A Brick o’ Qualia: Tononi, Phi, and the Neural Armchair

by rsbakker

Aphorism of the Day: The absence of light is either the presence of dark–or death. For every decision made, death is the option not taken.

Aphorism of the Day II: Things we see through: eyes, windows, words, images, thoughts, lies, lingerie, and excuses.


So Guilio Tononi’s new book Phi: A Voyage from the Brain to the Soul has been out for a few weeks now, and I’ve had this ‘review’ coalescing in my brain’s gut (the reason for the scarequotes should become evident in due course). In the meantime, as fate would have it, I’ve stumbled across several reviews of the book, including one that is genuinely philosophically savvy, as well as several other online considerations of his theory of consciousness. And of course, everyone seems to have an opinion quite the opposite of my own.

First, I should say that this book is written for the layreader: it is in fact, the most original, beautiful general interest book on consciousness I’ve read since Douglas Hofstadter’s Godel, Esher, Bach: The Eternal Golden Braid – a book I can’t help but think provided Tononi with more than a little inspiration – as well as a commercial argument to get his publishers on board. Because on board they most certainly were: Phi is literally one of the most gorgeous books I have ever purchased, so much so that ‘book’ doesn’t seem to do it justice. Volume, would be a better word! The whole thing is printed on what looks like #100 gloss text paper. Posh stuff.

Anyway, if you’re one of my fiction readers who squints at all this consciousness stuff, this is the book for you.

What makes this book extraordinary is the way it ‘argues’ across numerous noncognitive registers. Tononi, with the cooperation of his publisher, put a great deal of effort into the crafting the qualia of the book, to create, in a sense, a kind of phenomenal ‘argument.’ It’s literally bursting with imagery, a pageant of photographic plates that continually frame the text. He writes with a kind of pseudo-Renaissance diction, hyperbolic, dense with cultural references, and downright poetic at times. He uses a narrative and dialogic structure, taking Galileo as his theoretical protagonist. With various guides, the father of science passes through a series of episodes with thinly disguised historical interlocutors, some of them guides, others mere passersby. This is obviously meant to emulate Dante’s Inferno, but sometimes, unfortunately, struck me as more reminiscent of “A Christmas Carol.” Following each of these episodes, he provides ‘Notes,’ which sometimes clarify and other times contradict the content of the preceding narrative and dialogue, generating a number of postmodern effects in genuinely unprecedented ways. Phi, in other words, is entirely capable of grounding thoroughly literary readings.

The result is that his actual account, the Information Integration Theory of Consciousness (IITC), is deeply nested within a series of ‘quality intensive’ expressive modes. The book, in other words, is meant to be a kind of tuning fork, something that hums with the very consciousness that it purports to explain. A brick o’ qualia…

An exemplar of Phi itself, the encircled ‘I’ of information.

So at this expressive level, at least, there is no doubting the genius of the book. Of course there’s many things I could quibble about (including sexism, believe it or not!) but they strike me as too idiosyncratic to belong in a review meant to describe and evaluate the book for others.

What I’ve found so surprising these past weeks is the apparent general antipathy to IITC in consciousness research circles, when personally, I class it in the same category as its main scientific competitors, like Bernard Baars’ Global Workspace theory of consciousness. And unlike pretty much everyone I’ve read, I actually think Tononi’s account of qualia (the term philosophers use for the purely phenomenal characteristics of consciousness, the redness of red, and so on) can actually do some real explanatory work.

Most seem to agree with Peter Hankins’ assessment of IITC on Conscious Entities, which boils down to ‘but red ain’t information’! Tononi, I admit, does have the bad habit of conflating his primary explanans for his explandum (and thus flirting with panpsychism), but I actually don’t think he’s arguing that red is information as he’s arguing that information integration can explain red as much as it needs to be explained.

Information integration builds on Gerald Edelman’s guiding insight that whatever consciousness is, it has something to do with differentiated unity. ‘Phi’ refers to the quantity of information (in its Shannon-Weaver incarnation) a system possesses over and above the information possessed by its component parts. One photodiode can be either on or off. Add another, and all you have are two photodiodes that are on or off. Since they are disconnected, they generate no information over and above on/off. Integrate them, which is to say, plug them into a third system, and suddenly the information explodes: on/on, on/off, off/on, off/off. Integrate another, and you have: on/on/on, on/on/off, on/off/off, off/off/off, off/off/on, off/on/on, off/on/off, on/off/on. Integrate another and… you get the picture.

