candidate structures are compared by Pareto dominance, preserving the independence of distinct structural constraints.
The framework is positioned relative to recent work rather than insulated from it. IIT 4.0 formalizes intrinsic cause-effect power and Φ-structures [1], while Grasso, Hendren, and Tononi have recently developed a direct program connecting phenomenal content to intrinsic causal structure [3]. Mindt distinguishes intrinsic from extrinsic structure and dynamics [4], Nešić develops a neutral-structuralist approach to consciousness and selfhood [5], and Carvalho, Langa, and Soler-Toscano formalize intrinsic information through dynamical systems and information landscapes [20]. ICPS does not claim priority for the generic idea that consciousness is structural. Its distinct commitment is that a subject-candidate must be defined at the level of a temporally extended process equivalence class, with an explicit representation group, an explicit compositional structure, state/channel separation, and a non-scalar multi-criterion test of irreducibility, closure, persistence, and grain. Quantum mechanics is used as a stress test for nonseparability and representation, not as a speculative neural mechanism. The calculations establish properties of these structural diagnostics and expose counterexamples to common identifications. They do not prove that any calculated structure is conscious.

Orientation: existence, structure, and the level at which the question is asked

A theory of consciousness that begins from structure faces a double demand. It must be abstract enough to compare classical stochastic dynamics, quantum dynamics, and different physical grains without confusing notation with ontology. At the same time it must be concrete enough that its claims can fail. A merely verbal structuralism is too permissive: almost anything can be redescribed as a graph, a set of relations, or an information-processing system. The relevant notion of structure must therefore include modality, intervention, composition, and invariance. It must say which differences in a description correspond to physical differences, which transformations preserve the same organization, which partitions count as decompositions, and how a purported whole behaves when those relations are severed.
The basic object is accordingly not a static graph. A static graph suppresses state-dependence and temporal direction; a density operator suppresses the distinction between what a system is and what it can do; a transition matrix suppresses the question of which variables are physically legitimate units; an entanglement measure suppresses causal direction entirely. The more adequate object is a structured process. In finite settings it can be represented by an algebra of observables or degrees of freedom, a state on that algebra, a dynamical map, and an explicit compositional rule. The proposed ontology begins only after these ingredients are separated.
This separation also changes the role of quantum theory. Noncommutativity and entanglement are not treated as shortcuts to consciousness. They are useful because they make hidden assumptions visible. In a classical finite algebra, observables commute and a state can be represented as a probability distribution over a joint sample space. In a quantum algebra, there need not be a single joint probability distribution assigning simultaneous values to all observables. If a proposed account of intrinsic causal organization silently presupposes classical separability or a preferred basis, quantum theory exposes that presupposition. The question becomes what aspects of causal unity survive when the underlying algebra itself changes.
The central hypothesis is deliberately conditional. If phenomenal subjectivity is identical with a physical organization, then the relevant organization should not be destroyed by arbitrary relabeling, should not depend on an observer-selected coordinate system, should exhibit more than static correlation, should possess a nontrivial causal boundary, and should remain identifiable over a temporally extended process. These conditions do not derive phenomenal existence from equations. They make a structural identity proposal mathematically answerable enough to be criticized.
​

1. Phenomenology and the Problem of Existence

1.1 Experience is not first encountered as an object

The scientific problem of consciousness begins from an asymmetry. A star, a neuron, or a detector trace is encountered as an object available to multiple observers. Experience is encountered differently: its existence is given in the occurrence of the experience itself. This does not imply that experience is nonphysical, private in a metaphysically mysterious sense, or immune to scientific explanation. It does mean that a theory which begins exclusively from third-person observables risks changing the explanandum before it begins. Chalmers’ hard problem is one formulation of this concern: structural and dynamical truths, described externally, do not appear by conceptual entailment alone to yield truths about what experience is like [11]. The point is not that a physical identity is impossible, but that identity cannot be obtained merely by adding another correlation between a neural variable and a report.
IIT is unusual among scientific theories because it begins by characterizing properties that experience appears to possess from within and then asks what a physical substrate would have to be like if those properties were literally instantiated [1,2]. Existence, composition, information, integration, and exclusion are not treated merely as convenient descriptors. They constrain what counts as the physical substrate. That methodological reversal is valuable even if one rejects particular IIT postulates, particular measures, or IIT’s metaphysics. The 2025 adversarial collaboration between IIT and global neuronal workspace theory is especially important here: several predicted neural signatures received mixed support, and key predictions of both theories were challenged [13]. A metaphysical proposal inspired by IIT should therefore avoid making its validity depend on the empirical success of every current IIT neural prediction.

1.2 Four meanings of intrinsic

The word intrinsic is dangerously overloaded. Four senses must be separated. Epistemic intrinsicness concerns what can be specified without an external observer’s chosen coding. Causal intrinsicness concerns differences that the system makes to its own possible states rather than information available only to an external decoder. Representational intrinsicness concerns invariance under transformations that change notation or coordinates without changing the modeled physical organization. Ontological intrinsicness concerns what the system is in itself. The first three can be given increasingly precise operational or mathematical criteria. The fourth is a metaphysical claim. A central discipline of ICPS is that success with the first three is evidence relevant to, but not a proof of, the fourth. Mindt’s argument that not all structure and dynamics are equivalent is directly relevant [4], but the distinction must be sharpened into transformation rules rather than left at the level of an adjective.

1.3 The subject/object distinction as a constraint

A subject is not simply a complicated object described from outside. If subjectivity is physically instantiated, the candidate substrate must possess an organization whose unity is not imposed solely by an observer’s decision to draw a boundary around it. This motivates three formal demands. First, the candidate should be irreducible relative to physically meaningful partitions. Second, it should possess a degree of causal closure, so that its internal evolution is not merely a passive projection of a larger external driver. Third, it should persist as an identifiable process for a nonzero temporal interval. An instantaneous maximum of a scalar can fail the third demand even if it satisfies the first two.
Phenomenological constraint → not “find a large number,” but “identify a process whose unity, boundary, and temporal identity are invariantly specified.”

1.4 Identity is a bridge principle, not a computational output

No calculation below is licensed to infer phenomenal existence from complexity, information, entanglement, mutual information, irreducibility, or causal closure alone. The strongest metaphysical hypothesis considered is an identity thesis: that phenomenal subjectivity and a certain intrinsic process structure are two modes of access to one concrete occurrence. Identity hypotheses can be coherent and scientifically constrained without being derivable from lower-level equations by definition. The explanatory burden is therefore split. Mathematics can test whether a proposed physical structure has the right invariance, unity, boundary, and temporal behavior. Philosophy must still defend why phenomenal existence should be identified with that structure rather than merely correlated with it.

1.5 Phenomenology supplies constraints, not measurements

Phenomenology does not deliver a transition matrix. It does, however, place constraints on what a physical identity claim must explain. An experience is given as occurring, as this rather than every alternative, as internally differentiated, and as a single episode rather than a mere list of independent fragments. IIT turns closely related claims into axioms of existence, intrinsicality, information, integration, exclusion, and composition [1]. Whether those axioms are complete is contestable, but their methodological role is important: the target is not first a behavioral function and only later a metaphysical interpretation. The physical description is required to exhibit properties that answer to the structure of the explanandum.
A danger appears immediately. A phenomenological property can be translated too quickly into a mathematical surrogate. Unity can become graph connectivity, differentiation can become entropy, intrinsicality can become mutual information, and definiteness can become whichever threshold gives a convenient boundary. None of these identifications is justified merely by linguistic resemblance. The correct direction of reasoning is constraint, construction, and stress test. Phenomenological unity motivates irreducibility tests; it does not announce in advance which divergence, which partition rule, or which state space is uniquely correct.

1.6 Explanatory identity versus causal production

The familiar hard-problem schema can be written as a production question: given a physical description P, why or how does P produce an additional phenomenal fact E? If P and E are assumed to be ontologically independent relata, the arrow P → E asks for a bridge that ordinary causal explanation seems unable to supply. An identity theory rejects the extra production step. The relevant claim is instead that one concrete occurrence admits an extrinsic physical characterization and an intrinsic phenomenal characterization.
Pₑₓₜ ≅ Eᵢₙₜ
The symbol is intentionally an isomorphism-like sign rather than an equality sign. What would have to be established is not numerical equality between a physical variable and a quale. It is a defensible correspondence between the organization specified in an external causal model and the organization instantiated from the system's own causal perspective. IIT's explanatory identity is one developed version of this strategy [1,25]. The present framework remains weaker: it studies necessary structural conditions for such an identity while leaving open whether those conditions are sufficient.
This reframing relocates the explanatory burden. It no longer asks for a causal mechanism that manufactures phenomenality out of something wholly nonphenomenal. It asks why this physically instantiated structure should possess an intrinsic mode of existence at all, and why the intrinsic character should be phenomenal rather than merely another structural fact. That residual question is not a defect to be hidden. It is the precise point at which formal results end and metaphysical interpretation begins.

1.7 The subject is a persistence problem as well as a boundary problem

An instantaneous system can satisfy a cut-based irreducibility test and still be a poor model of a subject. If the minimum partition, candidate boundary, or structural signature changes discontinuously under arbitrarily small temporal evolution, then identifying the subject with the instantaneous maximum makes personal or experiential unity depend on the phase of a trajectory. The subject problem therefore contains a temporal identity problem. A candidate must be more than irreducible at time t; some invariant organization must remain trackable over an interval.
This motivates treating the subject as a process-equivalence class rather than a point in state space. The relevant persistence need not mean stasis. A melody persists as a pattern precisely by changing. A vortex persists while its constituent molecules are exchanged. Likewise, a causal subject could persist through state transitions if a suitable relational signature remains stable under those transitions. This will later motivate cut fingerprints, process tensors, and grain-persistence tests rather than a search for one privileged instantaneous scalar.

2. From Substances to Structures

2.1 The structural turn

Classical metaphysical pictures often begin with substances that possess properties and then stand in relations. Modern physics makes that order less secure. Quantum theory, gauge symmetry, permutation invariance, and spacetime physics repeatedly make the identity of physical entities depend on relational and symmetry structure. Ladyman’s distinction between epistemic and ontic structural realism was an attempt to take that lesson metaphysically seriously [7]. French develops a modal version in which laws and symmetries help determine the physical content of structure [8]. None of this establishes that objects do not exist. It does undermine the assumption that objecthood and intrinsic qualitative properties must be primitive in every adequate physical ontology.
Every Thing Must Go presses the methodological point more aggressively: metaphysics should be continuous with successful science rather than built from intuitions about middle-sized objects [6]. For consciousness, the lesson is not that “relations are conscious.” It is that a proposed ontology should not demand primitive substances if the best physical descriptions do not. A structural approach can therefore ask whether the subject is constituted by a physically realized organization of modal and causal relations. The word realized matters. Abstract isomorphism is not enough. A graph on paper, a software simulation, and a physical process may instantiate the same extensional pattern while differing in counterfactual causal organization.

