The other day, when discussing a paper that criticized IIT (the Integrated Information Theory of consciousness) as unscientific, I noted that IIT, while questionable as the ultimate answer for consciousness, could be useful in the more limited capacity of distinguishing degrees of consciousness in a brain. Apparently I’m not the only one who thinks that, as this preprint paper from Matthias Michel and Hakwan Lau discusses: On the dangers of conflating strong and weak versions of a theory of consciousness.
The authors identify two versions of the theory:
Empirical IIT is the view that some measures of complexity of brain networks can be used to detect states of consciousness, i.e., whether subjects have subjective experiences (such as during wakefulness, or dreams), or not (such as being in a coma, anesthetized, or dreamless sleep). That is, Empirical IIT takes complexity to be a marker of consciousness.
…On the other hand, Fundamental IIT is the view that a specific form of complexity (integrated information) is identical with consciousness (Oizumi et al., 2014). That is, Fundamental IIT posits that a specific form of complexity is the constituent of consciousness. Specific conscious experiences are also exactly identified with specific states of a network with this particular form of complexity (Tsuchiya et al., 2015; Tsuchiya, 2017).
The authors see merit in the empirical version of the theory, but are “unsure” about the latter. This actually matches my own intuition pretty closely. It’s the metaphysical and somewhat magical aspects of the stronger “fundamental” theory that turn me off, the identification of consciousness with complexity, without attempting to wire up how that complexity, in and of itself, leads to specific percepts.
Anyway, the paper more broadly urges that researchers strive to define the NCC (neural correlates of consciousness) in terms of the markers of consciousness, what shows its presence, rather than as the constituents of consciousness, that is, in terms of the signs of consciousness rather than the identity of it. It’s a caution for epistemic humility.
And it reminds me why I have instrumentalist rather than realist leanings, that is, viewing scientific theories as pragmatic prediction tools rather than metaphysical statements of reality. When discussing Stanislas Dehaene’s theory in the previous post, I noted its promising value from an instrumental viewpoint, that is, one focused on predicting observations. Taking that view, we can find ways to productively combine theories like IIT, GWT (Global Workspace Theory), HOT (Higher Order Theory), and many others.
But if we take each theory in its strongest incarnation, one that asserts that what it describes is consciousness, then there can be only one. And I highly doubt there will be only one. The collection of cognitive capabilities that trigger our intuitions of consciousness are too complex and varied. Much like the difficulties in defining life (described by an Aeon article out today), we will almost certainly need many theories.
Unless of course I’m missing something.
(Via tweet by Hakwan Lau)
11 thoughts on “Empirical vs Fundamental IIT and the benefits of instrumentalism”
Do you take requests? :o)
Other than curiosity, what is the goal of understanding consciousness? I would be interested in your thoughts on this. (I am curious … and….) Will understanding consciousness lead to new psychiatric care protocols? Will it lead to better education? There are a few hypothetical “goods” that may appear. On the other hand, will it allow more manipulation by “bad actors,” advertisers/marketers, political operatives, cultists, etc.? Are you aware of any speculation on this?
LikeLiked by 1 person
Good question, Steve.
Despite the fact I keep commenting on “consciousness” questions, I am beginning to think it may not be a scientifically coherent concept. We may be lumping together a bunch of different things brains and neurons do because subjectively it feels to us like it is a unified experience.
It seems to me that effective modeling and definitions for the “consciousness” term, will help transform our woefully soft mental and behavioral sciences into harder forms of study. We don’t yet effectively understand our nature, either personally (psychology) or socially (sociology).
Why should “hard” forms of these sciences matter? Though we’re extremely powerful today, we don’t yet grasp how to use our power effectively given how mysterious human nature happens to be to us. Once we do get a better handle on this, we should thus have theory from which to better lead our lives and structure our societies.
I consider us to have experienced four great power revolutions: oral language, specialization, written language, and hard science. The final revolution I speak of should be special however, since instead of just more power, this one should finally teach us how to live.
My original interest, years ago, was to determine if there was any evidence for dualism. That quickly got ruled out.
My interest flared a bit later as research for writing science fiction. But I’m pretty sure I’ve long since picked up enough there. (It is interesting how many sci-fi authors, who wouldn’t dream of screwing up their physics, make a mess of things when they write about the brain.)
Why do I remain interested? Hard to say. I just am.
