Comparison
Integrated Information Theory vs Global Workspace Theory
Integrated Information Theory (IIT) and Global Workspace Theory (GWT) are two of the most discussed scientific accounts of consciousness, and they start from opposite ends. Giulio Tononi built IIT from experience itself. He lists features every experience has, such as being unified and specific, and asks what kind of physical system could have them. His answer is a system with high integrated information, measured as Φ (phi), whose parts constrain each other in ways that cannot be split.
Bernard Baars proposed GWT in the 1980s, working from cognitive psychology. The brain runs many specialized processes in parallel, and information becomes conscious when it enters a limited-capacity workspace and is broadcast to all of them. Stanislas Dehaene and Jean-Pierre Changeux developed this into the Global Neuronal Workspace, in which conscious access appears as a sudden, brain-wide ‘ignition’ across prefrontal and parietal cortex.
For GWT, consciousness is what a brain does with information, so any system with the right architecture could have it. For IIT, consciousness is what a physical system intrinsically is, its cause-effect structure, so a conventional computer running brain-like software would have almost none. The theories also point to different brain regions, IIT to a posterior ‘hot zone’ and GWT to frontal and parietal networks.
Those clear differences made them natural targets for an adversarial collaboration. In the Cogitate Consortium, proponents of both theories agreed on predictions before the data came in. The results, published in Nature in 2025, contradicted key predictions of each theory and settled neither. After early results were presented in 2023, an open letter signed by more than a hundred researchers called IIT pseudoscience, a charge David Chalmers and others rejected.
Common ground. Both are scientific theories that try to say which physical systems are conscious and make predictions brain data can test. Both accept that most brain activity is unconscious and treat conscious experience as unified. Their proponents agreed to test them head to head under predictions fixed in advance.
Key differences
| Integrated Information Theory (IIT) | Baars’s and Dehaene’s Global Workspace Theory | |
|---|---|---|
| What consciousness is | Integrated information. An experience is identical to a system’s maximally irreducible cause-effect structure. | Global availability. Content becomes conscious when it is broadcast across the brain to many specialized processes. |
| Starting point | Phenomenology. Axioms about experience are translated into postulates about physical systems. | Cognitive psychology and neuroscience, especially lab contrasts between conscious and unconscious processing. |
| Where in the brain | Mainly posterior cortex, where IIT expects sustained, highly integrated activity. | A frontal and parietal network that ‘ignites’ when a stimulus reaches awareness. |
| Main objection | Counterintuitive predictions, such as consciousness in simple but highly integrated grids, and a Φ measure impractical to compute for real brains. | It explains access and reportability, and critics say it leaves open why broadcast information feels like anything. |
| AI consciousness | Conventional digital computers have negligible Φ, so they would not be conscious however intelligent they seem. | Possible in principle. A machine with a global workspace architecture could meet the theory’s conditions. |
On the big questions
| Integrated Information Theory (IIT) | Baars’s and Dehaene’s Global Workspace Theory | |
|---|---|---|
| AI consciousness | NoKoch, a proponent, argues that consciousness is not computable. Digital computers have negligible integrated information (Φ(Phi)) due to hardware… | YesSince the theory defines consciousness in terms of information-processing computations and architectures, and proponents suggest novel machine architectures… |
| Life after death | UncertainIf phenomenal consciousness requires a radically new structure (like IIT's postulates), the potential permanence of these 'inscrutable things' cannot be… | NoConsciousness is understood as being entirely dependent on specific, complex biological brain processes, which cease upon biological death. |
| Mind uploading | NoIf consciousness is a fundamental property tied to real cause-effect power, a digital simulation (uploading) would only create a 'perfect zombie'—an entity… | YesIf consciousness is wholly explained by the computational organization and information flow of the brain, a perfect duplication or uploading of these… |
| Meaning and purpose | UncertainWhile the measurement and analysis of consciousness do not convey ultimate meaning, the speculative nature and novelty of the innumerable nth dimensional… | Humanistic onlyAs a Materialist/Neurobiological theory, it does not mandate ultimate cosmic meaning or purpose, but consciousness provides essential functions (like memory… |
Integrated Information Theory (IIT)
IIT is an original, radical model that proposes a definition for conscious experience and specifies which physical systems can possess it, starting from the phenomenology of consciousness (axioms). It translates these axioms into physical postulates based on intrinsic cause-effect power, featuring a mathematical measure Φ(Phi) to quantify a system's…
Baars’s and Dehaene’s Global Workspace Theory
This theory posits that conscious awareness is constituted by specific multi-sensory, multi-cognitive information that achieves brain-wide presence and broad accessibility. It conceptualizes consciousness as an inner 'theater' where this information is globally broadcast, reconciling the limited capacity of momentary conscious content with vast long-term…