Grasping the mind: The explanatory power of Conifold Theory

Consciousness has been postulated to be the total amount of information, integrated across an entire neural network. How does consciousness happen?

There is a great mystery at the center of consciousness studies. We collect data about our external reality through our individual senses, and then we process these data in separate regions of the brain – yet we experience a bound, cohesive perception of reality. This mysterious discrepancy has been labeled ‘the binding problem’ [1, 2].

To generate a cohesive, qualitative perceptual experience, information must be bound together, or integrated, across the neural network. Indeed, one of the leading theories in the field posits that consciousness is the total amount of information held by the brain, integrated together. This concept is called Integrated Information Theory, or IIT [3, 4].

One of the leading theories in the field posits that consciousness is the total amount of information held by the brain, integrated together.

IIT posits that the total amount of information held by a neural network is proportional to the total amount of consciousness held by that neural network [5]. The amount of consciousness is then measured in terms of phi; any natural or synthetic entity with computing power can potentially be conceived to have some non-zero level of phi [6]. Phi is a useful notion for considering the structural and functional requirements for perceptual experience and self-awareness - at least it provides a starting point [7]! As a result, phi is a key central concept of Integrated Information Theory.

The axioms of Integrated Information Theory

To expand this framework, the proponents of IIT put forward several axioms to explain consciousness. Firstly, they posit that consciousness is real – that we each have a personal, private experience of thought. Secondly, proponents of IIT argue that consciousness is exclusive to the entity experiencing it. Thirdly, consciousness is understood to be a conceptual structure, composed of representational objects and events that form cohesive qualitative experiences. As such, it is distinguishable from other experiences, which contain other objects and other events.

Fourthly, proponents of IIT contend that momentary consciousness (signified by phi) is irreducible to its component parts, as individual aspects of the observed scene are unified to create a single cohesive experience. And finally, proponents of IIT assert that the nature of qualitative experience gives rise to the cause-effect structure of our perception of reality. In summary, this theory postulates that consciousness has several key characteristics: it is real, exclusive, distinctive, irreducible, and associated with the perception of cause and effect.

It is hard to argue these points. Certainly neuroscience offers no counter-evidence. However, there are several major criticisms of IIT. Firstly, the axioms provided are ‘self-evident’, rather than proven from first principles. As a result, they may provide a descriptive account of consciousness, but no explanation of how these criteria come to be. Secondly, while the principles of existence, exclusivity, irreducibility, the establishment of causal relationships within an information set, the capacity to differentiate between objectively different phenomena, and the capacity to integrate temporally-, spatially-, or categorically-linked subjects are necessary conditions for consciousness, these processes are not sufficient to produce consciousness. There must be physical parameters that facilitate the manifestation of perceptual experience, but these physical processes or building blocks are not specified in the theory, besides stating they could be either neurons or logic gates. Thirdly, the amount of phi does not provide any plausible explanation for the unique qualitative nature of consciousness. The theory simply does not provide a way to distinguish between individual conscious states and thus misses something about the nature of consciousness.

How these axioms come to be, as provided by Conifold Theory

Consciousness does seem to involve the integration of all the information held in the brain. However, Integrated Information Theory in its current form cannot explain how neural activity generates a cohesive, streaming, perceptual experience. There is no mechanistic explanation here – only a proclamation that consciousness exists; it is tied to a neural network, but not reducible to neural activity; it is representative of reality; and it has some causal structure. This descriptive explanation provides an excellent starting point, but a more complete theory is needed.

In this light, Conifold Theory is highly compatible with IIT. While IIT states several axioms, Conifold Theory demonstrates how these features naturally arise in systems that obey certain mathematical and physical laws. Firstly, Conifold Theory shows how a neural network with probabilistic gating mechanisms creates representational information content [8]. That is, consciousness is real. Secondly, it shows how this representational information content is tied to the neural network encoding it. That is, conscious states are exclusive. Thirdly, it proposes that streaming perceptual experience is uniquely defined by the neural network state - and the probabilistic particle movement underpinning that state - driven by activation of the sensory apparatus [9]. That is, consciousness represents reality. Fourthly, Conifold Theory specifies that consciousness, as representational information content, is irreducible, as it is mathematically equivalent to an integrated neural network-wide wavefunction or probability density function existing in higher-dimensional space. And fifthly, Conifold Theory specifies how consciousness gives rise to the cause-effect structure of reality by demonstrating how wavefunction collapse physically reduces the total set of probable neural network states into a single reality, which in turn influences subsequent behavior through causation [10].

In short, IIT asserts the necessary features of consciousness, but does not explain the mechanisms underpinning it. Conifold Theory accomplishes both of these tasks, and therefore provides a more complete theory of consciousness than IIT.

References Cited

1. A.K. Engel and W. Singer, Temporal binding and the neural correlates of sensory awareness. Trends Cogn Sci, 2001. 5(1): p. 16-25.

2. C. von der Malsburg, Binding in models of perception and brain function. Curr Opin Neurobiol, 1995. 5(4): p. 520-6.

3. G. Tononi, An information integration theory of consciousness. BMC Neuroscience, 2004. 2(5): p. 42.

4. G. Tononi, M. Boly, M. Massimini, C. Koch, Integrated information theory: from consciousness to its physical substrate. Nat Rev Neurosci, 2016. 17: p. 450–461.

5. A.K. Seth, A.B. Bennett, L. Barnett, Causal density and integrated information as measures of conscious level. Phil Trans A Math Phys, 2011. 369(1952): p. 3748-3767.

6. D. Balduzzi and G. Tononi, Qualia: The Geometry of Integrated Information. PLOS Computational Biology, 2009: p. 1000462.

7. Tononi, G., Integrated information theory of consciousness: an updated account. Arch Ital Biol. , 2012. 150(2-3): p. 56-90.

8. E.A. Stoll, Modeling the probabilistic behavior of electrons at the neuronal membrane yields a holographic projection of information content. Under review.

9. E.A. Stoll, Thermodynamic computing in neurobiological systems. Under review.

10. E.A. Stoll, The mechanics of non-deterministic computation in cortical neural networks. Under review.