2.2 Russell, Newman, and the danger of empty structure

Russell’s The Analysis of Matter argued that physics characterizes much of the world through mathematical and relational structure while leaving open questions about intrinsic nature [10]. Newman’s objection remains decisive against any naive structuralism: if “structure” means only an extensional pattern over a domain with the right cardinality, structural claims can become nearly trivial [9]. A consciousness theory that says only that experience is “some relation structure” therefore explains almost nothing. ICPS responds by requiring modal structure: transition rules, interventions, counterfactual dependence, temporal composition, and invariance constraints. The relevant structure is not merely which nodes are connected but what transformations the concrete system supports and what would change under physically defined cuts.

2.3 Neutral monism and structural monism

Neutral monism offers a route between substance dualism and the claim that phenomenal properties must already be distributed among fundamental particles. On a structural-monist reading, the physical and phenomenal are not two substances linked by a bridge law. They are two descriptions of one organized reality, provided the identity claim can be made precise enough to avoid verbal relabeling. Nešić explicitly argues that neutral monism fits structuralist approaches to consciousness better than straightforward panpsychism [5]. ICPS adopts this as a live metaphysical option, not as a premise built into the mathematics. The formalism is intentionally compatible with physicalist, neutral-monist, and some dual-aspect interpretations until additional bridge principles are supplied.

2.4 Why neither idealism nor panpsychism follows

IIT 4.0 has generated debate about the relation between its realist commitments and its idealist-sounding claims about existence [14]. The present framework does not infer that all matter is experiential and does not make experience ontologically prior to external reality. Panpsychism faces a combination problem: if microsubjects are fundamental, why and how do they compose into a macrosubject? Structural monism instead treats the subject-candidate as a maximal organized process at a particular boundary and grain. That move avoids assuming microsubjects, although it does not by itself solve the problem of why the selected structure is phenomenal.

2.5 The precise novelty claim

Recent literature makes a generic “consciousness is intrinsic structure” thesis too broad to count as a distinctive contribution. Grasso, Hendren, and Tononi explicitly pursue content-level correspondences between phenomenal structures and IIT Φ-structures [3]. Carvalho, Langa, and Soler-Toscano define intrinsic information using global dynamical structure, attractors, Morse decompositions, and information landscapes [20]. Esteban and colleagues earlier developed dynamical informational structures related to integrated information [18]. The distinctive move here is different: subjecthood is associated, provisionally, with an equivalence class of temporally extended causal processes under a factorization-preserving representation group, then constrained by multiple non-collapsed criteria. The mathematical innovation is therefore not a new synonym for Φ but a change in the object being evaluated.

2.6 Every Thing Must Go and the demand for objective modality

Ladyman and Ross's Every Thing Must Go is useful here for a reason more specific than its slogan against traditional substance metaphysics. Their ontic structural realism is tied to a principle of naturalistic closure: metaphysics earns credibility by articulating the ontology presupposed or supported by successful science rather than by protecting common-sense categories from scientific revision [6]. On this view, objective modal structure is fundamental. Relations are not merely patterns noticed by a spectator; they concern what systems can do, how quantities covary under laws, and what counterfactual dependencies remain invariant across legitimate representations.
That modal requirement matters for consciousness. A static relational pattern can be duplicated by a picture, a data table, or a simulation trace. If subjectivity is identified with structure while structure means only extensional pattern, the account immediately faces the simulation objection: why would a printed adjacency matrix not instantiate the same subject as the physical system it depicts? Modal structure supplies a principled answer. The physical process supports interventions, transitions, counterfactuals, and causal powers that the inscription merely represents. A formalism for subjectivity therefore has to retain dynamics and intervention, not merely graph topology.
Ladyman, Ross, Spurrett, and Collier, Every Thing Must Go: Metaphysics Naturalized

2.7 The Newman problem and why cardinality is not enough

Newman's objection to structural descriptions is especially severe for consciousness because a vacuous structuralism could fit almost any system. If all that is asserted is that there exists some relation with a specified abstract pattern over a domain of the right cardinality, the empirical content of the claim collapses. The cure is not simply to add more relations. The relations must be physically constrained: transition probabilities, commutation relations, intervention responses, symmetry groups, dynamical invariants, and admissible coarse-grainings cannot be assigned arbitrarily once the physical model is fixed.
The framework below therefore treats a structural description as a package of state space, composition, dynamics, and transformation rules. Two systems count as structurally equivalent only relative to an explicitly declared class of physical redescriptions. This is stronger than graph isomorphism and weaker than indiscriminate mathematical isomorphism. The equivalence group must preserve the composition that makes partitions meaningful. Otherwise an arbitrary global unitary could turn a product basis into an entangled basis and make separability itself look coordinate dependent.

2.8 Neutral monism without automatic panpsychism

Neutral monism becomes attractive when neither traditional material substance nor phenomenal substance is taken as fundamental. A neutral structural monism can instead hold that one concrete process admits physical and phenomenal characterizations without reducing either characterization to the vocabulary of the other. Nešić explicitly develops a neutral-structuralist direction in response to the limitations of both structural realism and panpsychist interpretations [5]. Mindt similarly argues that intrinsic structure and dynamics should be distinguished from merely extrinsic structure and dynamics [4].
Nothing in that move entails that every relation, every quantum state, or every integrated system is conscious. Panpsychism would require an additional thesis connecting fundamental physical organization to phenomenal character. The framework here withholds that thesis. It allows a spectrum of structural candidates to be analyzed while reserving phenomenal attribution for a further identity claim. This makes the mathematics useful even for readers who reject panpsychism, IIT, or neutral monism.

2.9 Relations require a composition rule

A recurrent weakness in structural talk is failure to specify what the relata are allowed to be. Integration is always relative to a decomposition. A matrix can look block diagonal in one tensor factorization and highly coupled in another. The compositional rule Γ therefore belongs in the ontology of the model rather than being treated as bookkeeping. Γ determines the candidate units, their admissible groupings, and the partitions against which irreducibility is evaluated.
This is also where OSR must resist a slide into pure mathematics. A physical factorization is not selected merely because it makes a quantity large. It is constrained by manipulability, locality, operational accessibility, conserved quantities, interaction terms, and the scale at which the model's units are physically individuated. The grain problem cannot be solved by optimizing over unconstrained factorizations of Hilbert space.

3. A Unified Structural Formalism

3.1 The process quadruple

Let a finite process be represented as 𝔖 = (𝒜, ω, {ℰt:s}, Γ). 𝒜 is an algebra of observables. In a finite classical model it can be understood through functions on a finite state space; in a finite quantum model it is a matrix algebra. ω is a normalized state: a probability distribution in the classical case or a positive trace-one density operator in the quantum case. {ℰt:s} is a family of admissible evolutions: stochastic maps classically and completely positive trace-preserving maps quantum mechanically. Γ is a declared compositional structure, such as a product decomposition into nodes or subsystems. Γ is not decorative metadata. Without Γ there is no unique fact about what counts as cutting a system into parts, and therefore no unique integration problem.
𝔖 = (𝒜, ω, {ℰₜ:ₛ}, Γ), t ≥ s.

3.2 Admissible redescription and process identity

Let GΓ be the group or groupoid of transformations that alter representation while preserving Γ. In a labeled classical n-node system, this includes node permutations accompanied by the induced permutation of state coordinates and transition probabilities. In a quantum n-partite model with fixed tensor factors, it includes local basis changes
V
V
1
⊗ ... ⊗
V
n
and permutations of isomorphic factors, provided states, observables, generators, and cut operations are transformed covariantly. An arbitrary entangling global unitary is not automatically a mere change of basis: it can change the tensor-factor decomposition itself. Treating every global unitary as representational would erase the very compositional facts that integration is supposed to evaluate.
[𝔖]GΓ = { g·𝔖 : g ∈ GΓ }.
The equivalence class [𝔖]GΓ, not one coordinate expression of 𝔖, is the candidate for representationally intrinsic structure. This is a stronger requirement than checking that a plotted graph “looks the same” after relabeling. Every proposed diagnostic must either be invariant under GΓ or have its transformation law explicitly stated.

3.3 Irreducibility, closure, persistence, and grain

For a partition π of Γ, define a cut process 𝔖π by a physically explicit intervention that removes cross-partition causal dependence while preserving within-part dynamics. Classical cuts are constructed from induced local transition kernels under a stated intervention baseline μ. Quantum cuts are constructed by removing generator terms whose support crosses π. Let D be a divergence appropriate to the process representation. The instantaneous cut irreducibility is Iπ(t) = D(ℰt, ℰtπ), and the minimum cut irreducibility is Imin(t) = minπ Iπ(t). A temporally thick quantity averages or otherwise summarizes this profile across an interval rather than identifying a subject with an isolated instant.
Īₘᵢₙ(T) = (1/T) ∫₀ᵀ minπ D(ℰₜ, ℰₜπ) dt.
Closure is treated independently from irreducibility. A subsystem may be internally irreducible yet heavily driven by its environment. The classical closure diagnostic used below compares normalized effective information internal to the subsystem with incoming and outgoing contextual sensitivities measured by Jensen-Shannon divergence. Temporal persistence is measured from the stability of the sorted cut-distance fingerprint across lags or times. Grain is handled by explicit coarse-graining maps and induced macro-dynamics. Each quantity has different assumptions; none is allowed to hide those assumptions inside a single coefficient.

3.4 No arbitrary master scalar: Pareto structure

Suppose a candidate subsystem S has a structural signature v(S) = (I(S), C(S), P(S), E(S)), containing irreducibility, closure, persistence, and normalized effective information or a grain-related quantity. A weighted sum λ1I + λ2C + λ3P + λ4E would introduce arbitrary exchange rates between conceptually different properties. ICPS instead uses Pareto dominance. S1 dominates S2 only if it is no worse on every declared criterion and strictly better on at least one. The non-dominated set is the structural frontier. This does not select a unique subject by fiat. It makes genuine underdetermination visible.
S₁ ≻ S₂ iff vᵢ(S₁) ≥ vᵢ(S₂) for every i, with strict inequality for at least one component.