Tononi argues that consciousness is a product of the combinatorial explosion of possible states that accompanies the kind of neuronal integration that seems to be going on in the thalamocortical system of the human brain. And he claims that this can explain what is going on with qualia, the one thing in consciousness research that seems to be heavier than Thor’s hammer.

Theoretically speaking, this puts him in a pretty pickle, because when it comes to qualia, two warring camps dominate the field: those who think qualia are super special, and those who think qualia are not what we make of them, conceptually incoherent, or impossible to explain without begging the question. Crudely put, the problem Tononi faces with the first tribe is that as soon as he picks the hammer up, they claim that it wasn’t Thor’s hammer after all, and the problem he faces with the second tribe is that they don’t believe in Thor.

The only safe thing you can say about qualia is that they are controversial.

Tononi thinks the explanation will look something like:

The many mechanisms of a complex, in various combinations, specify repertoires of states they can distinguish within the complex, above and beyond what their parts can do: each repertoire is integrated information–each an irreducible concept. Together they form a shape in qualia space. This is the quality of experience, and Q is its symbol. (217)

The reason I think this notion has promise lies in the way it explains the apparent inexplicability of things like red. And this, to me, seems as good a place to begin as any. Gary Drescher, for instance, argues that qualia should be understood by analogue to gensyms in Lisp programming. Gensyms are elements that are inscrutable to the program outside of their distinction from other elements. Lisp can recognize only that a gensym is a gensym, and none of its properties.

Similarly, we have no introspective access to whatever internal properties make the red gensym recognizably distinct from the green; our Cartesian camcorders are not wired up to monitor or record those details. Thus we cannot tell what makes the red sensation redlike, even though we know the sensation when we experience it. (Good and Real, 81-2)

Now I think this analogy fails in a number of other respects, but what gensyms do is allow us to see the apparent inexplicability of qualia as an important clue, as a positive feature possessing functional consequences. Qualia qua qualia are informatically impoverished, ‘introspectively opaque,’ so much so you might almost think they belonged to a system that was not designed to cognize them as qualia – which, as it turns out, is precisely the case. (Generally speaking, theoretical reflection on experience is not something that will get you laid). So in a sense, the first response to the ‘problem of qualia’ should be, Go figure. Given the exhorbitant metabolic cost of neural processing, we should expect qualia to be largely inscrutable to introspection.

For Tononi, Q-space allows you to understand this inscrutability. Red is a certain dedicated informatic configuration (‘concept’) that is periodically plugged into the larger, far more complex succession of configurations that occupy the whole.

Now for all it’s complexity, it’s important to recall that our brains are overmatched by the complexity of our environments. Managing the kind of systematic relationships with our environments that our brain does requires a good deal of complexity reduction, heuristic mechanisms robust enough to apply to as many circumstances as possible. So a palette of environmental invariants are selected according to the whims of reproductive success, which then form the combinatorial basis for ‘aggregate heuristic mechanisms’ (or ‘representations’) capable of systematically interacting with more variant, but recurrent, features of the environment.

So red helped our primate ancestors identify apples. As thalamocortical complexity increased, it makes sense that our cognitive capacities would adapt to troubleshoot things like apples instead of things like red, simply because the stakes of things like light reflected at 650nm are low compared to things like apples. Qualia, you could say, are existentially stable. Redness doesn’t ambush or poison or bloom or hang from perilous branches. It makes sense that the availability of information and corresponding cognitive resources would covary with the ‘existential volatility’ of a given informatic configurations (prerepresentational or representational).

What Tononi gets is that red engages the global configuration in a fixed way, one that does not allow the it nearly so many ‘degrees of dynamic reconfiguration’ relative to it as opposed to apples. Okay, so this last bit isn’t so much Tononi as the way IITC plugs into the Blind Brain Theory (BBT). But his insight provides a great starting point.

