An Intriguing and Controversial Theory of Consciousness: IIT

Source: DPS studio / AdobeStock

This is Part 4 of a five-part series.

Integrated information theory (IIT) is a theory of consciousness. It was first formulated by the neuroscientist and psychiatrist Giulio Tononi in the early 2000s and further developed in collaboration with the neuroscientist Christof Koch.¹

IIT suggests that consciousness arises from the integration of information in a system. A system is considered conscious if it possesses a high degree of differentiation and integration.

IIT differs in a number of ways from higher-order theories (HOTs) and global workspace theories (GWTs). In contrast to theories that see consciousness as an emergent property from complex computations or neural activities, IIT posits that consciousness exists intrinsically and attempts to define it from that perspective—from the system’s own “point of view,” it is about “intrinsic information”—the information a system has about itself, for itself.^2,3

Tononi explains some of the guiding principles of IIT:⁴

• It starts from phenomenology, not behavioral, functional, or neural correlates.
• It identifies the essential phenomenal properties that are true of every conceivable experience—the essential properties that characterize every experience. These are the axioms of IIT.
• It translates those essential phenomenal properties (the axioms) into essential physical properties (properties of physical existence, having cause-effect power) that physical systems must satisfy to account for the presence of experience. These essential physical properties are the postulates of IIT.
• It aims at accounting for all phenomenal properties in terms of physical properties, including the specific properties of specific experiences (properties of spatial extendedness, temporal flow, objects, colour, sounds, etc.)
• It has explanatory, predictive, and inferential power.

Axioms and Postulates of IIT

IIT proposes five axioms and five corresponding postulates (for clarity, the following is a simplified paraphrased version⁵).

Axioms of IIT: These are asserted as self-evident essential properties of every conscious experience—phenomenological truths about consciousness.

1. Intrinsic Existence: Consciousness exists for the system itself, not for an external observer; i.e., consciousness exists in a real and personal way for you (if you’re the system). Its real intrinsic existence is demonstrated by its cause-effect power on itself.
2. Composition: Consciousness is structured, consisting of multiple distinct but interrelated elements (each with causal power); i.e., conscious experiences can be made up of multiple parts, like how a movie scene has images, sounds, and feelings all at once.
3. Information: Conscious experiences are specific, meaning that among all the possible experiences you could have, you are having a very particular one; i.e., every conscious moment is unique (differentiated)—like watching a specific scene in a movie out of all possible scenes.
4. Integration: Conscious experiences are unified and cannot be reduced into independent components; i.e., any single conscious experience you have can't be split in half such that one part is independent of the other.
5. Exclusion: At any given time, there's a particular experience that excludes others. And it occurs at a particular scale of time and space; i.e., at any moment, you're conscious of certain things and not others, and there's a specific "size" or "scale" to what you're conscious of—not too detailed, not too vague.

Postulates of IIT: These are the proposed mechanisms that physical systems must satisfy to have the above properties of consciousness.

1. Intrinsic Existence: A physical system has intrinsic cause-effect power. This means it can have an effect on itself and is not just influenced by external factors.
2. Composition: Consciousness is not a single undifferentiated whole but is instead structured and composed of a set of interrelated experiences that together form the overall conscious experience.
3. Information: A system's current state must specify a particular cause-effect structure, differing in its specific way from other possible structures (differentiation); e.g., the experience of seeing a red apple is distinct from smelling a rose, and this distinction carries specific information that differentiates the two experiences.
4. Integration: The cause-effect power of a system must be unified, meaning it cannot be divided into independent components (irreducibility); i.e., the system works as a united whole—it must unite different parts in a way that they cannot be separated into independent parts.
5. Exclusion: Out of all the possible overlapping sets of elements that could contribute to a conscious experience, there is a single set that is maximally irreducible—meaning it cannot be divided into independent parts without losing its essential properties. This set is what specifies the conscious experience to the exclusion of all other sets.