3.5 Computational definitions


3.6 Algebraic unification: commutative and noncommutative realizations

Let a candidate process be written
𝔖 = (𝒜, ω, ℰ, Γ).
Here 𝒜 is an algebra of observables or physically individuated degrees of freedom, ω is a state on that algebra, ℰ is a dynamical map or family of maps, and Γ is a compositional structure specifying legitimate subsystems and tensor or product factorizations. The inclusion of Γ is essential. The triple (𝒜,ω,ℰ) is sufficient for many dynamical questions, but not for integration, because integration asks how a whole compares with specified parts. Without Γ, the word 'partition' is undefined.
In a finite classical system with sample space X, the algebra can be represented by the commutative algebra C(X) of functions on X. A state is a positive normalized linear functional, equivalently a probability distribution p(x). A discrete-time dynamics can be represented by a Markov kernel T(x'|x). In the Schrödinger picture, p is mapped to pT; in the Heisenberg picture, observables are pulled back by the transition operator. The same physical process can therefore be described on states or on observables without changing its causal organization.
For a finite-dimensional quantum system, 𝒜 can be the noncommutative matrix algebra B(H). A state is represented by a density operator ρ ≥ 0 with Tr ρ = 1, while physical evolution is represented by a completely positive trace-preserving map ℰ. The Heisenberg dual ℰ* is completely positive and unital. The algebraic difference is not cosmetic. C(X) is commutative, whereas B(H) generally is not, so no single classical joint sample space reproduces all quantum observables while preserving their operator relations.
Classical: ω ↔ p(x), ℰ ↔ T(x′|x), [A,B] = 0.
Quantum: ω ↔ ρ, ℰ: ρ ↦ ℰ(ρ), [A,B] may ≠ 0.
The shared notation makes a more disciplined comparison possible. The question is no longer whether quantum systems have a mysterious extra ingredient. It is what happens to candidate intrinsic causal structure when the observable algebra changes from commutative to noncommutative while the requirements of composition, intervention, irreducibility, and invariance are held fixed.
In[]:=
Graphics3D[{​​{Opacity[.05],RGBColor[.38,.23,.58],Sphere[{1.3,0,0},.62]},​​{RGBColor[.08,.32,.58],PointSize[.022],Point[Tuples[{-1,1},3].42+ConstantArray[{-1.3,0,0},8]]}​​},Boxed->False,ImageSize->500]
Figure 1A. Algebraic stress test, drawn schematically. The left cube represents a classical joint state space with simultaneously definite coordinates. The right sphere represents incompatible quantum directions. The point is not geometric literalism; it is that the underlying observable algebras support different compositional and probabilistic structures.

3.7 States are not processes

The pair (𝒜,ω) specifies what is actual at an instant but not the counterfactual organization of transitions. Two physical systems can occupy the same density operator and possess different Hamiltonians; two stochastic systems can share the same stationary distribution and possess radically different transition kernels. Conversely, the same dynamics can act on different states. Any proposal that identifies consciousness directly with a state therefore throws away causal information unless dynamics is recoverable from independent physical constraints.
This distinction is particularly sharp in quantum mechanics. A Bell state has strong nonseparability, but the identity channel acting on it has no interaction between the subsystems. A product state entering an entangling interaction begins with no state entanglement, yet the channel has nonlocal causal structure relative to the declared tensor factors. State structure and process structure must therefore be reported separately.

3.8 Composition, partitions, and cut models

Let Γ provide a family of admissible subsystem decompositions. A partition π = {S1,...,Sk} induces a comparison process 𝔖π in which specified cross-part relations are removed or replaced by an intervention baseline. The generic irreducibility functional is
Iπ(𝔖) = D(𝔖, 𝔖π), I(𝔖) = minπ Iπ(𝔖).
The symbol D deliberately hides no metaphysics. Different theories choose different state spaces, repertoires, intervention baselines, and difference measures. IIT 4.0 derives particular intrinsic-information quantities from its postulates [1,22]. The structural diagnostics below use bounded Jensen-Shannon divergence for classical transition kernels and trace distance for channel comparisons because those choices are transparent and stable. Wherever the exact IIT intrinsic difference or quantum intrinsic difference can be isolated without pretending to implement the entire structure, it is computed separately and labeled as such.

3.9 Exact classical ID and quantum QID as bridge quantities

IIT 4.0's intrinsic difference for classical distributions P and Q is a state-specific quantity rather than an average divergence. For discrete distributions it can be written as the maximum weighted pointwise information difference
ID(P,Q) = maxα { pα log₂[pα/qα] }.
Albantakis, Prentner, and Durham extend this idea to finite-dimensional quantum mechanisms [15]. If ρ = Σi pi |i⟩⟨i| and σ = Σj qj |j⟩⟨j|, with Pij = |⟨i|j⟩|², the quantum intrinsic difference is
QID(ρ∥σ) = maxᵢ { pᵢ [ log₂ pᵢ − Σⱼ Pᵢⱼ log₂ qⱼ ] }.
QID reduces to classical ID when ρ and σ commute and equals quantum relative entropy when ρ is pure [15]. It is also invariant under a common unitary change of basis applied to both arguments. These properties make it useful for the invariance discussion. It is still not by itself system Φ, mechanism φ, or a measure of consciousness. It is one theory-motivated difference primitive that can be compared with the more operational trace-distance diagnostics.

3.9 Intrinsic structure as an orbit, not a coordinate description

Let GΓ be a group of transformations interpreted as admissible redescriptions that preserve the declared composition Γ. The candidate intrinsic object is then not one matrix representation but an orbit
[𝔖]GΓ = { g·𝔖 : g ∈ GΓ }.
A structural diagnostic F is representation-invariant when F(g·𝔖)=F(𝔖) for every admissible g. This condition looks elementary, but it forces a difficult philosophical decision: which transformations belong to GΓ? Node relabelings and graph isomorphisms usually do. Local basis changes in quantum subsystems usually do when states, observables, and dynamics are transformed together. A generic entangling unitary generally does not, because it can change the tensor factorization that Γ declares to be physically meaningful. The equivalence class is therefore constrained by physics, not by mathematical convenience.
Figure 1. Orbit-space view of representation. Blue points are coordinate descriptions connected by admissible transformations; the gold center denotes the abstract equivalence class they represent, not an additional physical object.
The figure is intentionally geometric. It should not be read as claiming that the equivalence class literally occupies a higher-dimensional physical space. Its purpose is to distinguish an invariant object from its coordinate representatives. A consciousness-relevant quantity that varies around this orbit would be tracking description rather than intrinsic organization.

4. Classical Integrated Causal Structures

4.1 Finite stochastic systems and intervention-defined cuts

For a finite n-node classical process with state x ∈ X = X1 × ··· × Xn, let K(y|x) be a row-stochastic transition kernel. Given a bipartition A|B and an intervention baseline μ over source states, the local kernel KA(yA|xA) is obtained by averaging over external source variables xB according to μ conditioned on xA and marginalizing external outputs yB. KB is defined analogously. The disconnected comparison kernel is Kπ(y|x) = KA(yA|xA) KB(yB|xB). This construction makes a methodological choice explicit: the meaning of “cut” depends on what is done to severed inputs. The baseline μ is therefore part of the analysis and can be varied.
Iπcl(K; μ) = Σx μ(x) JS( K(·|x) || Kπ(·|x) ).
Jensen-Shannon divergence is bounded, symmetric, and finite for finite distributions. It is not asserted to be uniquely privileged by phenomenology or identical to IIT intrinsic difference. Its role is to provide a transparent, stable diagnostic for how much the full conditional dynamics differ from a factorized intervention-defined approximation. The dependence on μ is a feature to be studied, not a nuisance to be concealed. Barrett and Mediano’s critique of underdetermined Φ-like measures is directly relevant to this discipline [19].

4.1 Two- through five-node systems

The purpose of comparing several sizes is not to reward larger systems. It is to expose how the minimum partition can migrate as recurrent structure, sign, and modular organization change. Effective information and cut irreducibility are reported separately because a channel may transmit substantial information while being decomposable, or be weakly informative yet still resist a particular partition.

4.2 Partition profiles rather than a single number

A sorted cut spectrum is a representation-invariant fingerprint under node relabeling, whereas the identity of a labeled minimum cut is not. The spectrum deliberately retains more information than its minimum. Two systems can share the same minimum while differing markedly in the rest of their partition structure. This is one reason a single Φ-like scalar is too lossy for the ontological role being considered.

4.3 Boundaries as closure, not visual enclosure

A boundary should not be inferred merely because a cluster is visually dense. For a subset S, ICPS computes an induced internal kernel KS, its effective information per retained bit, and two contextual sensitivities: how the distribution of S at the next step changes when the outside context is varied while the present internal state is held fixed, and how the outside changes when S is varied while the present outside state is held fixed. The closure score is the normalized internal term divided by the sum of internal, incoming, and outgoing terms. This is still a model-relative diagnostic, but it removes the size bias of simply counting internal edge weights.

4.4 What a cut actually means

A partition operation is an intervention, not an eraser applied to a picture. In a classical stochastic system, removing the dependence of A on B requires specifying what replaces the severed B-inputs. Uniform noise, stationary-distribution averaging, fixed background states, and causal marginalization are different operations. They answer different counterfactual questions and can produce different minimum partitions. Any cut score that suppresses this dependence behind a single function call risks mistaking a convention for an intrinsic fact.
The classical structural diagnostic used here constructs local conditional kernels by averaging external source variables against an explicit baseline μ and marginalizing external outputs. The disconnected kernel is the product of those local conditionals. This creates a well-defined comparison process, but the baseline remains visible as a modeling choice. Uniform μ corresponds to treating possible source states symmetrically; a stationary μ emphasizes states visited by the autonomous process. Reporting both is more informative than choosing one silently.
Kπ(y|x) = KA(yA|xA; μ) KB(yB|xB; μ).
Iπcl(K;μ) = Σx μ(x) JS[K(·|x) || Kπ(·|x)].

