Matteo Grasso and Consciousness as Intrinsic Structure
Integrated Information Theory has long claimed to explain not just how much consciousness a system has but what it is like, the quality of experience, and critics have pointed out that this second half of the claim has stayed comparatively abstract. Matteo Grasso, Jeremiah Hendren, and Giulio Tononi’s August 2026 preprint, “Consciousness as Intrinsic Structure: Towards a Chemistry of Experience” (arXiv:2608.11398), takes a direct step toward closing that gap, proposing that qualia should be treated as intrinsic structures with the same kind of systematic, combinatorial organization chemistry uses to relate molecules to their properties.
| Chemistry analogy | Chemical concept | Proposed consciousness analog |
|---|---|---|
| Basic building blocks | Atoms and their bonding rules | Elementary distinctions and the causal relations between them |
| Combinatorial structure | Molecules built from atoms in specific configurations | Cause-effect structures built from distinctions in specific configurations |
| Predicted properties | A molecule’s structure determines its chemical properties | A cause-effect structure’s shape determines the quality of experience it specifies |
Structure as the Missing Link Between Phi and Quality
Integrated Information Theory’s central quantity, phi, measures how much a system’s causal structure is irreducible to the sum of its parts, and has functioned mainly as a scalar answer to how much consciousness a system has. IIT’s fuller claim, that the shape of a system’s cause-effect structure, not just its magnitude, specifies the actual quality of the experience, what it is like rather than merely how much of it there is, has been harder to state in a way that yields specific, checkable predictions. Grasso, Hendren, and Tononi’s proposal is to treat that shape as an intrinsic structure in a formal sense borrowed from mathematics, a network of distinctions, relations between elements of a system that make a causal difference, and relations, how those distinctions combine and constrain one another.
The chemistry analogy is meant literally rather than decoratively. Just as a molecule’s specific arrangement of atoms and bonds determines its chemical properties, in a systematic, predictable way that lets chemists infer properties from structure without measuring each molecule from scratch, the paper proposes that a cause-effect structure’s specific arrangement of distinctions and relations should systematically determine the quality of the experience it specifies, in principle allowing researchers to infer qualitative properties of an experience from the structure’s shape rather than only from a system’s raw phi value.
What a Chemistry-Like Theory Would Need to Predict
For the analogy to do real theoretical work rather than remain a metaphor, the framework needs to specify rules connecting structural features to qualitative properties with the same kind of systematicity chemistry’s periodic table and bonding rules provide, predicting that a certain kind of structural motif corresponds to, for instance, the phenomenal character of a particular color relation or the specific quality of a spatial experience. The preprint develops this programme in the mathematical language of category theory and structural composition rather than claiming the full predictive apparatus is complete, presenting the chemistry framing as a research programme for how such predictions could eventually be derived rather than a finished table of correspondences.
This connects directly to the inference problem Kallum Robinson, Tononi, Naotsugu Tsuchiya, and Grasso formalize in a companion August 2026 preprint, covered in Naotsugu Tsuchiya on the Rosetta Stone problem for another mind’s experience, since establishing a principled correspondence between two systems’ experiences, in that paper’s category-theoretic sense, would be far more tractable if each system’s cause-effect structure could first be decomposed into the kind of well-behaved intrinsic structure this preprint proposes.
Where This Fits Against Field Formulations and Decomposition Methods
Giulio Tononi’s parallel work reformulating IIT’s mathematics for continuous rather than discrete systems, covered in Giulio Tononi on IIT field formulation for continuous AI architectures, and Pedro Mediano’s methods for decomposing integrated information into synergistic and redundant components, covered in Pedro Mediano on integrated information decomposition and synergy, both supply mathematical machinery the intrinsic structure programme would likely draw on, since decomposing a cause-effect structure into interpretable substructures is a prerequisite for mapping structural features onto specific qualitative predictions. Melanie Boly’s empirical work relating specific IIT cause-effect structures to the phenomenology of meditative states, covered in Melanie Boly on IIT and the neural correlates of pure presence, is closer to the kind of structure-to-quality correspondence this preprint’s programme would eventually need to generate predictions for, and test against.
Consequences for AI Systems Specifically
If the chemistry-of-experience programme succeeds even partially, the implication for artificial systems is that a nonzero phi value would not by itself say anything about the quality of whatever minimal experience a system might have, only its magnitude. Two systems with matching phi but differently shaped cause-effect structures would, on this account, be predicted to have qualitatively different experiences, or in the more likely case for current AI architectures, structures too impoverished or too unlike any biologically studied structure for the chemistry analogy’s rules, once developed, to generate a meaningful prediction at all. This raises the evidentiary bar for any claim about what an AI system’s experience would be like even in the hypothetical case where its phi were nonzero, since magnitude alone would no longer be treated as informative about content.
Comparison to The Consciousness AI
The IIT phi computation in this project’s Global Workspace layer, described on the architecture page, currently reports a scalar value derived from causal gate states rather than a decomposed cause-effect structure of the kind the chemistry-of-experience programme analyzes. No structural decomposition of the architecture’s cause-effect relations, distinctions and the relations between them in the paper’s sense, has been attempted, which means the project currently has no basis for saying anything about what quality of experience, if any, its measured phi values would correspond to, only a magnitude. Extending the existing Phi/EI measurement pipeline to output an intrinsic structure rather than a scalar would be a substantial, currently unplanned addition, and the preprint’s own programme is not yet complete enough to specify what such an extension would need to predict.
A Research Programme, Not Yet a Finished Theory
Grasso, Hendren, and Tononi present the chemistry analogy as a direction for future work rather than a settled correspondence between structural features and specific qualia, and the preprint is explicit that deriving the detailed bonding-rule equivalent for consciousness, the systematic mapping from structural motif to phenomenal quality, remains substantially undone. What the paper establishes is the formal target such a mapping would need to hit and the mathematical vocabulary, distinctions, relations, intrinsic structure, in which it would need to be stated, which is a meaningfully more precise place to start than treating phi’s qualitative claims as a promissory note. The broader landscape of competing accounts of what determines the character of experience is indexed on the state of the field on AI consciousness.