So what explains the ‘redness’ of red, the raw, ineffable feel of pain? This is where qualiaphiles will likely want to jump ship. From Tononi’s Q-space perspective, a given space (heuristic configuration) simply is what it is – ‘irreducible,’ as he puts it. Thanks to evolution, we inherited a wild variety of differentiating shapes, or qualia, by happenstance. If you want to understand what makes red red, let me refer you to the anthropic principle. It’s part of basic cable. These are simply the channels available when cable first got up and running.

Returning to BBT, the thing to appreciate here is what I call encapsulation. Even though the brain is an open system, conscious experience only expresses information that is globally broadcast or integrated. If it is the case that System 2 deliberation (reflection) is largely restricted to globally broadcast or integrated information, then our reasoning is limited to what we can consciously experience. Our senses, of course, provide a continuous stream of environmental information which finds itself expressed in transformations of aggregate heuristic configurations, representations. With apples we can vary our informatic perspective and sample hitherto unavailable information to leverage the various forms of dynamic reconfiguration that we call cognition.

Not so with red. Basic heuristic configurations (combinatorial prerepresentations or qualia) are updated, certainly. Green apples turn red. Blood dries to brown. But unlike apples, we can never get up and look at the backside of red, never access the information required to effect the various degrees of dynamic reconfiguration required for cognition.

It’s a question of informatic ‘perspective.’ With qualia we are trapped in our neural armchair. The information available to System 2 deliberation (reflection) is simply too scant (and likely too mismatched to the heuristic demands of environmental cognition) to do anything but rhapsodize or opine. Red is too greased and cognition too frostbitten to do the juggling that knowledge requires. (Where science is in the business of economizing excesses of information, phenomenology, you could say, is in the business of larding its shortage).

But this doesn’t mean that qualia can’t be naturalistically explained. I just offered an outline of a possible explanation above. It just means that qualia are fundamentals of our cognitive system in a manner perhaps similar to the way the laws of physics are fundamentals of the universe. (And it doesn’t mean that an attenuated ‘posthuman’ brain couldn’t be a radical game changer, providing our global configuration with cognitive resources required to get out of our neural armchair and ‘scientifically’ experiment with qualia). The qualification ‘our cognitive system’ above is an important one. What qualia share in common with the laws of physics has to do with encapsulation, which is to say, constraints on information availability. What qualia and the laws of physics share is certain informatic inscrutability, an epistemological profile rather than an ontological priority. The same way we can’t get out of our neural armchair to see the backside of red, we can’t step outside the universe to see the backside of the Standard Model.*

But the fact is the kind of nonsemantic informatic approach I’m taking here marks a radical departure from the semantic approaches that monopolize the tradition. Peter, in his Conscious Entities critique of IITC linked above, references the Frank Jackson’s famous thought experiment of Mary, the colour-deprived neuroscientist. The argument asks us to assume that Mary has learned all physical facts about red there is to know while sequestered in a black and white environment. The question is whether she learns a new fact, namely what red looks like, when she encounters and so experiences red for the very first time. If the answer is yes, as intuition wants to suggest, then it seems that qualia constitute a special kind of nonphysical fact, and that physicalism is accordingly untrue.

As Peter writes,

And this proves that really seeing red involves something over and above the simple business of wavelengths and electrical impulses. Doesn’t it? No, of course not. Mary acquired no new knowledge when she saw the rose – she had simply had a new experience. Focussing too exclusively on the role of the senses as information gatherers can lead us into the error of supposing that to experience a particular sight or sound is merely to gain some information. If that were so, reading the label on a bottle of wine would be as enjoyable as drinking it. Of course experiencing something allows us to generate information about it, but we also experience the reality, which in itself has nothing to do with information.

The reason he passes on IITC is that he thinks qualia obviously involves something over and above ‘mere information,’ what he calls the ‘reality’ of the experience. This is a version of a common complaint you find levelled against Tononi and IITC, the notion that information and experience are obviously two different things – otherwise, as Peter says, “reading the label on a bottle of wine would be as enjoyable as drinking it.” Something else has to be going on.

