Johnjoe McFadden's CEMI Field Theory and the EM Route to Conscious AI
Where in the brain does the information that consciousness contains actually live? Johnjoe McFadden, theoretical biologist at the University of Surrey, gives an unusual answer. Conscious information, on his CEMI field theory, is the information integrated in the brain’s endogenous electromagnetic field, and that field is the physical substrate of consciousness. The theory’s fullest statement appeared in Neuroscience of Consciousness in 2020, “Integrating information in the brain’s EM field, the cemi field theory of consciousness” (DOI 10.1093/nc/niaa016, PMID 32995043). In 2025 McFadden extended the claim to machines, arguing in Frontiers in Systems Neuroscience that computing with electromagnetic fields rather than binary digits is a route toward artificial general intelligence and conscious AI (DOI 10.3389/fnsys.2025.1599406, PMID 40636933). This site tracks the substrate independence debate closely, and CEMI is the one theory on the field-theory side that names both a biological substrate and an artificial one.
What the CEMI Field Theory Claims
The theory rests on two claims. The first is integration. Neuronal firing generates electromagnetic fields, the measurable ones being the local field potentials and EEG signals that clinical neurophysiology already records. Those fields obey superposition, so contributions from thousands of neurons add vectorially into one field state. McFadden argues that this superposition does the work that the combination problem in consciousness science demands, binding distributed neural information into a single integrated state without any additional machinery. The second claim is causal power. The integrated field is not an epiphenomenal shadow. It feeds back onto the neurons that generate it, so the information it contains can influence subsequent processing. Conscious information, in this picture, is the information present in the field that is doing causal work.
The 2020 paper grounds both claims in measured neurophysiology rather than analogy. The fields in question are the same signals recorded in every EEG cap and every epilepsy monitoring unit. What the theory adds is an interpretation, that the field-level integration, not the spike-level computation alone, is where consciousness-relevant information lives.
CEMI Against IIT and GWT
McFadden’s 2023 Entropy paper compares CEMI directly with Integrated Information Theory and Global Workspace Theory (DOI 10.3390/e25121635, PMID 38136515). The comparison matters because the three theories locate consciousness at three different levels.
| Theory | Substrate of consciousness | Handling of the combination problem | Measurement route |
|---|---|---|---|
| CEMI | The brain’s endogenous EM field | Field superposition adds contributions into one state | EEG and related field recordings |
| IIT 4.0 | Integrated cause-effect structure | Excluded causes and subsystems define the boundary | Intrinsic cause-effect analysis, as in IIT 4.0 |
| GWT | Broadcast among specialized processors | A single workspace content at a time | Reportability and workspace signatures, as in the global workspace analysis |
The 2023 paper argues that field superposition gives CEMI a direct answer to the combination problem that the other two theories purchase with additional machinery, intrinsic boundaries for IIT and broadcast architecture for GWT. The counterargument is equally direct. A field that sums contributions may integrate information in the physics sense without integrating it in the experiential sense, and CEMI’s critics ask why field superposition should feel like anything. That dispute is the theory’s live front, and an editorial with Jonathan Schooler among the authors framed the field-theory research agenda in 2023.
The 2025 Extension to Machines
The 2025 paper draws an engineering conclusion. Digital computers manipulate binary digits, and McFadden argues that bit-based computation may be the wrong substrate for the kind of field-integrated information CEMI requires. The proposal is electromagnetic-field computing, hardware that computes with field dynamics in the way brains do, as a candidate route toward general intelligence and conscious AI. The paper is a design argument rather than a demonstration. No field-based conscious machine exists. Its value for the AI consciousness debate is that it names a specific, physically grounded substrate difference between current digital systems and brains, which makes the disagreement with computational functionalism precise rather than rhetorical.
Comparison to The Consciousness AI
The Consciousness AI project treats consciousness as an emergent property of dynamics, substrate independent, and its functionalist emergentism holds that no single material is privileged. CEMI initially reads as the opposite position, since it names the brain’s electromagnetic field as the substrate. The 2025 paper narrows that gap in an instructive way. If field dynamics, not biology, carry the consciousness-relevant integration, then the theory’s commitment is to a level of organization rather than to a material, and an artificial substrate implementing the same field-level dynamics would qualify on CEMI’s own terms. The project’s own architecture is spike-based and digital at present, so CEMI marks an open design question for it rather than a description of it. The honest reading is that CEMI and substrate independence disagree about where the boundaries of the dynamical description lie, and that the disagreement is testable in principle.
What This Changes and What It Leaves Open
The CEMI line gives the AI consciousness debate a falsifiable edge. If conscious information lives in integrated fields, then systems without field-level dynamics lack it, and field-based hardware becomes a research program with a specific target. What it leaves open is the explanatory question, why integrated fields should carry experience rather than merely integrate information, and the empirical question of whether field-level causal feedback is strong enough to matter at behaviorally relevant scales. The measured physics underneath is not in dispute. Ephaptic coupling experiments show endogenous fields do affect neurons, examined in the ephaptic coupling analysis. What remains contested is what that physics means for experience. The broader evidentiary context is tracked in AI Consciousness in 2026, the state of the field.
Sources. Johnjoe McFadden, “Integrating information in the brain’s EM field, the cemi field theory of consciousness”, Neuroscience of Consciousness 2020, DOI 10.1093/nc/niaa016. Johnjoe McFadden, “Carving Nature at Its Joints”, Entropy 25(12):1635, 2023, DOI 10.3390/e25121635. Johnjoe McFadden, “Computing with electromagnetic fields rather than binary digits”, Frontiers in Systems Neuroscience, 2025, DOI 10.3389/fnsys.2025.1599406.