The Integrated Information Theory (IIT) is the leading theory of consciousness in Neuroscience. Proposed in 2004 by Giulio Tononi, the theory has consistently been advanced in the previous decade as technologies that allow us to look into the brain continuously improve. The IIT is now on its third iteration, and has offered new axioms and postulates to guide consciousness studies. The article below, written by Hedda Hassel Morch in an interview with Tononi, attempts to explain what integrated information is, how the theory can explain consciousness, and what its implications are.
Based on a previous article [we have been introduced to IIT in What Is Consciousness? which summarizes the fundamental nature of consciousness], integrated information can be represented by the mathematical symbol phi or Φ. According to IIT, a system that has high Φ has high consciousness, which means that their experiences are both complex and meaningful. Meanwhile, systems with low Φ has low consciousness, their experiences are simple and rudimentary. Those with zero Φ are not conscious at all. Based on this theory, if we are able to measure a system’s Φ, then we would be able to determine the level of consciousness of that system, and in effect, would be able to understand the unique role of consciousness in life and creation.
But what exactly is integrated information, and how does it explain consciousness? To understand this, we must first define the two terms – information and integration.
According to Tononi, all systems contain information. There are systems like books, computers, and photographs, which contain information information about other things. A book may contain information about other books, but very little information about itself.
Meanwhile, there are systems, like the human being which contain information about the outside world, but there is something within us that signals when our body is in need of fuel, or when it is at the edge of energy depletion. According to IIT, the only kind of information that matters for consciousness is the information a system has about itself. [Using this point of view, the unique capacity of plants to ensure their survival may be considered as more than mere adaptation, plants may be considered to be conscious. For more information about this idea, read Plants Are Intelligent, Says Plant Neurobiologists].
Meanwhile, integration, according to IIT, refers to the interconnections between the parts of the system. One must ask: how much information will be lost if a system is divided? Consider tearing a page of book into two – will any information be lost? If we put the two parts of the page, we will still be able to read what it contains. Hence, in essence, no information will be lost. Even if we cut the page into even more parts, as long as we keep every part, put them together, the information it contains will be intact.
Now consider cutting the brain into two – will any information be lost? Death of one part of the brain will definitely affect the over all capacities of a human being, which means that the brain contains an intricate structure that does not exist in the pages of a book. Disconnecting one neuron from the system could imply a different state for the brain.
The third requirement of consciousness – maximality
According to IIT, the third and final requirement for consciousness to exist is that the system must contain the maximum amount of integrated information. In the case of the human being, the posterior cortex (which studies suggest as the part of the brain which directly supports consciousness) must have higher Φ than any other neuron in the brain, or molecule in the body. It must even have higher Φ compared to the entire brain, the human body, human societies, or the internet! Why is this important?
Consider this: if a smaller group of neurons in the human brain surpasses the Φ of the bigger area of the brain, then it could develop its own consciousness that is separate from the bigger whole. Meanwhile, if the the Φ of this smaller group of neurons drop below the Φ of the bigger area, the separate consciousness it once had will dissolve to the consciousness of the bigger whole.
When we sleep, IIT claims, Φ in the conscious part of the brain drops, and is dissolved into multiple smaller systems with lower Φ. Our consciousness becomes fragmented as the part of ourselves which is able to identify the “I” in our system drops its Φ during sleep. In this sense, we may say that we don’t truly lose consciousness during sleep (since the rest of our bodily organs, which themselves are integrated, continues to function. Hence we are able to “come back” to our bodies when we wake up), only the part of ourselves which is able to identify who “we” are.
Discovering the IIT
What is the proof of IIT? Aside from the various tests made in the lab, Tononi came to this theory through the study of his own experience. The first-person perspective, according to Tononi, is the only perspective from which consciousness can be directly observed.
Implications for AI
Needless to say, the IIT, if proven to be true, has many radical implications. For one, we can say that computers today have very low Φ. Their computational capacity may have exceeded that of the human being, but remove one transistor from the mother board, replace it with another one exactly like it, and the computer continues operating. This cannot be done on the human brain. We cannot simply remove and replace a neuron, and expect the human to continue functioning as if nothing happened. That being said, as long as computers are made with traditional hardware, then their Φ will remain negligent, and it will not be able to develop its own consciouness that will equal that of the human being [creating a conscious machine is one of the bigger goals of AI, read With Mind Uploading, There Is No Need For Religions and What Is Transhumanism?].
Second, there is no computer in the world that contains information about itself. Whatever information it spews forth, is an integration of all the information it was fed. The past and future states of a computer, no matter how “intelligent” it is, will be determined by external factors. It is unable to determine things for itself [read about the implications of this reality in AI Will Never Trully Understand Ethics].
Third, if the human being contains maximum Φ in our known universe, is it possible that there is a bigger Φ than our own? If such a bigger Φ exists, then human consciousness would not be able to detect it, simply because we are subsumed under it. Could the human consciousness be a smaller group that has separated itself from the bigger whole that is the cosmos? [To understand this context, see Human Life And Its Connection To All Of The Universe’s Creation].
This point is important as the results of astrophysics is showing us that our universe, and the existence of all life is impossible. The laws are the universe are so fine tuned it can only be so for a specific reason. Is it possible that we are aborting this bigger purpose with the creation of a conscious AI?
Read Original Article
Click the button below if you wish to read the article on the website where it was originally published.
Click the button below if you wish to read the original article offline.
2 thoughts on “What is the Integrated Information Theory?”
If indeed there is a greater Phi, and it’s gateway is coherence within our neurobiology, then interfacing that biology with computational devices devoid of consciousness seems troubling on an existential scale.