Why Information and Integration are not enough? A constitutive critique to computational models

Integrated Information Theory (IIT) is arguably the most formally developed theory of consciousness available today. Proposed by Giulio Tononi, it defines consciousness as integrated information, quantified as Φ (phi), and holds that a system is conscious to the degree that its parts are informationally integrated into an irreducible whole. IIT is only one example of a broader class of computational models built on information and integration, but it is the most explicit in treating integration as foundational.

The following argues that integration is not sufficient to serve as the foundation of consciousness, and that a prior condition - cohesion - is required: the condition from which differentiation, and implicitly information, arise.

Contrast Presupposes Relation

Contrast is typically treated as the basis of differentiation. I argue that the moment a contrast is invoked, a minimal relational condition is already implied. For a contrast to appear, A and B must be jointly accessible within a common frame. Without such co-presence, no contrast can exist as difference. Therefore, differentiation does not precede relation but presupposes it: to differ is always to differ from something.

This is not a minor technicality. It concerns what is logically required before information can exist at all. Information, in any formal or intuitive sense, requires that states be distinguishable. Distinguishability requires contrast. And contrast requires that the elements in question are already given together within a condition that allows comparison. If they are not held together, if there is no common frame, then no contrast can appear, not even in the abstract.

Integration Cannot Supply Its Own Preconditions

For a relation to hold between elements, those elements must already be given under a common condition in which the relation can apply. Without this, no relation can be established. This condition cannot be supplied by integration, because integration already presupposes elements that are both related and distinguishable. Integration therefore presupposes the very condition it would need to explain.

Integration operates on elements that are already differentiated and already in relation. It binds what is already given as distinct and co-present. But it does not, and cannot, account for why those elements are co-present in the first place. To integrate is to unify across differences, but differences must already exist, and they must already be held within a common condition, for integration to have anything to work with.

Cohesion as the Missing Condition

A prior condition is required - cohesion - understood as that which allows multiple elements to exist under a common condition, making relation, differentiation, and information possible. Relation is cohesive in its operation, but presupposes cohesion as its condition.

This is not a mystical or vague addition. Cohesion names something precise: the condition under which elements are jointly accessible. Without it, there is no common frame for relation. Without relation, there is no differentiation. Without differentiation, there is no information. The chain of constitutive dependence is clear, and at no point can integration step in to supply what it already requires.

The Frame and the Water

Consider a frame and a body of water. To differentiate them - solid vs fluid, fixed boundaries vs adaptable boundaries - one must already hold them in relation. Now, how does one integrate frame with water? Let's throw the frame in the water. Both are internally integrated and stable when taken separately, but a plastic frame and a wooden frame interact differently with water. A plastic frame resists penetration and maintains its form. A wooden frame is porous: water penetrates it and gradually disrupts the relations between its components.

Despite each being integrated on its own, the outcome of their integration is constrained by the character of the difference between their components - by the relation that precedes integration. The specific way integration unfolds between two systems is not determined by integration itself, but by the prior relational and differential character of the elements involved. Integration does not dictate the terms of its own operation. It inherits them.

The Scope of the Objection

This objection is not limited to IIT specifically. It applies to any framework that treats information as its primitive or foundational concept. If information requires differentiation, and differentiation requires relation, and relation requires a prior condition of co-presence that neither information nor integration can supply, then no information-first ontology can be self-grounding. IIT is the most explicit target because it is the most formally developed, but the structural point holds generally: information cannot be the starting point of an account of consciousness, because information already presupposes conditions it does not contain.

IIT as Explanatory Theory, Not Merely a Measure

If IIT presented itself only as a measure - a way of quantifying consciousness in systems that already exist - the objection that it does not explain how those systems came to be would be real but arguably beside the point. A thermometer does not explain how heat arises.

However, IIT does not present itself merely as a measure. It presents itself as an explanatory theory of what consciousness is. Its five axioms - existence, composition, information, integration, exclusion - are intended to specify the essential properties of experience, from which the postulates of the theory are derived. IIT claims to identify the fundamental nature of consciousness, not just to measure it.

If integration presupposes conditions it cannot supply - cohesion and differentiation - then integration cannot be the foundational property of consciousness, even if it is an essential one. IIT correctly identifies integration as essential but I argue it misidentifies it as foundational. The theory begins with an already integrated and established system, but it does not account for the conditions that make such a system possible.

Co-dependence and Asymmetry

The chain of constitutive dependence runs: without cohesion, no common condition for relation; without relation, no differentiation, and implicitly no information; integration presupposes all of this and cannot supply it.

But I do not argue that cohesion is simply prior and self-sufficient. Cohesion without differentiation is equally incoherent: a condition under which elements are held together, but in which nothing is distinguishable, is not a condition under which anything can be registered, not even cohesion itself. A pure, undifferentiated unity would have no internal structure, no contrast, no means by which its own holding-together could appear as anything at all. It would be indistinguishable from nothing.

I argue that cohesion and differentiation are co-dependent. Neither can stand without the other. Cohesion without differentiation collapses into featureless indistinction. Differentiation without cohesion scatters into unrelated fragments that cannot constitute anything. Integration is their process and dynamic, the ongoing maintenance of their relation, with integration being constitutively dependent on cohesion and differentiation in a way they are not dependent on it.

However, co-dependence does not mean symmetry. There is an asymmetry between cohesion and differentiation that the co-dependence framing, taken alone, would obscure. Differentiation occurs within cohesion, but cohesion does not occur within differentiation. Differentiation is always differentiation of something that is already held together. One cannot differentiate a scattering, one can only differentiate what is already co-present. Cohesion is the condition within which differentiation operates. Differentiation is not the condition within which cohesion operates.

The failure modes confirm this asymmetry. Differentiation without cohesion does not simply fail to exist, it degrades into: fragmentation, scattering, elements that are distinct but unrelated, present but unnavigable. Cohesion without differentiation, by contrast, does not degrade into a corresponding state, it simply cannot appear at all, because there is nothing within it to register its own holding-together. Fragmentation is the loss of cohesion while differentiation remains partially operative. There is no corresponding condition in which cohesion is operative and differentiation is entirely absent, because such a state would be experientially invisible, indistinguishable from nothing.

This means cohesion and differentiation are co-dependent for existence, but asymmetric in function. Cohesion provides the ground; differentiation is what appears within that ground. One provides the condition, the other operates within it. Co-dependence and asymmetry are not contradictory. Two things can require each other while having different structural roles, just as a magnetic field and an electric field are mutually dependent in electromagnetic propagation, yet are not interchangeable in what they do.

This is not a temporal claim. Cohesion does not exist first in time, followed by differentiation, followed by integration. The priority is constitutive: it concerns what is structurally required for what, not what came first chronologically. All three are always co-present in any actual system. But their constitutive dependencies are asymmetric: integration requires cohesion and differentiation; cohesion and differentiation require each other; and within that co-dependence, cohesion has the character of ground while differentiation has the character of articulation within that ground.

What This Means

I argue that any theory of consciousness built on information and integration alone is structurally incomplete. Such theories correctly identify essential features of conscious systems, that they are informationally rich and that their parts are unified, but I argue they mistake these features for the foundation. The foundation lies in the co-dependence of cohesion and differentiation, with integration as the process that maintains their relation.

This does not make IIT wrong. It makes it incomplete in a specific and identifiable way. The correction is not to discard integration but to recognise that it cannot ground itself, and that the conditions it presupposes - cohesion and differentiation - must be included as co-foundational in any account of consciousness that aspires to be explanatory rather than merely descriptive.