I do think understanding the brain will lead to all kinds of much needed medical treatments. And it will help with artificial intelligence, regardless of whether we ever bother to mass produce machines that meet our intuitions of consciousness. For instance, understanding how grid cells in the hippocampus and entorhinal cortex track location and time positions will give us insights into doing the same thing with robots.
But it will also likely lead to all the bad things you list, and many more. Knowledge is always a double edged sword.
For consciousness in particular, I’m with James. I think the word “consciousness” is hopelessly vague. We can make it a little less so by prefixing or qualifying it (exteroceptive consciousness, interoceptive consciousnes, affect consciousness, access consciousness, etc) or focusing on more narrow capabilities (attention, semantic memory, episodic memory, metacognition, etc). Each of these can be defined objectively enough to study empirically and reach relatively non-controversial conclusions.
But no matter how you study consciousness itself, there will always be someone who says you’re not studying the real thing. And a bunch of people who say science has no hope of doing anything with it.
On the other hand, there is no one definition of life, and biologists know a lot about it and are constantly learning more, and have a good handle on how it fundamentally works. I think the same will eventually be true for consciousness. In truth, most neuroscientists stay away from the subject, even though it’s basically what they’re studying, to keep the philosophers off their back.
As usual, I enjoyed your post and read the Aeon article with much gusto. I was intrigued by your suggestion that the difficulties in defining consciousness are “much like the difficulties in defining life”. The Aeon essay suggests that life might be an integrated reflexive information system feeding back into itself. The actual information content is critical and highly specific although it can vary greatly across different biological or physical systems. This has been a common thread running through many of your posts on consciousness. It occurred to me that I’ve often heard it said that Nature is economical in its solutions; that is, when it solves a particular problem elegantly, it tends to repeat that solution elsewhere. Just as the genes encode instructions on how to build all the components of our cells, including the ribosomes that read those instructions to build the proteins required for those cellular structures, our neurons might encode instructions for constructing the functional circuitry for perception, memory, conception, imagination, language, awareness, accessible consciousness, etc., with feedback loops so that virtual systems are layered on top of physical systems, and virtual systems on top of virtual systems with increasing levels of integration on each feedback loop. Complexity in itself is not enough. Only certain specific configurations are workable (survivable), but they probably would vary widely across different biological or physical systems. Anyway, this was my intuition after reading your recent posts on the subject.
LikeLiked by 1 person
That’s an interesting intuition. It is a fact that we have many common innate circuits for perception. For example, newborns come out of the womb with an ability to recognize faces and a strong penchant for doing so. And all primates have an aversion for slithering snake like shapes.
The usual thinking is that this ultimately traces back to genes, or gene expression. But of course, any individual gene only codes for a protein. Which means somehow the combinations of proteins leads to everything we see.
It’s conceivable that some of these combinations may lead to certain neuron types, and the epigenetic factors may control the prevalence of certain types in certain regions. How all of this leads to the specific circuits, the combinations of axons, chemical synapses, electrical synapses, and everything else seems like a bewildering problem.
And of course we can’t separate genes, gene expression, and proteins from the environment in which we develop, initially the womb, and later our overall environment.
LikeLiked by 1 person
There are two kinds of “real essence” of jade. I don’t see why there couldn’t be two kinds of consciousness.
I think the distinction between “empirical” and “fundamental” theories generally collapses, and consciousness studies may well wind up being a case in point. If an empirical theory of X is successful enough and explains enough, it amounts to the discovery of the fundamental nature of X.
The jade comment’s going over my head. But I would say there are far more than two kinds of consciousness. Admittedly it depends on how you divvy it up.
Scientific theories are typically successful within certain domains, but often fail as those domains are exceeded. Newtons laws were once taken to be fundamental. They remain extremely useful for many purposes in non-relativistic conditions, but are now known not to be the final answer. General Relativity breaks down in the singularity of a black hole, and doesn’t address gravity at the quantum level.
IIT may be successful within a brain. But outside of that, it makes predictions about other systems being conscious, such as a configuration of inactive logic gates: https://www.scottaaronson.com/blog/?p=1799
(This isn’t a strawman example. Tononi endorsed it.) I think it’s fair to say there’s no way to test this.
The truth of any scientific theory will always be provisional.
Thanks foor writing this
LikeLiked by 1 person