Note on Integrated Information Theory
10/14/20232 min read
IIT has been hailed for promising a fully mechanistic, even quantitatively defined physicalist explanation of consciousness. What is strange to me about this is the patent absurdity of attempting to explain consciousness without any reference to its content and functional relationship to the rest of the mind. Few theorists (and even fewer civilians) would disagree that a crucial aspect of what we call being conscious is having a sense of presence, the sense that there is some sort of a 'me' that is observing, witnessing 'what is going on'. It is the structuring of the perception/understanding of mental processes within a tripartite framework of an (explicit or implied) observer, the observed and properties of the process of observing (paying attention, being absorbed, trying to avoid, responding emotionally etc.) that turns processing an input into an 'experience'. Indeed, one hardly needs to be a fully committed analytical functionalist to agree that the very meaning of the word 'experiencing' is to perceive/understand/analyze unfolding mental activities as things happening to/within/by an agent (whatever that construct exactly is).
The problem with IIT (and some other theories) is the failure to realize that the 'agent' which does the observation is necessarily a virtual one and not the physical substrate which implements it. The observing 'me' must be a construct represented (displayed) in the mind because that is the only way its relationship to and interaction with all other contents of the mind can be perceived or experienced. To see how IIT fails because of the category error of equating the implementation with the content that it represents, consider one of its central tenets, the unity of conscious experience. It is certainly true that our experiences do not appear as disconnected fragments. But why should we expect their perceived unity to result from (and according to IIT, to require) the physical "gluing" together of neuronal stuff that underlies the components of the experience ? The story, meaning and emotional arc of a movie form a coherent whole but this requires nothing more than explanatory connectedness of their elements (which Mr. Nolan should sometimes be reminded of). The scenes and characters in a story are distinct but their distinction does not fragment the story into unrelated parts because they and their actions are thematically accounted for. The perceived unity of conscious experience is established by the same coherence of the narrative of the 'here and now' (past and present or I and thou). What we perceive is a rabbit, not separate, co-travelling ears, legs, body and tail because the object segmentation performed by the visual system conceives of the limbs etc. as constituents of a whole, and the whole rabbit as the thing which is made up of these particular parts working together the characteristic ways the do (Fodor).
While IIT may be an important mathematical/physical research program and may even provide us with useful tools for analyzing neuronal data, if it aims to be a theory of something that is not recognizable as consciousness. In the end, IIT seems to me as a synesthetic version of alchemy: it's practitioners devise networks with features that elicit the same vague feelings as introspective meditation on the nature of experience does and then - keeping their fingers crossed - wait to see the gold of consciousness miraculously appear in the frying pan of function-free computations.
Published as a comment to Albantakis et al., (2023) Integrated information theory (IIT) 4.0: Formulating the properties of phenomenal existence in physical terms. PLOS Computational Biology, https://doi.org/10.1371/journal.pcbi.1011465, @ https://journals.plos.org/ploscompbiol/article/comment?id=10.1371/annotation/bd778039-85c1-4b29-b437-f8ad4574f3c3