Recap previous blogpost
In the previous blogpost I talked about the neural infrastructure of the brain and how knowledge is represented therein. I argued that the neurons, the synapses that connect them and the networks they form, operate in a radically different manner than is often assumed in popular parlance. For one, the brain is not a computer and, as such, no information is stored or processed in the brain. Knowledge, or internalized information, is inherent to the wiring of neural networks; it is, therefore, procedural and not, as is the case with computers, declarative. When a memory from a certain event is recalled it is not being retrieved from a specific location in the brain, instead it is generated in the process of activating the neural networks that were active when the memory was ‘formed’.
“Procedural knowledge means that it is the procedure of ‘retrieving’ the information that is simultaneously producing the information”
If the recall of a memory was a lock being opened by a key, then the traditional way of viewing emphasised the lock as a container of information which could be retrieved by unlocking it with a key. While the more truthful view would be that it is the act of unlocking that brings forth the experience of recalling a memory. In the act of unlocking it matters how fast the key is inserted, the steadiness of the movement, the exact way the teeth of the key scroll over the slots of the lock, etcetera. This is what I mean with procedural knowledge, it is the procedure of ‘retrieving’ the information that is simultaneously producing the information. It is only a metaphor, but it might make the argument a bit less fuzzy.
Integrated Information Theory – its significance
This blogpost will be dedicated to a paper – and growing body of research – called Integrated Information Theory, or IIT for short. In my opinion it is the most promising and comprehensive theory about the nature of consciousness that, as opposed to many earlier efforts, has mathematical models of its predictions which can be tested and validated.
“In my opinion it is the most promising and comprehensive theory about the nature of consciousness that, as opposed to many earlier efforts, can actually be tested and validated”
The theory is relatively new, the first paper dating from 2004, meaning that it has a lot of potential, but also that it has not yet been able to withstand the test of time. Nevertheless, I find it a dazzlingly insightful theory, which truly expanded my understanding of consciousness and the brain. Although the theory is aimed at explaining the mechanisms of consciousness I think it is also telling a great deal about the way our brains function in general.
Do note that I am only reiterating what others, like Dr. Giulio Tononi (the original author) and neuroscientist Christof Koch, have extensively described in their papers. I am, therefore, using much of their vocabulary together with some of my own elaborations. If you are interested in reading more about this theory I would advise you to read the Scholarpedia page, which is probably the most concise and understandable page about IIT.
Integrated Information Theory – the theory
Lets start with a thought experiment: you are sitting in front of a blank screen and are told to say “light” when it turns on and “dark” when it turns off. In front of another screen that is synchronised with your screen a photodiode is placed that sounds a “beep” when the screen is on and does nothing when the screen is off. Essentially you are doing exactly the same thing as the photodiode, but while the photodiode is performing his preprogrammed behaviour that will never voluntarily change, you are consciously seeing a white or black screen and have countless of possible behaviours to act on at any moment. Why do we say that you are conscious of the alternating screen and the photodiode is not? According to the theory the photodiode has only two possible states to be in and, thus, generates only one bit of information, while you, on the other hand, have millions upon millions of possible states to be in at any one moment and so the amount of information that you generate is way beyond the photodiode’s capacity.
To illustrate this fact even more powerfully think about a one Mega-pixel camera. Think of each pixel as being a photodiode as described earlier. Taken together the camera must then have 2^1.000.000 possible states. Still it would be hard to believe that the camera is conscious. Why is that so? What is the difference between you and the camera? The theory predicts that it is the degree of integrated information that gives rise to consciousness. The individual photodiodes of the sensor have no causal effect on each other, they are completely independent, meaning that no integration between the photodiodes takes place. The neural correlates in your brain, however, are incredibly complex and interconnected, thus generating a vast amount of information that makes us able to feel and perceive, and have innumerable different experiences.
“As such consciousness is the capacity of a system of interacting elements to integrate information as a whole”
As such consciousness is the capacity of a system of interacting elements to integrate information as a whole and implies that infants and animals are conscious and that it doesn’t matter if a system is biological or artificial, as long as the system has the right properties.
In order for IIT to develop a general theory of consciousness it needs to identify the fundamental properties of consciousness and the mechanisms of a physical substrate that accounts for them. As an attempt IIT has developed an axiomatic system describing these properties and, henceforth, infers postulates explaining the physical mechanisms underlying them. An axiomatic system is a system of ‘laws’ or ‘rules’ that are so self-evident that they need no further proof; an axiom in geometry is, for example, that the sum of the angles of a triangle in Euclidian space always amount to 180 degrees. The five axioms are as follows:
Intrinsic existence – consciousness is actual. It is the only absolute fact I know, all my other knowledge is secondary and inherently uncertain, even my scientific knowledge is but a likely inference. The theme of this axiom, that we can be sure of our existence, but can never be sure of it’s nature, is, for example, addressed in the Matrix and the Simulation Argument by Nick Bostrom.
Composition – consciousness is structured: each experience is made-up out of elementary and higher order phenomenological distinctions, such as a book, a blue shape, a blue book, a left side, a blue book on your left side, etcetera.
Information – consciousness is specific: each experience is the way it is, and, thereby, differs from all other possible experiences you could have had instead. An experience can be seen as a unique set of a phenomenological distinctions, both positive and negative. For example, seeing a bicycle as opposed to seeing no bicycle, or a blue bicycle as opposed to a red one. Together these positive and negative concepts make-up each experience.
“When seeing the word ‘DAMAGE’ in the middle of a blank page you do not read the words ‘DAM’ and ‘AGE’ separately”
Integration – consciousness is unified: each experience is an integrated whole that is irreducible to non-interdependent parts. When seeing the word ‘DAMAGE’ in the middle of a blank page you do not read the words ‘DAM’ and ‘AGE’ separately. In the same way it is impossible to see the colour blue of a blue book apart from the book. All the phenomenological distinctions are seen as one unified visual field.
Exclusion – consciousness is definite, in content and spatio-temporal grain. I am not experiencing less or more visual distinctions, not a subset nor a superset. Also my experience is not going any faster or slower than the speed that it has, roughly a hundred milliseconds. The edge of my peripheral vision is unperceptive for colour, but I am unaware of that; my experience tells me I see everything in colour.
Now we know what the five essential properties of consciousness are the real magic happens, because at this point IIT attempts to postulate how each of the properties can be explained by the mechanisms of a physical substrate, therefore bridging philosophy with science. In the end IIT is able to calculate both the quantity and the quality of consciousness of a particular system, which has already been successfully applied in a handful of empirical studies.
To be continued…