The motivation of this article is that this theory of consciousness, the ITT, will be tested with its rival, the Global Workspace (GWT), in an experiment supposed to separate them. I will answer the following questions in succession:
-What is it the ITT?-What are i
ts strengths and weaknesses?-How is it very differ
ent from the previous ones, especially the GWT?-What is this experience that can separate them?-
In conclusion a prognosis on the result.-And as a bonus: What is thi
s experience that can decide them?-In conclusion a prog
nosis on the result.-And as a bonus: What is it missing from the ITT and its competitor to really explain the conscience?
What is ITT?
Everything is said in the name: integrated information. What does 'integration' mean for information? Information has a different value (in terms of storage bits) depending on whether it is isolated or associated with others.
For example, you look at a red object. What is the value of 'red' information? 1st case: your eye perceives only two colors, red or white. The information fits in a single bit: 1/0 (red/white). 2nd case: your eye differentiates thousands of colors. Red must be encoded in the middle of all these possibilities. That takes up a dozen bits. The same qualitative information (the object is red) can thus cover a variable amount of information. The degree of quantitative integration of qualitative differs.
The information is integrated if when the eye says 'red' it also says 'it is none of the other contingencies'. It's a fusion of all that information. Non-integrated information, conversely, would be provided by thousands of sensitive sensors each at a specific wavelength. Only one of them would react to the red color of the object. The qualitative information is provided but its degree of integration is zero. The positive output of the sensor does not integrate the negative outputs of others. The amount of information recovered is again on a single bit (sensor turned on/off).
For Tononi's ITT (2008), the brain is conscious because it integrates information from all neural patterns. Accurate information is difficult to locate because it relies on the presence of all others. Tononi formalized the degree of consciousness of a neural system by the Phi coefficient, which is the additional information generated by the integration of the parts, added to the simple set of interactions of the parties. It is a quantification of emerging information.
What are the strengths and weaknesses of the ITT?
The ITT explains very well why a camscope or computer are not aware, despite the same information as the brain. They are just recording this information. They don't integrate them.
Some of the criticisms come from people who have not really understood the heart of the theory. For example, those who claim that a country or the internet meet the ITT's criteria of awareness. It's stupid. The information forming the 'country' entity is not linked by any physical integration. The internet is a collection of information that is made available to users. The network does not integrate them by itself, or when it starts to do so (integration of information about a user) it is still extremely frustrating and without integration of data about other users.
When I log in to Google, he is not aware that I am Jean-Pierre Legros. There is only information stored on Google's servers about me, awakened by my connection. They're independent of other news. For Google to be aware of me, there would have to be something in it that simultaneously knows that I am none of the other billion Internet users it is able to recognize. There's going to be something in him to pay attention to me.
A more solid critique is that of the mathematical formalization of the ITT. A first attempt by Griffith (2014) was based on a destructive compression of information. It involved replacing the constituent information with its integration and thus a loss of memorial elements. Memories couldn't be kept.
A second attempt (Maguire, Moser, Maguire, Griffith 2018) uses algorithmic information theory and proposes loss-free integration. It demonstrates that such integration requires non-computable functions.
This is not, strictly speaking, a denunciation of the ITT. The conclusion is: If integrated consciousness exists, it cannot be the subject of a computational model. In a sense this is good news for those who seek to escape consciousness from the hard sciences and computers. It is not possible to transfer consciousness into our current machines. The ITT is certainly the theory that can appeal to proponents of qualia independent of matter.
How is ITT different from global workspace theory (GWT)?
ITT is derived from information theory while GWT is a purely neuroscientific theory. The GWT comes from observing the flow of trade in the city. It's topographical. Some areas of the brain are activated in correlation with consciousness. They reverberate with the same data. The consciousness seen by the GWT is the observation by the neural patterns of their own functioning.
The GWT paradigm is more conventional and unable to explain the consciousness phenomenon. It assumes that mirrored exchanges of neurons account for consciousness but is powerless to give reason for it. The ITT paradigm is radically different. Revolutionary and transdisciplinary, he had great difficulty establishing himself in neuroscientists. But things are starting to change.
What is the experience that can separate them?
The GWT says that information becomes conscious when it is accessible to the entire overall workspace, which includes a priority area: the prefrontal cortex. Consciousness results in fMRI by a flare-up of these neural areas, 300 to 400ms after a stimulus.
While the ITT defines consciousness as proportional to the degree of integration of information. However, it is the posterior cortex that incorporates the maximum senitivo-motor information. And in this hypothesis consciousness starts as soon as the information begins to be integrated, without delay of transmission.
To decide between the two theories, experiments must not only collect the signals associated with consciousness, but do so with great finesse both spatially and temporally. There is no technique between the two. The studies that begin this year will therefore pair several devices, MRI, magnetoencephalography, intracranial electroencephalography.
Prognosis on the result
The result can be interpreted as desired in favour of either theory. Because these theories do not say the same thing and do not speak of the same consciousness.
What we call 'consciousness' is the amalgamation of two things: the whole and its parts. The level of integrated information and the constituent information. The first forms the thoughts and the seconds elevate a pyramid that corresponds to the conscious experience.
GWT theorizes the higher level of information: conscious thoughts. The ITT explains the conscious experience, the phenomenon itself. These are more complementary theories than rivals. ITT is closer to the truth because it does not reduce consciousness to the posterior cortex; it says that the whole brain is involved, and more so in areas where information is most integrated.
What is missing from these two theories
The GWT is not rooted in the structure of physical reality. She simply observes neurons at work. The ITT is more ambitious with its anchorin in information theory. All we see of physical reality is information. But it obliterates the complex dimension, which is a hierarchy. Information is not continuously integrated but discontinuous. These levels give independence relative to the integrated levels. The appearance of a quality above the amount of information signals the presence of one of these levels.
To use the example of colors seen by the eye, each color is also an integrated information over subatomic interactions. This quality is a fusion that cannot be mixed with that of 'red object' because they do not correspond to the same level of integration. There is a qualitative leap. The two integrations are not mixed but superimposed.
This theory called Stratium solves all ITT problems and integrates GWT without difficulty. The non-computability of integrated consciousness is expected since there is no current mathematical formalism to link integrations of a different nature. This is not to say that it does not exist. I explain elsewhere the inadequacies of algorithmic information theory.
That GWT supporters find the most awakened consciousness in the prefrontal cortex is expected since it is the top of the integrative hierarchy. They look at the activity of the upper floors. But as in a computer these floors would not provide authentic conscious experience if they were not perched on the complete pyramidal integration of neurons, as indicated by the ITT.
What ITT forgets to say is that this experiment is based on the presence of neurons as a medium of exchanges. They themselves are integrated information. Any other support would provide a different experience.
Finally, the ITT posits that the information is not localized, because of its integration. For example, it would not be possible to make the brain forget information by removing a neuron. This is not entirely true, always because the hierarchy is neglected. The relative independence of the levels produces symbolic values specific to each neuron, in the same way that the tops of a graph have different weights depending on their links.
Removing a neuron can remove a memory. However, the information constantly generates its own integration. If the stimuli that have created the memory are repeated, they quickly determine a new neuron-symbol. The direct stimulation of it will instantly awaken the memory in consciousness, involving its underlying integrative pyramid.
That is to say, the Stratium could also receive experimental confirmation… if it was included in the hypotheses to be tested.