What Is Integrated Information Theory?

Direct Answer

Integrated information theory (IIT) is a scientific and philosophical theory of consciousness developed by the neuroscientist and psychiatrist Giulio Tononi. Its central claim is that consciousness is integrated information: a physical system is conscious if and only if it is a maximally integrated causal whole, and the quantity of its consciousness is measured by a number called phi (Φ). The more a system's parts constrain one another — the more the whole is irreducible to its parts — the higher its phi and the richer its experience.

The theory starts from phenomenology: consciousness has essential properties that any theory must account for. Experience is structured (there is a specific shape to every experience), it is specific (each experience is this one, not that one), it is unitary (all aspects of an experience occur together, for one subject), and it is definite (there is always a particular experience, with a particular content). From these phenomenal axioms, IIT derives corresponding postulates about physical systems: a system must have mechanisms that generate information (differences that make a difference), and these must be integrated (irreducible to independent parts) to produce a unified whole. The whole is a "complex" of mechanisms, and its phi measures the irreducibility of its cause-effect structure.

IIT is distinctive in treating consciousness as a graded property of any system with the right causal organization. It is not restricted to brains: any physical system — biological or artificial — that instantiates the required causal structure has consciousness to the degree its phi dictates. This yields famous consequences: a digital computer that merely implements computations (like a simulated mind, or a "zombie" architected as a feedforward network) may have zero or near-zero phi and hence no consciousness, because its parts are not causally integrated in the required way; whereas some simple biological structures may have surprisingly high phi. The theory also identifies a "postulate of exclusion": a system's consciousness is that of its maximally integrated complex, which solves the binding problem by stipulating that experience belongs to the whole, not the parts.

Historical Context

IIT emerged from Tononi's work on the neuroscience of consciousness in the 1990s and 2000s. Earlier work with Gerald Edelman had developed the "dynamic core" hypothesis: conscious states are characterized by a specific kind of distributed, integrated neural activity. Tononi formalized this intuition: if consciousness requires both information (the system distinguishes many states) and integration (the parts hang together), then consciousness can be measured as the amount of integrated information. The formal measure phi was introduced in 2004, and the theory was presented in its canonical form in Tononi's 2008 manifesto "Consciousness as Integrated Information" and his 2012 book Phi: A Voyage from the Brain to the Soul.

The theory gained prominence partly through its relationship to the neuroscience establishment: Christof Koch, a leading consciousness researcher and longtime collaborator of Francis Crick, became a vocal champion of IIT, and the two co-authored the 2015 paper in which Koch publicly abandoned his earlier "neural correlates" agenda for IIT. The theory also connects to a long philosophical lineage: its claim that consciousness comes in degrees and is tied to the causal structure of complex systems resonates with panpsychism and with the Russellian monist idea that intrinsic properties of physical systems may be experiential. Tononi himself emphasizes the "pyramids" of consciousness — the idea that a simple system such as a photodiode could be conscious — which has made IIT a flashpoint in the debate between panpsychists, physicalists, and functionalists.

Key Arguments & Debates

The argument for IIT is that it is a theory in the strong sense: it starts from indubitable facts about experience, derives constraints on physical systems, provides a quantitative measure, makes testable predictions, and offers explanations of known phenomena. It explains why consciousness correlates with certain neural states (those with high phi), why it is graded (phi is graded), why it is unified (the whole complex is the unit), and why feedforward or purely computational systems are not conscious (their causal structure lacks integration). Its proponents argue that it is the only theory that has successfully derived the phenomenology from the physics.

The objections are numerous and intense. The "phi" measure is enormously difficult to compute for real systems, so the theory's quantitative predictions have been tested only in toy cases. The claim that a photodiode or grid of logic gates could be conscious strikes many as absurd, and critics argue that IIT's axioms are question-begging. The theory has been criticized for being unfalsifiable in practice and for making predictions that conflict with neuroscience (e.g., concerning the cerebellum, which has enormous integration capacity but seems not to support consciousness — a problem Tononi's defenders dispute). Philosophers note that IIT identifies consciousness with a quantity without explaining why integration should give rise to experience — the hard problem appears to remain, merely relocated into the definition of phi. Finally, the "combination problem" that plagues panpsychism recurs: how do the phi-complexes of parts combine into the single consciousness of a person? IIT's answer (the exclusion postulate) is itself contested.

Contemporary Relevance

IIT is one of the "big three" scientific theories of consciousness (with global workspace theory and predictive processing), and it has driven a major research program: attempts to measure phi in the human brain using transcranial magnetic stimulation and EEG (the "perturbational complexity index"), clinical applications in diagnosing disorders of consciousness, and theoretical work on the nature of consciousness in artificial systems. The theory has direct implications for AI: if consciousness requires causal integration of a particular kind, then many current AI systems — trained feedforward networks, transformer models with attention — may lack it entirely, and "machine consciousness" would require building systems with the right causal architecture, not just the right behavior. This has made IIT central to debates about AI ethics and the possibility of machine suffering.

IIT also anchors contemporary philosophical debates: its panpsychist-flavored conclusions are cited by proponents and opponents of panpsychism alike; its status as a "scientific theory" is contested by philosophers who argue it conflates measurement with explanation; and its implications for the hard problem — whether IIT dissolves it or inherits it — remain unresolved. Whatever its fate, IIT has set the agenda: a theory of consciousness must now say something precise about integration, information, and the causal structure of the systems that experience.

Related Concepts

• Phenomenal consciousness: what IIT aims to explain and quantify.
• Global workspace theory: its main scientific rival.
• Panpsychism: the philosophical view IIT's "consciousness everywhere in degree" conclusion resembles.
• The hard problem vs easy problems: whether IIT addresses the hard problem.
• The binding problem: which IIT claims to solve via integration and exclusion.

Further Learning

• Internet Encyclopedia of Philosophy: Integrated Information Theory of Consciousness
• Stanford Encyclopedia of Philosophy: Consciousness
• Giulio Tononi, "Consciousness as Integrated Information: A Provisional Manifesto," Biological Bulletin 215 (2008).
• Giulio Tononi and Christof Koch, "Consciousness: Here, There and Everywhere?," Philosophical Transactions of the Royal Society B 370 (2015).