4.5 Structural shell and the geometry of a cut

The three-node recurrent model contains excitatory and inhibitory directed couplings. A structural shell makes the topology visible, but the diagram is not itself the causal analysis. Edge thickness represents coupling magnitude; color separates sign. The calculation beneath it uses the full transition kernel generated by those couplings. The same graph with a different response nonlinearity or noise level would generally have a different cut spectrum.
Figure 2. Three-node recurrent structure. The shell is a visual aid for the weighted directed interaction matrix; irreducibility is evaluated on the induced transition kernel, not on visual connectivity.
Now color one side of the partition {1}|{2,3} separately and mark the cross-boundary arrows. What disappears under the cut is not simply the red arrows as graphical objects. Their causal contribution is replaced by the intervention baseline inside the factorized comparison kernel. The figure therefore corresponds to an operation on a stochastic model, not merely deletion in an adjacency list.
Figure 3. One candidate partition of the same physical process. Cross-part interactions are highlighted. The cut score is the distributional effect of severing them under the declared intervention rule.

4.6 Repertoires, effective information, and why one number is not enough

Effective information measures how informative interventions on present states are about future states under a specified intervention distribution. Cut irreducibility measures how poorly a factorized process approximates the whole. These quantities can vary independently. A deterministic feed-forward circuit may have high effective information while being cleanly decomposable across a directionally appropriate partition. A weak but recurrent stochastic system may be difficult to factorize while transmitting relatively little state information. Treating both as 'integration' obscures their causal meanings.
For the subject problem, the separation is useful. Specificity without unity is insufficient because a lookup table can be highly selective while decomposing into independent parts. Unity without specificity is also insufficient because a strongly coupled but almost random system can resist a cut while specifying little about its own alternatives. IIT's formalism handles this through multiple postulates and nested maximizations [1,22]. The present diagnostics preserve the distinction rather than compressing it into a new surrogate Φ.

4.7 A limited IIT 4.0 layer

The intrinsic-difference primitive can be evaluated exactly on classical repertoires even when the full IIT 4.0 unfolding is computationally out of scope. This matters because Jensen-Shannon divergence and intrinsic difference answer different questions. JSD averages a symmetric divergence between two distributions. ID selects the state-specific difference that matters most from the constrained distribution's perspective. Agreement between them is not expected in general.
The two values should be read side by side, not ranked as competing estimates of consciousness. Their disagreement quantifies the fact that 'difference between whole and cut' is mathematically underdetermined until a theory specifies what aspect of difference is intrinsic. This is exactly why general-purpose integration measures should not be renamed Φ without reproducing the postulates and optimization structure that give Φ its intended meaning.
IIT 4.0 reference image. The figure illustrates the theory’s separation of existence, intrinsicality, information, integration, and exclusion. It is included as a visual reference to the formal architecture summarized below [1].

4.8 IIT 4.0: from scalar irreducibility to an unfolded causal structure

The strongest lesson to import from IIT 4.0 is methodological rather than terminological. A system-level irreducibility value is only one layer of the analysis. IIT begins from an interventional transition probability model, evaluates a candidate substrate from its intrinsic perspective, identifies the cause and effect states that maximize intrinsic information, searches directional partitions for a minimum partition, and only then unfolds the internal mechanisms of a maximal substrate into causal distinctions and relations [1]. That sequence matters. It prevents a single divergence between a whole and a cut model from being silently promoted into a complete ontology.

System-level intrinsic information

For a current system state s and a possible effect state s̄, IIT 4.0 weights pointwise informativeness by selectivity. The unconstrained effect probability is obtained by averaging over interventions on all current states with a uniform prior. The effect-side intrinsic information therefore takes the state-specific form below. The cause side is defined analogously from the interventional cause repertoire [1].
iiₑ(s,s̄) = pₑ(s̄|s) log₂[pₑ(s̄|s) / pₑ(s̄)]
The maximal cause/effect state is selected by maximizing intrinsic information. Integration then asks whether that selected causal constraint survives a disintegrating partition. For a mechanism m, purview Z, selected effect state zₑ⁰, and partition θ, the effect-side integrated information is
φₑ(m,Z,θ) = πₑ(zₑ⁰|m) log₂[ πₑ(zₑ⁰|m) / πₑθ(zₑ⁰|m) ]
and the minimum-information partition is the partition that minimizes the normalized loss of irreducibility. This is not the same operation as the generic cut distances used elsewhere in ICPS. The latter remain structural diagnostics; the IIT quantities are theory-specific constructions tied to IIT’s postulates and intervention rules.

Distinctions, relations, and the geometry of composition

A mechanism becomes a causal distinction only when it specifies a definite maximally irreducible cause and effect. Distinctions can then participate in relations when their cause/effect purviews overlap congruently over common units. IIT 4.0 assigns integrated information to both distinctions and relations and defines the structured information of the unfolded Φ-structure by summing the irreducibility values of its causal components [1]. Schematically,
Φ = Σc∈C φ(c), where C contains the distinctions and relations of the unfolded structure.
This compositional layer is especially important for the present project. Two systems may share a similar system-level minimum cut while differing radically in the number, specificity, overlap, and higher-order binding of their internal causal distinctions. That means that “how irreducible is the whole?” and “what internally articulated structure exists within the whole?” are different questions. ICPS therefore treats a cut spectrum as only one coordinate of a richer structural fingerprint.

A 3D projection as an explanatory device

IIT 4.0 visualizes cause–effect structures by projecting them into three dimensions: mechanisms are linked to cause and effect endpoints, link thickness can encode distinction irreducibility, and second- or third-degree relations can be represented by edges or translucent faces [1]. The projection is pedagogically powerful, but the Euclidean coordinates of the drawing are not themselves intrinsic. Rotating, shearing, or otherwise re-embedding the same incidence structure should not create a new causal entity. This makes the IIT visualization convention an unusually useful bridge to the invariance principle developed later in the notebook.
The following figure is an original Wolfram construction inspired by that visualization methodology, not a reproduction of IIT Figure 6 and not a claim that the displayed toy values constitute a computed IIT Φ-structure. Black points are mechanisms, red points are cause endpoints, green points are effect endpoints, orange tubes represent distinction links with thickness proportional to a toy irreducibility weight, and translucent blue faces represent higher-order relational binding. Its role is to make composition geometrically inspectable while keeping the ontology in the incidence and weights rather than in the chosen coordinates.
Figure 3A. Original IIT-inspired causal-geometry schematic. The same abstract distinction–relation complex can be embedded in many coordinate systems; only the causal incidence pattern and attached values are intended to carry structural meaning.
A useful methodological extension follows immediately. For every such visualization, the notebook should be able to store the abstract hypergraph of mechanisms, cause/effect purviews, relation faces, and weights independently of the 3D embedding. The figure is then a rendering of an invariant combinatorial–causal object, rather than the object itself. This provides a direct computational test of representational invariance: perturb the coordinates while preserving the hypergraph, and verify that every consciousness-relevant diagnostic remains unchanged.

4.9 Temporal lifting: from one-step kernels to causal fingerprints

One-step irreducibility can miss organization that appears over multiple updates. For lag ℓ, the effective transition kernel is K^ℓ. The ordered collection of minimum-cut spectra across lags forms a temporal fingerprint. Rather than treating the subject as whichever subset wins at a single update, a process-level analysis asks whether an approximately stable fingerprint persists across neighboring lags.
Σcl(ℓ) = sort{ Iπ(K^ℓ) : π admissible }.
Sorting removes dependence on partition labels while retaining the multiset of cut sensitivities. It is therefore useful for representation-invariant persistence, although it deliberately discards which named partition produced each value. A richer implementation could align partitions by graph automorphism or optimal transport on partition space. The current fingerprint should be understood as a conservative invariant summary, not a complete process identity.
Figure 4A. Temporal cut spectrum for the recurrent three-node classical process. The rapid drop after one step shows that one-step irreducibility and long-lag causal distinctness can diverge; mixing can erase a cut signature even when immediate recurrence is nontrivial.
Figure 4. Layered causal geometry across five updates. The vertical direction is time. A subject-candidate is associated with a stable organization of transitions across layers, not with one horizontal slice alone.

5. Quantum Integrated Structures

5.1 State structure and process structure are different objects

A quantum state ρ describes statistical and relational properties of a system at an instant. A quantum channel ℰ describes a transformation. Conflating them is especially dangerous in consciousness discussions because entanglement is a property of states or processes relative to a factorization, while causal irreducibility concerns what the dynamics would do under interventions or cuts. The quantum analysis therefore reports three layers separately: static state descriptors, channel-level cut distance, and state-conditioned response to the same physical cut.
Iπq(t) = ½ ‖ J(U(t)) − J(Uπ(t)) ‖₁.

5.2 Product, Bell, GHZ, W, cluster, and mixed states

These quantities characterize the states, not the causal organization that produced them. In particular, negativity depends on a bipartition and detects one form of nonseparability; mutual information includes both classical and quantum correlations; l1 coherence depends on the chosen basis and therefore is not representationally intrinsic by itself. Their inclusion is diagnostic: they make it possible to see when a dynamic irreducibility result is being mistakenly read off from a static state property.

5.3 Embedding the states in actual dynamics

The Choi cut profile depends on the channel and the declared cut operation, not on which initial state happens to be supplied. The state-conditioned response asks a different question: for this particular ρ, how distinguishable are the outputs of the full and cut dynamics? A state can therefore have high entanglement and weak response to a given dynamic cut, or zero initial entanglement and strong response.

5.4 Temporally thick quantum structure

Unitary dynamics can make an instantaneous cut distance oscillate strongly. A subject ontology tied to the maximum at one instant would therefore risk making subject identity flicker with phase. ICPS treats time as part of the candidate object: the average magnitude and the persistence of the invariant cut fingerprint across a window are independent constraints. Process-tensor methods make this temporal emphasis natural in quantum theory because multitime processes can themselves be represented and reconstructed operationally [16].

5.5 The noncommutative change of language

The quantum case is not obtained by replacing probabilities with amplitudes in the classical formulas. The state space, composition law, and admissible interventions change. Density operators encode mixtures and coherence; completely positive maps encode physical transformations; reduced states are obtained by partial trace; and subsystems are defined relative to a tensor factorization. Because observables may fail to commute, there need not be a classical joint distribution that preserves the full algebraic structure.
This makes the quantum case a genuine stress test for any claim about intrinsic organization. If a proposed measure depends on a preferred basis even after the state, dynamics, and observables are transformed consistently, it is not representation-invariant. If it reports 'integration' solely because a state is entangled under identity dynamics, it has conflated state nonseparability with causal interaction. If it changes under a global refactorization of Hilbert space, one must decide whether the refactorization is a mere redescription or a change in physical composition.

