Consciousness as Information Integration
Updated: Jun 16, 2019
I just finished reading “The Great Unknown: Seven Journeys to the Frontiers of Science”, a book authored by Marcus du Sautoy, a British mathematician and a prolific writer.
There was a section which described the "Integrated Information Theory" (IIT) research being done at the Center for Sleep and Consciousness at the University of Wisconsin-Madison, lead by the neuroscientist Guilo Tononi which focuses on developing a mathematical model of consciousness.
IIT applies to all forms of matter, not just brains, and it implies that panpsychism (the doctrine that consciousness is a property not just of brains but of all matter) might be true. Panpsychism may sound absurd at first go, but the idea that the subjective experience is an emergent phenomenon of the structure of matter is not so far-fetched, so I decided to go into it in a little more depth.
Starting off with the three basic axioms that any conscious experience needs to be structured, specific/differentiated and integrated, Tononi proposes that we can identify any entity's (person, animal, or even a computer) consciousness from the level of “information integration” that is possible in the brain (or CPU). According to his theory, the more information that is shared and processed between many different components to contribute to that single experience, the higher would be the level of consciousness.
Therefore, just like entropy is a measure of the information or randomness possessed by a system, the output of this theory (denoted by Φ or Phi) can also be viewed as a quantitative measure which determines some sort of cohesive information gain that occurs due to the structure and connectivity present in the network. Phi corresponds to the feedback between and interdependence of different parts i.e. the synergy of a system. In other words, it measures the degree to which a system is “more than the sum of its parts.” This means that if a system has Φ = 0, then its cause–effect power is completely reducible to that of its parts - it cannot lay claim to existing. On the other hand, If Φ > 0, then the system cannot be reduced to its parts, so it exists in and of itself.
IIT is not without its own set of controversies and skeptics. Computer scientist Scott Aaronson's complaint with IIT is its claim that high Phi produces consciousness. “Phi may be a necessary condition for consciousness, but it is certainly not a sufficient condition”, he says. This is because we can demonstrate some systems for which the model unavoidably predicts vast amounts of consciousness but that no sane person would regard as particularly “conscious” at all (like the Vandermonde system). Moreover, IIT is computationally unfeasible in most occasions, a problem which is further exacerbated by how little we know of the interconnection network of the human brain. It is also argued by some that IIT avails the circular reference and solepsicm problems. There are even further issues that crop up when we enter the amorphous world of Ontology and Phenomenology.
What I find interesting is that even if we don't go into the metaphysical realm of consciousness and only borrow the mathematical framework of this research, it could help us understand the ways in which we interact with each other which lies at the heart of economic modeling. Thomas Malone, director of the Massachusetts Institute of Technology's Center for Collective Intelligence has applied this theory to teams of people in the laboratory and in real-world, including the editors of Wikipedia entries. He has shown that the estimates of the integrated information shared by the team members could predict group performance on the various tasks. This shows that integrated information can be a useful way of characterizing a certain kind of interactional complexity that, at least sometimes, predicts group performance. In this sense, Phi can be viewed as a potential metric of effective group collaboration.
"Group consciousness" may seem like an outlandish concept to explain the above phenomenon, but Tononi’s theory might help us to understand how large bodies of people sometimes begin to decide and react as one entity - the so called “hive mind” at work. I am reminded of the extensive literature on "information cascades" and the "wisdom of the crowds" concept in economics which tries to explain this very phenomenon in relation to socio-economic activity such as the creation of a bubble so it would be a nice thought experiment to see how information integration in group networks plays out.