The Information Integration Theory of Consciousness

The Information Integration Theory of Consciousness
By: Giulio Tononi
Analysis By: James Bresnahan

      Tononi takes one of the older and more defined definitions of consciousness, saying it is the wakeful state that leaves us every night as we go to sleep.  Tononi goes on stating essentially exactly what Chalmers did in defining the two types of problems. The difference is that Tononi believes Neuroscience can address the problem commonly referred to as "the hard problem."  There is one slight difference-- Tononi does really state "why" the hard problem comes about.  I think he does a nice job of defining and establishing how it does so-- although again, I'm not entirely convinced that these aren't really the same problems being approached from two different angles upon initial dissection.  

       Tononi goes on to describe how and why the brain cannot be related to a simple machine as typically described.  Often in the field of consciousness we are forced at looking at a hypothesis where we view someone saying "light" when a bulb is turned on and "dark" when the light is turned off as being the same as a photodiode that beeps when the light is turned on and is silent when the light is turned off.  The problem with this hypothesis, and described by Tononi, is that the number of conscious states that a human uses to differentiate and spit out the word "light" is far more than that of a photodiode.  Essentially, the photodiode never experiences the phenomenology. 

       Even though this difference is quite large, it isn't in and of itself enough to dispute a subjective experience.  However, take the same example and apply it to an HD digital camera.  In an effect, this came may have a million different possible states to describe the transient scene which you are also describing, yet no one in their right mind would describe the camera as conscious.  Why?  Tononi suggested that the lack of inter connectivity is the reason. If you took every pixel that a camera takes in and process that Information separately then put it back together you are essentially left with the same picture.  This is completely false when it comes to the human subject experience. Your brain is constantly integrating your experience, leading to the subjective feelings associate with each experience, often referred to as qualia or phenomenology. We can see the disastrous effects of separating the components of our brain when we do things like study split brain syndrome, or destroy specific neural tracts.  

       You may ask, how does one even go about systematically quantifying the number of tracts present in the conscious system. Tononi proposes that in order to measure information integration, it is essential to know whether a set of elements constitute a causally integrated system, or whether they can be broken down into a number of independent or quasi-independent subsets among which no information can be integrated.  Tononi proposes how to do so in a series of theoretical publications (2001).  Think of this simple system:  there is a small circuit of interconnected neurons in the brain that are completely isolated from external input such as present during anesthesia or sleep.  Assume the group of neurons is something like a cortical mini column, bringing information between layers of the cortex.  Assume each layer can have different firing levels lasting a few milliseconds each.  We try to figure o up all the possible combinations of firing patterns as outputs from 'A', and establish how differentiated is the repertoire of firing patterns they produce in 'B'.  This can be viewed as the entropy caused in B when the outputs from A are substituted by independent noise sources.  Of course, the same can be done for the effective information between B and A. 

Here you can see an information, minimum information bipartition, and analysis.  This allows us to assess how much information can be integrated within a system of elements.  It is key to realize that these are two independent subsets and not a single integrated subset. To measure the capacity of a subset, the minimum information bipartition is used. The symbol shown is used to determine the extent of the integration all ability of the subset.  If we do this for many subsets while discarding those with external reference, we are left with complexes of information.  Complexes are individual entities that can integrate information.  Whatever complex has the lowest minimum information bipartition, that complex is considered the main complex.  Elements of a complex that inevitably connect with other elements not part of it are referred to as port-ins and port-outs.  In the case of the brain, the spatial elements and time scale that maximize minimum information bipartition, are likely to be local collections of neurons such as mini columns and periods of time comprised between tens and hundreds of milliseconds respectively.  This system can be used to analyze and identify complexes.  Complexes are the subjects of experience,  essentially, since these complexes are not branching out to receive input or output, it is the self reflection that builds conscious experience, according to Tononi.  


       There are a few systems with very low interactions among modules, and therefore complex size and minimum information bipartition are necessarily low, yet there is no conscious experience.  Why? Something along this line like the cerebellum should, theoretically, have some kind of conscious experience.  Interesting, it is thought that some of these circuits may be purposely eliminated of external conscious experience in order to maintain homeostatic functions, and things like sleep.  

       Tononi goes on further to explain how his hypothesis is different from past hypotheses such as Crick and Koch. He states that while their hypotheses lead to an all-or-nothing response, the information integration theory of consciousness is one that allows a graded response.  For instance, the meaning of each and every quale is subjective and revolves around what we associate with based on previous experiences.  This therefore allows for a graded and person experience to each and every qualia.  Yet another important clarification depends on understanding that even within a complex where not every neural fiber is activated, there is still comes contribution towards conscious experience.  Essentially, what counts is how much is how much integrated information is generated, not how widely it is disseminated.  Every artifact is conscious to the extent that it houses a complex of high minimal information bipartition.


Comments