5.6 Quantum intrinsic difference and operational trace distance

Two difference measures appear for distinct reasons. Trace distance
Dtr(ρ,σ) = ½ ‖ρ − σ‖₁.
has a clear operational interpretation: it determines optimal state distinguishability under measurement. For channel dynamics represented by normalized Choi states, it gives a stable diagnostic of how much the full and cut channels differ. QID, by contrast, is designed to implement IIT's intrinsic-information desiderata for quantum mechanisms [15]. It is state-specific in the eigenbasis of the constrained repertoire and invariant under a common unitary basis change. Neither should silently substitute for the other.

5.7 Product, Bell, GHZ, W, cluster, and mixed states inside mechanisms

Static states are embedded in explicit interacting mechanisms. For each state, three conceptually different quantities are reported: mutual information across a selected bipartition, negativity across that bipartition, and the minimum state-conditioned response to cutting the Hamiltonian couplings. The first measures total correlation, the second detects a class of quantum entanglement, and the third asks how much the evolved state changes when the causal mechanism is severed. Coincidence among the three is not assumed.
Figure 5. Three-axis diagnostic space for representative states embedded in explicit interactions. Similar entanglement coordinates can coexist with different dynamical cut responses; the product state has zero initial correlation yet a nonzero response to cutting the subsequent interaction.
The geometry makes the logical independence visible. Bell, GHZ, and cluster states sit at nearly the same static first-versus-rest correlation coordinates, yet their cut responses differ because the mechanisms in which they are embedded differ. The product state begins at the origin of the static-correlation plane but does not lie at zero on the process axis. A state-only theory would miss that distinction.
Figure 5A. Quantum channel cut spectrum through time. The channel-level irreducibility increases smoothly in this window, while the ordering of cuts remains separated. The figure demonstrates why a process must be studied over an interval rather than sampled at one arbitrary phase.

5.8 Choi states encode channels, not experiences

This is a useful example of the larger methodological rule: representation can make a structural property computable without becoming the ontology of that property. Similar caution applies to qualia spaces, causal graphs, and state-space embeddings.

6. Entanglement Is Not Integration

6.1 Three explicit non-implications

The following cases are not rhetorical examples. They are direct countermodels to three common identifications: entanglement with causal integration, correlation with causal integration, and coherence with causal integration. A successful theory must survive these separations.
The first row has nonzero entanglement but exactly factorized identity dynamics. The second has correlation without entanglement and again no dynamic integration. The third begins from a product state yet evolves under a cross-partition interaction, producing nonzero channel irreducibility and a nonzero state-conditioned response. Therefore none of the implications “entangled ⇒ integrated,” “correlated ⇒ integrated,” or “unentangled ⇒ unintegrated” is valid.

6.2 Why this matters philosophically

Quantum-consciousness proposals often gain rhetorical force from the nonclassicality of entanglement. But nonseparability is not automatically the kind of unity demanded by a theory of subjectivity. Entanglement can be inherited from preparation history while current dynamics factorize; conversely an unentangled state can sit inside a strongly interacting process. The ontological candidate must therefore be process-sensitive. This is also why the quantum calculations here are not a claim that brains require coherent quantum computation. They are a test of conceptual distinctions under the strongest available examples of nonseparability.

6.3 A four-quadrant independence test

The slogan entanglement ≠ integration becomes sharper when written as a four-quadrant test. Consider whether state entanglement E is zero or nonzero and whether process irreducibility I is zero or nonzero. All four combinations are conceptually available.
◼
  • E = 0, I = 0: a product state under factorized dynamics.
    • ◼
  • E > 0, I = 0: an entangled state preserved by noninteracting local dynamics, including identity evolution.
    • ◼
  • E = 0, I > 0: a product input entering an interacting channel whose cut version differs from the full dynamics.
    • ◼
  • E > 0, I > 0: an entangled state embedded in genuinely interacting dynamics.
    • The existence of the middle two cases blocks either implication. Entanglement is a property of a state relative to a subsystem factorization. Process irreducibility is a property of dynamics relative to a declared cut operation and composition. Their mathematical domains overlap, but neither subsumes the other.

    Proposition 6.1. State nonseparability does not entail dynamical irreducibility.

    Let ρAB be any entangled state and let the evolution be the identity channel IA ⊗ IB. The full channel already factorizes across A|B, so the cut channel is identical to the full channel. Any faithful channel distance therefore returns zero, while an entanglement measure such as negativity can remain strictly positive. The counterexample requires no controversial interpretation of quantum mechanics.

    Proposition 6.2. Dynamical irreducibility does not require initial entanglement.

    Let the initial state be |00⟩ and let U = exp(-it Z⊗Z - it h(X⊗I + I⊗X)) with nonzero interaction strength. The initial negativity is zero. Removing the Z⊗Z interaction changes the output for generic t and h, producing a strictly positive state-cut response and a nonzero Choi-channel distance. The causal difference belongs to the process even before the input has acquired entanglement.
    These propositions matter philosophically because quantum holism is often invoked too quickly. Entanglement demonstrates that the state of a whole need not reduce to independently possessed pure states of its parts. That is relevant to structural realism. But the kind of unity relevant to a subject is more demanding: it concerns the system's own counterfactual causal organization, its boundary, its grain, and its persistence.

    7. The Invariance Principle

    7.1 Classical permutation invariance

    Proposition 1. Let σ be a permutation of node labels and let Kσ and μσ be the pushforwards of a finite transition kernel K and intervention distribution μ. Then the multiset of bipartition cut divergences is invariant: SpecCut(K, μ) = SpecCut(Kσ, μσ). Proof sketch: σ induces a bijection on bipartitions and on source and destination states. Jensen-Shannon divergence is invariant under a common permutation of outcome coordinates, and the μ-weighted sum is unchanged under pushforward. Therefore individual labeled partitions move, but the sorted cut spectrum is unchanged.

    7.2 Quantum local-basis invariance

    Proposition 2. Let V = ⊗i Vi preserve the declared subsystem factorization Γ. If H and every cut Hamiltonian Hπ transform covariantly as H′ = VHV† and Hπ′ = VHπV†, then the Choi trace distance between the full and cut unitary channels is invariant. This follows because vectorization converts the conjugation into a unitary action on the Choi vector, and the trace norm is unitarily invariant. The proposition does not say that every global unitary is ontologically innocent. A global entangling transformation can change the factorization itself and therefore changes the decomposition question rather than merely its coordinates.

    7.3 Invariance is stronger than aesthetic symmetry

    A numerical quantity can be basis invariant yet still fail to be intrinsic if the basis transformation changes what counts as a subsystem. The correct question is not “is this formula invariant under all transformations?” but “what transformation group corresponds to the same physical organization?” This is why Γ is part of the process quadruple. The invariance principle is therefore two-layered: diagnostics must be invariant under admissible redescriptions, and the admissible class itself must be justified by the ontology of composition. French’s emphasis on laws, symmetries, and modal structure is useful precisely because invariance has physical content only relative to a theory of what transformations preserve the system [8].

    7.4 Not every unitary is a mere change of basis

    Quantum mechanics tempts an overly strong invariance principle: because Hilbert-space representations are basis-dependent, perhaps every unitary transformation should preserve subject identity. That conclusion is false unless composition is ignored. A local unitary U1⊗...⊗Un changes bases inside already identified subsystems. A generic entangling unitary can map those subsystems into new nonlocal factors. If Γ encodes physical units, locality, or manipulable components, the latter transformation changes the decomposition against which causal integration is defined.
    The admissible group GΓ should therefore preserve the physical compositional structure. This is analogous to gauge reasoning more generally: invariance is meaningful only after the redundancy group has been identified. Declaring every mathematical isomorphism a redundancy would erase precisely the physical structure the theory is trying to characterize.

    7.5 Classical graph isomorphism

    For a node permutation p, the stochastic kernel must be transformed by the induced permutation of global states. A cut-spectrum invariant should then be unchanged as a multiset even though the named partitions are relabeled. The notebook tests this numerically rather than merely asserting it.

    7.6 Local-basis covariance of the quantum cut comparison

    Let H be the full Hamiltonian and Hπ its cut counterpart. Under a common local unitary V = ⊗i Vi, define H' = VHV† and H'π = VHπV†. Then U'(t)=VU(t)V† and U'π(t)=VUπ(t)V†. Because trace distance is invariant under common unitary conjugation, the Choi-distance cut diagnostic is unchanged. This is a genuine representation invariance rather than a numerical coincidence.
    Dtr[J(U), J(Uπ)] = Dtr[J(VUV†), J(VUπV†)].
    A residual near machine precision is the expected result. Crucially, the Hamiltonian and its cut are transformed together. Rotating the state while leaving the dynamics fixed would describe a different physical preparation, not merely a coordinate change.
    Figure 6. Representation orbit. The physically meaningful candidate is associated with the entire orbit under admissible redescriptions, while inadmissible transformations that alter Γ belong to a different orbit.

    8. The Grain of the Subject

    8.1 Coarse-graining must induce dynamics

    A macro-description is not obtained by averaging edge magnitudes. Let q:X→Y be a surjective coarse-graining map from microstates to macrostates. Given an intervention baseline μ, the macro kernel is obtained by conditioning μ on the fiber q−1(y), averaging the micro transition rows in that fiber, and summing destination probabilities over the destination fiber q−1(y′). The result is row-stochastic by construction. This makes explicit which microstates realize a macrostate and how interventions distribute probability within that realization class.
    Kq(y′|y) = Σx∈q−1(y) μ(x|y) Σx′∈q−1(y′) K(x′|x).
    This definition also exposes a familiar problem in causal-emergence work: macro results can depend on the intervention distribution. Hoel’s effective-information account shows how macro causal descriptions can sometimes become more informative than micro descriptions [17], while later critiques emphasize the significance of averaging and intervention assumptions. ICPS therefore reports those assumptions rather than treating the macro level as automatically privileged.

    8.2 A four-node micro/macro experiment

    8.3 Intervention sensitivity at the macro level

    If the macro result changes materially when μ changes, that dependence is substantive. It means the proposed grain has not been specified independently of how interventions are distributed within macrostate fibers. Rather than hide this dependence, a rigorous analysis should either justify μ physically or demand stability across a family of admissible μ.