Phi

In contrast to most other theories of consciousness, IIT is a mathematical theory. To quantify the degree of consciousness of any system (not just a biological nervous system), IIT introduces the concept of Φ (phi). Phi measures the amount of integrated information in a system (more specifically, a maximum of irreducible, intrinsic cause-effect power). If a system has a high value of phi, it is considered to be highly integrated and, therefore, has a high degree of consciousness. On the other hand, if phi is zero, the system is not conscious.

IIT in the Brain

IIT is framed at an abstract mathematical level and does not easily capture specific cognitive neuroscience phenomena. Furthermore, IIT’s Φ cannot be easily computed from large-scale data, making it difficult to test the theory.⁶ It can only be measured in simple model systems⁷—it cannot come even remotely close to measuring the integrated information in a brain.⁸

Anatomically, in contrast to HOTs and GWTs, IIT primarily links consciousness with posterior regions of the cortex, referred to as the “posterior hot zone,” which includes the parietal, temporal, and occipital areas. This is partly based on the argument that these regions display neuroanatomical properties that are thought to be well-suited for producing high levels of integrated information.¹⁰ The prefrontal cortex, according to IIT, is not essential for the conscious experience itself but instead contributes to post-conscious cognitive processing, such as planning or verbal reports.^10,11

Does IIT Explain Qualia?

IIT claims that the specific content of a conscious experience (what it is like to have that experience, or its "qualia") is determined by, and indeed identical to, the configuration or geometric "shape" of the integrated information—the informational relationships within the conceptual structure.¹² In other words, the particular way in which information is integrated and organized within the structure determines the subjective quality of the experience.

Consciousness Essential Reads

Attention as a Criterion of Free Will

What Does It Actually Mean to Be Conscious?

Panpsychism

Controversially, IIT implies a form of panpsychism, the idea that consciousness is a fundamental property of the universe, similar to space, time, and matter.

IIT posits that any system that generates a non-zero maximum of irreducible integrated information has some degree of consciousness. This would suggest that there may be many non-biological systems that possess consciousness.¹³ According to IIT, even simple systems can have some minimal form of consciousness, provided they have a certain degree of information integration: “Even circuits as simple as a ‘photodiode’ made up of a sensor and a memory element can have a modicum of experience."¹⁴

The reasons why IIT leads to such conclusions—so outrageously implausible in the view of critics—can be traced, according to technical critiques, to some of its starting assumptions, including the ways in which Phi is defined mathematically. Among other problems, integrated information might be a necessary but not sufficient criterion for consciousness.¹⁵

On the other hand, perhaps the claim that consciousness is a fundamental property of the universe explains part of the alluring appeal IIT has enjoyed with the media and the public. It seems to flatter us that our most cherished property is indeed special after all, reversing the repeated narcissistic injuries that scientific materialism has inflicted upon us in telling us that we are accidental by-products in an indifferent, purely physical universe.

Harsh Criticism by Fellow Scientists

In part 3, I discussed the mixed results of the adversarial collaborative study comparing IIT and GWT, announced in June 2023, and the debatable interpretation that the results favored IIT. I mentioned that the media attention resulting from that interpretation prompted a controversial letter in September 2023 signed by a large number of prominent consciousness scientists calling IIT “pseudoscience.” This is a particularly severe label to assign a rival theory, one that made other critics of IIT, who had not signed the letter, decidedly uncomfortable.¹⁶

Value of IIT: What It Adds to Our Understanding of Consciousness

While the property of integrated information may not be sufficient for consciousness, it is undoubtedly a crucial ingredient. And while some of its assumptions may be incorrect, IIT has made a significant contribution in being the first substantial attempt at defining and attempting to measure integrated information. It has also rightly emphasized that subjective phenomenology cannot be ignored.

It has ambitiously attempted to define a theory of consciousness from the intrinsic perspective of the information system itself, from the inside-out—an approach that complements the more traditional extrinsic perspective—analyzing and measuring the system from the outside in.

IIT is a bold attempt, but its claim of being “the only really promising fundamental theory of consciousness"¹⁷ now seems a regrettable and premature overstatement.

Part 1 of this series provided an overview of the leading theories of consciousness. Part 5 concludes the series by examining re-entry and predictive processing theories, and reconsidering the overall approach to the "hard problem" of consciousness.