    8.4 The subject-candidate as a Pareto frontier

    A unique “complex” need not emerge. Multiple non-dominated subsets can survive because one is more closed while another is more irreducible or more temporally stable. This is not a computational failure. It is an honest diagnosis that the chosen principles do not yet determine a unique ontology. Any theory that nevertheless declares one winner must add another principled constraint rather than an arbitrary weighted sum.

    8.5 Quantum grain

    The analogous quantum construction requires a coarse-graining channel 𝒞 and a reconstruction or preparation map ℛ. An effective macro channel has the form ℰmacro = 𝒞 ∘ ℰmicro ∘ ℛ. Different choices of 𝒞 and ℛ correspond to physically different claims about which degrees of freedom are ignored, encoded, or controlled. A projection onto a code subspace, a partial trace over an environment, and a renormalization-group map are therefore not interchangeable notions of grain. The algebraic formalism accommodates them, but the physical choice must come from the model rather than from a desire to maximize integration.

    8.6 Coarse-graining as a quotient of dynamics

    A macro-variable is not obtained by averaging coupling weights. Let q:Xmicro -> Xmacro be a surjective map from microstates to macrostates. A macro transition kernel must be induced from the micro dynamics together with a rule for weighting microstates inside each macrostate. With baseline μ, the induced kernel is
    Tq(b|a) = Σx:q(x)=a μ(x|a) Σy:q(y)=b T(y|x).
    This equation makes two facts explicit. First, coarse-graining is a mapping of state spaces, not merely a grouping of nodes. Second, macro dynamics depends on how microstates within a macrostate are treated. Uniform conditional weighting and stationary conditional weighting can disagree when the micro dynamics spends unequal time in the constituent states. That dependence should be inspected rather than hidden.

    8.7 Intrinsic units and the 2026 grain result

    Marshall, Findlay, Albantakis, and Tononi extend IIT 4.0 in 2026 to systems containing candidate macro units and show in simulated examples that system integrated information can be greater at a macro grain than at the corresponding micro grain [21]. This result is philosophically important because it blocks a simplistic identification of ontological privilege with the finest available description. If cause-effect power is the criterion, the relevant unit can be constituted by finer units while being causally more definite at a coarser grain.
    The present framework adopts the question but not an automatic maximization rule. A sharp maximum can be informative, yet a subject that jumps grains under tiny perturbations would have unstable identity. Structural persistence therefore asks whether a family of nearby coarse-grainings supports approximately the same causal organization. The ontologically interesting object may be a robust plateau or equivalence class across grain rather than an isolated numerical peak.
    Figure 7. A coarse-graining map as a many-to-one mapping from micro units to macro units. The arrows encode constitution of macro variables; they do not by themselves establish that the macro grain is causally superior.
    Illustrating micro-, meso-, and macroscale organization in biological systems. It motivates the grain problem but does not by itself identify the scale at which a conscious subject exists.

    8.8 Grain persistence

    Let Cλ denote a family of physically licensed coarse-grainings indexed by scale λ. Each produces a process 𝔖λ and an invariant cut fingerprint Σλ. A scale-stability functional can compare neighboring fingerprints,
    Pgrain = exp[ − ⟨ ‖Σλ+δ − Σλ‖ / (‖Σλ‖ + ε) ⟩λ ].
    High grain persistence does not mean every scale is equivalent. It means that the causal identity of the candidate is not an artifact of a knife-edge descriptive resolution. Low persistence is equally informative: it suggests that the purported subject boundary or integration maximum is highly dependent on how units are bundled.

    9. Decoherence and the Classical Limit

    9.1 Destroying coherence continuously

    Decoherence is a useful stress test because it separates several forms of relational organization. A phase-damping channel can suppress off-diagonal coherence in a chosen pointer basis while leaving classical populations and some mutual information intact. If an alleged measure of “intrinsic unity” simply tracks l1 coherence, it will vanish with dephasing even when nontrivial classical causal structure remains. Conversely, if a channel-level cut diagnostic remains large while all state entanglement disappears, that demonstrates again that the two quantities answer different questions.

    9.2 What survives classicalization

    The dephasing path should not be read as a literal model of the brain. Its conceptual role is cleaner: it demonstrates which relational properties are basis-sensitive quantum coherence, which are entanglement, which are total correlations, and which are inherited from the interaction structure of the dynamics. A structural ontology that aims to apply continuously from quantum microphysics to effectively classical macrodynamics should not make subjecthood depend on a quantity that disappears merely because environmental decoherence selects a stable pointer basis, unless it also supplies a biological reason to privilege coherent quantum states.

    9.3 An analytic dephasing trajectory

    The cleanest quantum-to-classical stress test begins with the Bell state and applies local phase damping. Under the dephasing map used here, each one-qubit off-diagonal acquires a factor (1-γ); the Bell coherence |00⟩⟨11| therefore acquires (1-γ)². At γ=1 the state becomes the classical correlated mixture 1/2(|00⟩⟨00|+|11⟩⟨11|).
    ρBell(γ) = ½[ |00⟩⟨00| + |11⟩⟨11| + (1−γ)²(|00⟩⟨11| + |11⟩⟨00|) ].
    Several structural quantities now separate cleanly. The l1 coherence is (1-γ)². The negativity is (1-γ)²/2. Both vanish at complete dephasing. Mutual information does not vanish: it falls from 2 bits for the pure Bell state to 1 bit for the perfectly classical correlated mixture. Decoherence therefore destroys specifically quantum coherence and entanglement while leaving a nontrivial relational structure.
    Figure 8. Analytic decoherence trajectory in the space of coherence, negativity, and mutual information. The path ends with zero entanglement and zero coherence but nonzero total correlation.

    9.4 Does causal irreducibility survive dephasing?

    The more important experiment embeds the state in interacting dynamics and applies the same dephasing to the outputs of both the full and cut mechanisms. The state-conditioned cut response is then tracked alongside coherence, negativity, and mutual information. If the cut response remains substantial after entanglement has vanished, the irreducible effect of the interaction is not reducible to quantum coherence. If it collapses in lockstep with coherence for a particular mechanism, quantum structure is contributing specifically to that mechanism's cut sensitivity.
    The experiment is intentionally neutral about brains. It does not require or suggest long-lived macroscopic neural entanglement. Its purpose is conceptual discrimination. Quantum mechanics supplies a continuous path along which specifically quantum relations can be removed while classical correlation and causal interaction are monitored separately.

    9.5 The classical limit is not simply γ = 1

    Complete dephasing in a chosen basis is not a universal derivation of classical mechanics. Classicality also involves environment-selected stable observables, suppression of interference between relevant histories, restricted measurements, and often effective stochastic dynamics. The γ-family should therefore be interpreted as a controlled coherence-destruction experiment, not as a complete theory of the quantum-classical transition.
    This caution reinforces the invariance principle. A dephasing basis becomes physically meaningful only when the environment or interaction Hamiltonian singles it out. Otherwise 'remove coherence in this basis' can itself be an observer-imposed operation. The pointer basis is part of the dynamics, not a metaphysical gift.

    10. The Ontology of the Subject

    10.1 Six metaphysical readings of the same mathematics

    The calculations do not uniquely determine a metaphysics. Reductive physicalism can read the invariant process structure as an objective physical organization and deny that anything further is required. Property dualism can accept the same organization as the physical base while adding irreducible phenomenal properties. Idealism can reverse the order and treat physical structure as an abstracted pattern within experience. Panpsychism can distribute experiential or proto-experiential properties more fundamentally and treat the macrostructure as a combination problem. Neutral monism can take the concrete process as neither intrinsically mental nor intrinsically physical prior to description. Ontic structural realism can emphasize that the relevant physical reality is modal and relational rather than substance-first.
    ICPS is most naturally paired with a structural neutral monism: one concrete process, public physical description on one side, intrinsic phenomenal presentation on the other. But this pairing is not encoded into the code. The mathematical constraints are intended to make the metaphysical claim harder to state loosely. The identity thesis is acceptable only if it respects compositional structure, causal closure, representation invariance, temporal continuity, and the distinction between a simulation of a process and the process’s own counterfactual causal organization.

    10.2 A comparison without a winner

    Reductive physicalism offers ontological economy but must explain why a complete structural-dynamical description is not merely correlated with phenomenology. Dualism preserves the apparent gap but faces interaction or causal-exclusion problems. Panpsychism moves phenomenality downward and inherits combination or decombination problems. Idealism avoids a mind-to-matter bridge by making experience primary but must recover the apparent autonomy and stability of external physical structure. Neutral monism avoids two primitive substances but requires a principled account of the neutral base. Structural monism adds the proposal that the neutral base is constituted by modal relational organization, but then faces Newman-style triviality unless that organization is causally and dynamically substantive. The point of the formal work is to sharpen these liabilities, not to hide them behind terminology.

    10.3 Subjecthood as temporally thick structural identity

    The most distinctive ontological suggestion is that a subject should be modeled as a temporally thick process rather than an instantaneous state. What persists is not a fixed microstate but an invariant organization of counterfactual transitions, boundaries, and internal differentiation. A person can remain the same subject while every neural microstate changes; an instantaneous-state ontology makes that persistence obscure. Process identity allows the relata to change while higher-order organization remains within an equivalence class. The persistence diagnostic below is only a toy realization of this idea, but the conceptual target is clear: identity through structured transformation, not identity through frozen material configuration.

    10.4 The subject as a causally closed perspective

    A physical subject is neither perfectly isolated nor merely a subsystem selected by an observer. Let S be a candidate embedded in a larger universe U. Three directional quantities can be distinguished: internal self-determination, contextual sensitivity to U\\S, and causal influence from S to U\\S. Closure is high when internal organization dominates both boundary terms; autonomy is high when internal organization dominates incoming dependence even if the system strongly affects its environment.
    B(S) = ( Cself(S), Cin(U∖S → S), Cout(S → U∖S), A(S), K(S) ).
    This vector avoids a common mistake: equating consciousness with physical isolation. A brain, organism, or artificial agent can be open to energy, matter, and information while still possessing a relatively self-determining internal causal organization. Conversely, a subsystem can be shielded from the environment yet internally decomposable. Boundary, closure, and irreducibility are separate questions.
    Figure 9. A candidate causal boundary embedded in a larger network. Blue nodes lie inside the candidate, gray nodes outside. Internal, incoming, and outgoing arrows remain visible; closure is dominance of internal causal organization, not disconnection.

    10.5 The self as boundary condition rather than hidden object

    This picture removes the need for a homuncular self located at a point inside the system. The 'self' is a dynamical boundary condition: a region whose internal causal organization supports a stable perspective from which differences make differences primarily within the region. The boundary can move, sharpen, blur, or reorganize without requiring a persisting substance underneath it.
    A structural account of selfhood therefore resembles other process ontologies. A flame, vortex, or biological organism maintains identity through organized turnover rather than material sameness. The analogy is limited because phenomenality adds an explanatory burden absent from ordinary process identity, but it shows that persistence without a self-subsistent bearer is not conceptually incoherent.

    10.5 Metaphysical positions under the same formal evidence

    The calculations underdetermine metaphysics. Physicalism can treat ICPS quantities as higher-level organizational properties realized by physical dynamics. Neutral monism can treat the same process as admitting physical and phenomenal aspects without privileging either vocabulary. Russellian monism can argue that physics captures relational/dispositional structure while phenomenal character supplies categorical nature. OSR can deny that a further categorical bearer is needed and place ontology in objective modal structure. Panpsychism can add the thesis that some fundamental structures already possess phenomenal character. Dualism can accept every structural result while denying that structure exhausts phenomenal reality.
    The formal evidence therefore constrains these positions asymmetrically but does not choose among them by numerical score. Any view that makes subject identity depend on arbitrary labels is disfavored by invariance. Any view that treats static entanglement as sufficient is refuted by the counterexamples. Any view that requires a perfectly isolated subject conflicts with open-system causal organization. The residual disagreement concerns what ontological significance to assign the surviving intrinsic process structure.

    11. Intrinsic Structure and the Explanatory Gap

    11.1 The bridge principle stated without concealment

    The strongest claim available to structural monism is an explanatory identity: the concrete intrinsic process is the phenomenal occurrence, while physics describes that same occurrence under an observer-independent representational scheme. This is coherent only if “intrinsic process” does more work than “whatever causes the reports.” The process must be independently characterized by its algebra, dynamics, factorization, invariants, intervention structure, and temporal organization. Even then, the identity is a bridge principle. Mathematics can show that the candidate has unity and invariance; it cannot derive the first-person character of existence from a trace norm.

    11.2 Newman’s challenge revisited

    Newman’s objection blocks an easy escape: if the proposed structure is purely extensional, any sufficiently large domain can realize it [9]. ICPS therefore treats transition possibilities, interventions, and transformation symmetries as constitutive. The modal facts matter. Two isomorphic snapshots can differ as processes because different counterfactual transitions are physically available. This also distinguishes a static representation from an instantiated causal organization. The response is not a proof that structuralism is true; it is a constraint on what “structure” must mean if the view is to remain substantive.

    11.3 The unfolding argument

    Doerig and colleagues argue that causal-structure theories face an unfolding problem: systems can be constructed that preserve input-output function while altering internal recurrent causal structure, forcing a theory either to accept divergent consciousness despite functional equivalence or to weaken its causal claims [12]. ICPS does not evade this objection by definition. It accepts the dilemma as a discriminating test. If phenomenology is identical to intrinsic causal process structure, then sufficiently faithful functional equivalence at the behavioral boundary need not guarantee phenomenal identity. That is a substantive commitment. To make it scientific, one must identify interventions or conditions in which structurally different but behaviorally comparable systems generate divergent predictions that are not reducible to verbal reports alone.

    11.4 Empirical constraint after the 2025 adversarial collaboration

    The 2025 Nature adversarial collaboration found support for some predictions of IIT and GNWT while substantially challenging key claims of both [13]. For a structural identity program, the lesson is methodological. Neural localization claims and metaphysical identity claims should not be welded so tightly that one failed cortical synchronization prediction refutes every structural thesis. Conversely, metaphysics should not retreat from experiment whenever a prediction fails. ICPS therefore separates the abstract structural conditions from hypotheses about which biological variables instantiate them. A future empirical version would need preregistered mappings from neural interventions to predicted changes in the process signature.

    11.5 Falsification and discrimination conditions

    The framework would be undermined if its diagnostics were unstable under transformations that clearly preserve the same physical organization; if the selected boundaries depended mainly on arbitrary coding choices; if temporally adjacent states of an evidently continuous process produced unrelated structural identities without a principled timescale; if coarse-graining results could be reversed by innocuous changes of intervention convention; or if every sufficiently connected system occupied the same Pareto region. More strongly, a structural identity theory would face empirical pressure if interventions that selectively alter invariant causal organization while preserving relevant behavior never altered conscious content, or if large changes in reported phenomenal organization occurred while every plausible intrinsic process invariant remained fixed. These are difficult experiments, but they state what failure would mean.

    11.6 The hard problem after structural reformulation

    Structural reformulation does not dissolve the hard problem by changing vocabulary. It changes the target of the explanatory question. Instead of asking how an externally characterized material state produces an ontologically separate phenomenal item, it asks whether the external causal description is incomplete because it describes one concrete process only from outside. If so, phenomenality could be the intrinsic mode of existence of that process rather than an effect added to it.
    The central unresolved question is then sharper: what licenses the move from causal intrinsicality, representational invariance, and self-determination to phenomenal intrinsicality? None of the computations supplies that license. They can eliminate bad candidates and clarify what a coherent identity thesis would require. They cannot make the final ontological identification true by definition.

    11.7 The unfolding argument as a stress test

    Doerig and colleagues argue that causal-structure theories face a dilemma because functionally equivalent systems can be constructed with different internal causal organization [12]. A theory identifying experience with causal structure must either predict different consciousness despite matched input-output behavior or retreat to functional equivalence. ICPS accepts the first horn at the level of principle: if internal causal organization is part of the identity claim, behavioral equivalence cannot guarantee phenomenal equivalence.
    That response preserves internal consistency but creates an empirical challenge. If two systems are externally indistinguishable by every feasible behavioral test, what evidence could discriminate the consciousness claim? The framework therefore needs interventions or structural measurements that access internal causal organization, not report alone. This is an epistemic difficulty, not a formal contradiction, but it limits empirical reach and should remain explicit.

    11.8 From instantaneous dynamics to process tensors

    A one-step channel assumes that the relevant causal state can be summarized at one time. Open quantum systems and biological systems often violate that assumption. Environmental memory can make future evolution depend on earlier interventions even after the current reduced state is fixed. Process tensors provide a natural language for such multitime causal structure: they map sequences of interventions to outcome states while retaining temporal correlations [16,28,29].
    This is especially relevant to a temporally thick subject. If subject identity belongs to an equivalence class of processes rather than instantaneous states, then multitime intervention structure is closer to the target than a single density operator or TPM. A future formalism could define cuts not only across spatial subsystems but across temporal legs of a process tensor, allowing temporal integration and memory to be analyzed directly.
    Figure 10. Schematic process-tensor geometry. The blue rail is the system, the gray rail an environment carrying memory, and gold nodes mark interventions. Cross-time system-environment paths make clear why the current reduced state need not contain the whole causal history.
    The diagram also clarifies why persistence should not be defined as similarity of successive states alone. Two trajectories can pass through the same reduced state while carrying different environmental correlations and therefore different future counterfactuals. Process identity is richer than state identity.

    12. What the Mathematics Does Not Establish

    ◼
  • The calculations do not establish that any classical or quantum toy system is conscious.
    • ◼
  • The Jensen-Shannon cut divergence is a transparent diagnostic, not a derivation of IIT 4.0 intrinsic information or system Φ.
    • ◼
  • The Choi trace-distance diagnostic measures full-versus-cut channel distinguishability. It is not quantum IIT and is not a measure of phenomenal intensity.
    • ◼
  • The representation group GΓ must be physically justified. Declaring a transformation “mere redescription” can itself smuggle in an ontology.
    • ◼
  • The coarse-graining map q and intervention baseline μ are assumptions. Macro conclusions are conditional on them unless stability across admissible choices is shown.
    • ◼
  • High closure and irreducibility do not entail ontological autonomy in an absolute sense; every finite subsystem of a larger physical world may remain environmentally coupled.
    • ◼
  • Temporal persistence is measured here from cut fingerprints. Other process invariants, including genuine multitime process tensors, may provide stronger notions of persistence.
    • ◼
  • Pareto maximality prevents arbitrary scalarization but can leave several non-dominated candidates. A unique subject requires further principle or empirical constraint.
    • ◼
  • Structural monism does not by itself dissolve the explanatory gap. The phenomenal-physical identity remains a substantive metaphysical hypothesis.
    • ◼
  • Quantum mechanics is used as a conceptual and mathematical stress test. Nothing here establishes that specifically quantum coherence or entanglement is required for biological consciousness.
    • These limits are not peripheral caveats. They define the exact inferential boundary of the work. The value of the formalism lies in making that boundary sharper: one can now ask which claims follow from the process mathematics, which depend on intervention conventions, which depend on a model of composition, and which remain philosophical bridge principles.

    13. Temporal Thickness and Process Identity

    13.1 Why an instantaneous subject is unstable

    Suppose a candidate system has irreducibility I(t) that oscillates under otherwise smooth dynamics. If subjecthood is identified with the instantaneous maximizer, small phase shifts can produce rapid switches in which subset or grain counts as the subject. This is not merely a numerical inconvenience. It implies an ontology in which subjects appear and disappear at frequencies set by microscopic evolution unless an additional persistence rule is supplied.
    A temporally thick criterion instead treats a finite interval [t0,t1] as part of the individuating structure. Let Σ(t) be an invariant structural fingerprint, for example the sorted cut spectrum together with boundary and state-conditioned quantities. A process distance can compare adjacent fingerprints after quotienting out admissible symmetries. Persistence is high when the trajectory remains within a small neighborhood of an equivalence class even though the microscopic state changes substantially.

    13.2 Temporal composition is not merely repeated spatial composition

    Spatial integration asks whether parts at one time can account for the whole. Temporal integration asks whether a process over an interval can be decomposed into causally independent temporal segments. A Markov chain with no memory can still exhibit strong spatial integration at each step; a non-Markovian process can retain information across time even when instantaneous spatial coupling is weak. These are different axes of organization.
    This opens a new family of questions. Does a candidate subject require a minimum temporal depth? Can a process be temporally irreducible while spatially reducible? Can coarse-graining in time increase causal specificity just as spatial coarse-graining can? How should temporal partitions interact with IIT's exclusion principle? None of these questions is settled by a one-step Φ-like number.

    13.3 A process-equivalence criterion

    A stronger subject-candidate can now be written as an equivalence class [𝔖]GΓ over a temporally extended process 𝔖, together with a neighborhood of grains for which four conditions remain jointly nontrivial: irreducibility, closure, persistence, and representational invariance. No single weighted sum is imposed. Candidate processes are compared by Pareto dominance because improving one criterion should not automatically compensate for destroying another.
    𝒞(𝔖) = ( I(𝔖), K(𝔖), P(𝔖), Gstab(𝔖) ), compared by Pareto order.
    The Pareto choice is philosophically substantive. A weighted consciousness score would smuggle a theory of tradeoffs into arbitrary coefficients. Pareto order instead identifies candidates that are unambiguously dominated while leaving genuinely multidimensional cases unresolved. Indeterminacy is preferable to false precision when the phenomenological constraints do not fix exchange rates among unity, closure, and persistence.
    Figure 11. Candidate-subsystem Pareto geometry for the five-node classical model. The embedded static view rescales axes only for visibility; the executable plot above uses the raw irreducibility, closure, and persistence values. A frontier replaces an arbitrary weighted consciousness score.

    15. Conclusion: Intrinsic Process Structure Without Mysticism

    The strongest conclusion supported by the formal development is conditional and structural. Classical and quantum systems can be described within a shared algebra-state-dynamics-composition language. Within that language, state correlation, entanglement, causal irreducibility, closure, grain, and persistence are demonstrably distinct. Representation invariance can be made into an explicit mathematical demand rather than a philosophical slogan. Coarse-graining can be defined on induced dynamics rather than on averaged edge weights. Decoherence can continuously remove specifically quantum coherence while leaving classical correlation and, depending on the mechanism, nontrivial causal cut responses. Temporal extension reveals that subject identity cannot plausibly be reduced to an instantaneous extremum without an account of persistence.
    These results do not establish that invariant causal process structure is consciousness. They establish a much narrower point: if consciousness is to be identified with physical structure, the candidate structure must be richer than a static state, richer than entanglement, richer than generic information, and richer than one integration scalar. It must include the modal organization that distinguishes instantiation from representation, the compositional structure that makes partitions meaningful, the invariances that remove arbitrary coordinates, the boundary relations that define a perspective, and the temporal organization that lets a subject persist through change.
    The resulting metaphysical picture is compatible with a disciplined structural monism. Reality need not be conceived as self-subsistent objects to which relations are subsequently attached. A subject can instead be a stable, causally self-determining process pattern whose identity is fixed by invariants across admissible descriptions. What remains unresolved is exactly the question that should remain unresolved: why the intrinsic existence of such a process is phenomenal. That question is not answered by renaming irreducibility as consciousness. It becomes sharper because the mathematical work removes several easier confusions around it.

    References

    1. Albantakis, L., Barbosa, L., Findlay, G., Grasso, M., Haun, A. M., Marshall, W., Mayner, W. G. P., Zaeemzadeh, A., Boly, M., Juel, B. E., et al. (2023). Integrated information theory (IIT) 4.0: Formulating the properties of phenomenal existence in physical terms. PLOS Computational Biology, 19(10), e1011465. arXiv:2212.14787. https://doi.org/10.1371/journal.pcbi.1011465
    2. Oizumi, M., Albantakis, L., & Tononi, G. (2014). From the phenomenology to the mechanisms of consciousness: Integrated Information Theory 3.0. PLOS Computational Biology, 10(5), e1003588. https://doi.org/10.1371/journal.pcbi.1003588
    3. Grasso, M., Hendren, J., & Tononi, G. (2026). Consciousness as Intrinsic Structure: Towards a Chemistry of Experience. arXiv:2608.11398.
    4. Mindt, G. (2021). Not All Structure and Dynamics Are Equal. Entropy, 23(9), 1226. https://doi.org/10.3390/e23091226
    5. Nešić, J. (2022). Towards a Neutral-Structuralist Theory of Consciousness and Selfhood. International Studies in the Philosophy of Science, 35(3–4), 243–259. https://doi.org/10.1080/02698595.2022.2092825
    6. Ladyman, J., Ross, D., Spurrett, D., & Collier, J. (2007). Every Thing Must Go: Metaphysics Naturalized. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199276196.001.0001
    7. Ladyman, J. (1998). What is Structural Realism? Studies in History and Philosophy of Science Part A, 29(3), 409–424. https://doi.org/10.1016/S0039-3681(98)80129-5
    8. French, S. (2014). The Structure of the World: Metaphysics and Representation. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199684847.001.0001
    9. Newman, M. H. A. (1928). Mr. Russell’s “Causal Theory of Perception.” Mind, 37(146), 137–148. https://doi.org/10.1093/mind/XXXVII.146.137
    10. Russell, B. (1927). The Analysis of Matter. Harcourt, Brace; Kegan Paul, Trench, Trübner & Co.
    11. Chalmers, D. J. (1996). The Conscious Mind: In Search of a Fundamental Theory. Oxford University Press.
    12. Doerig, A., Schurger, A., Hess, K., & Herzog, M. H. (2019). The unfolding argument: Why IIT and other causal structure theories cannot explain consciousness. Consciousness and Cognition, 72, 49–59. https://doi.org/10.1016/j.concog.2019.04.002
    13. COGITATE Consortium, Ferrante, O., Gorska-Klimowska, U., Henin, S., Hirschhorn, R., Khalaf, A., Lepauvre, A., et al. (2025). Adversarial testing of global neuronal workspace and integrated information theories of consciousness. Nature, 642, 133–142. https://doi.org/10.1038/s41586-025-08888-1
    14. Cea, I., Negro, N., & Signorelli, C. M. (2023). The Fundamental Tension in Integrated Information Theory 4.0’s Realist Idealism. Entropy, 25(10), 1453. https://doi.org/10.3390/e25101453
    15. Albantakis, L., Prentner, R., & Durham, I. (2023). Computing the Integrated Information of a Quantum Mechanism. Entropy, 25(3), 449. https://doi.org/10.3390/e25030449
    16. Pollock, F. A., Rodríguez-Rosario, C., Frauenheim, T., Paternostro, M., & Modi, K. (2018). Non-Markovian quantum processes: Complete framework and efficient characterization. Physical Review A, 97, 012127. https://doi.org/10.1103/PhysRevA.97.012127
    17. Hoel, E. P. (2017). When the Map Is Better Than the Territory. Entropy, 19(5), 188. https://doi.org/10.3390/e19050188
    18. Esteban, F. J., Galadí, J. A., Langa, J. A., Portillo, J. R., & Soler-Toscano, F. (2018). Informational structures: A dynamical system approach for integrated information. PLOS Computational Biology, 14(9), e1006154. https://doi.org/10.1371/journal.pcbi.1006154
    19. Barrett, A. B., & Mediano, P. A. M. (2019). The Phi measure of integrated information is not well-defined for general physical systems. arXiv:1902.04321.
    20. Carvalho, A. N., Langa, J. A., & Soler-Toscano, F. (2026). On the Shape of Intrinsic Information: A Unified Framework from Dynamical Systems. Foundations of Science. https://doi.org/10.1007/s10699-026-10042-7
    21. Marshall, W., Findlay, G., Albantakis, L., & Tononi, G. (2026). Intrinsic units: identifying a system’s causal grain. Neuroscience of Consciousness, 2026(1), niag013. https://doi.org/10.1093/nc/niag013
    22. Barbosa, L. S., Marshall, W., Streipert, S., Albantakis, L., & Tononi, G. (2021). Mechanism Integrated Information. Entropy, 23(3), 362. https://doi.org/10.3390/e23030362
    23. Zanardi, P., Tomka, M., & Campos Venuti, L. (2018). Towards Quantum Integrated Information Theory. arXiv:1806.01421.
    24. Balduzzi, D., & Tononi, G. (2009). Qualia: The Geometry of Integrated Information. PLOS Computational Biology, 5(8), e1000462. https://doi.org/10.1371/journal.pcbi.1000462
    25. Tononi, G., Boly, M., Massimini, M., & Koch, C. (2016). Integrated information theory: from consciousness to its physical substrate. Nature Reviews Neuroscience, 17, 450-461. https://doi.org/10.1038/nrn.2016.44
    26. Tegmark, M. (2016). Improved Measures of Integrated Information. PLOS Computational Biology, 12(11), e1005123. https://doi.org/10.1371/journal.pcbi.1005123
    27. Marshall, W., Albantakis, L., & Tononi, G. (2018). Black-boxing and cause-effect power. PLOS Computational Biology, 14(4), e1006114. https://doi.org/10.1371/journal.pcbi.1006114
    28. White, G. A. (2022). Characterization and control of non-Markovian quantum noise. Nature Reviews Physics, 4, 287. https://doi.org/10.1038/s42254-022-00446-2
    29. Cygorek, M., Cosacchi, M., Aceituno Chávez, A., Lovett, B. W., Keeling, J., Gauger, E. M., et al. (2026). Process Tensor Approaches to Non-Markovian Quantum Dynamics. Physical Review X, 16, 020502. https://doi.org/10.1103/1ncg-11hz
    30. Pearl, J. (2009). Causality: Models, Reasoning, and Inference, 2nd ed. Cambridge University Press.
    31. Levine, J. (1983). Materialism and Qualia: The Explanatory Gap. Pacific Philosophical Quarterly, 64, 354-361.
    32. Alter, T., & Nagasawa, Y. (Eds.). (2015). Consciousness in the Physical World: Perspectives on Russellian Monism. Oxford University Press.
    33. Beni, M. D. (2019). Structuring the Self: A Philosophical Approach to the Self with Structuralism and Information Theory. Palgrave Macmillan.
    34. Chalmers, D. J. (2003). Consciousness and its Place in Nature. In S. Stich & F. Warfield (Eds.), Blackwell Guide to Philosophy of Mind. Blackwell.
    35. Ay, N., & Polani, D. (2008). Information flows in causal networks. Advances in Complex Systems, 11(1), 17-41. https://doi.org/10.1142/S0219525908001465

    CITE THIS NOTEBOOK

    Beyond ϕ : invariant causal process structures as the substrate of subjectivity​
    by Rihaan Shah​
    Wolfram Community, STAFF PICKS, September 24, 2026
    ​https://community.wolfram.com/t/28167