This is an example of a demand I have only ever seen in qualia debates: the notion that the explanans must somehow be the explanandum. Critics always focus on how strange this demand looks when mapped onto other instances of natural explanation. Should chemical notations explaining grape fermentation get us drunk? Should we reject them because they don’t? But the interesting question, I think, is why this move seems so natural in this particular domain of inquiry. Why, when we have no problem whatsoever with the explanatory power of information regarding physical phenomenal, do we suddenly balk when it’s applied to the phenomenal?

In fact, it’s quite understandable given the explanation I’ve given above. Rather than arising as an artifact of the radical (and quite unexplained) disjunct between mechanistic and phenomenal conceptualities as most seem to assume, the problem rather lies with the neural armchair. The thing to realize (and this is the insight that BBT generalizes) is that qualia are as much defined by their informatic simplicity as they are by the information they provide. Once again, qualia are baseline heuristics (prerepresentations): like gensyms, they are defined by the information they lack. Qualia are those elements of conscious experience that lack a backside. Since the province of explanation is to provide information, to show the backside, as it were, there is a strange sense in which we should expect our explanations will jar with our phenomenal intuitions.

Rethinking the Mary argument in nonsemantic informatic terms actually illustrates this situation in rather dramatic fashion. So Mary has, available for global broadcasting or integration (conscious processing), representations (knowledge of the brain as object) leveraged via prerepresentational systems lacking any colour. Suddenly her visual systems process information secondary to light with the wavelength of 650nm.  Her correlated neurophysiology lights up. In informatic terms, we have two different sets of channels–one ‘access’ and one ‘phenomenal’–performing a variety of overlapping and interlocking functions matching her organism to its environments. For the very first time in her brain’s history, red is plugged into this system and globally broadcast or integrated, becoming available for conscious experience. She sees ‘red’ for the very first time.

Certainly this constitutes a striking change in her cognitive repertoire, and so, one would think, knowledge of the brain as subject.

From a nonsemantic informatic perspective, the metaphysical implications (the question of whether physicalism is true) are merely symptomatic of what is really interesting. The Mary argument raises an artificial barrier between what are otherwise integral features of cognition, and so pits a fixed prerepresentational channel against a roaming, representational one. Through it, Jaskson manages to produce a kind of ‘conceptual asymbolia,’ a way to calve phenomenality from thought in thought, and so throw previously implicit assumptions/intuitions into relief.

The Mary Argument demonstrates something curious about the way information that makes it to global broadcasting or integration (conscious awareness) is ‘divvied up’ (while engaging System 2 deliberation (reflection), at any rate). The primary intuition it seems to turn on, the notion that ‘complete physical knowledge’ is possible absent prerepresentational components such as red, suggests a powerful representational bias, to the point of constituting a kind of informatic neglect. We have already considered how red is dumbmute, like a gensym. We have also considered the way deliberative cognition possesses a curious insensitivity to information outside its representational ambit. In rank intentional terms, you could say we are built to look through. The informatic role of qualia is left mysterious, unintergrated, unbroadcast–almost entirely so. We might as well be chained in Plato’s cave where they are concerned, born into them, unable vary our perspective relative to them.

The Mary argument, in other words, doesn’t so much reveal the limitations of physicalism as it undermines the semantic assumptions that underwrite it. Of course ‘seeing red’ provides Mary with a hitherto unavailable source of information. Of course this information, if globally broadcast or integrated will be taken up by her cognitive systems, dynamically reconfiguring ‘K-space,’ the shape of knowledge in her brain. The only real question is one of why we should have so much difficulty squaring these platitudinal observations with our existing understanding of knowledge.

The easy answer is that these semantic assumptions are themselves prerepresentational heuristics, kluges, if you will, selected for their robustness, and matched (in the ecological rationality sense) to our physical-environmental cognitive systems. But this is a different, far more monstrous story.

Ultimately, the thing to see is that Tononi’s Phi is a kind of living version of the Mary Argument. He gives us a brick o’ qualia, a book that fairly throbs with phenomenality, so seating us firmly in our neural armchair. And through the meandering of rhapsody and opinion, he gives our worldly cognitive systems something to fasten onto, information nonsemantically defined, allowing us, at long last, to set aside the old dualisms, and so range from nature to the soul and back again, however many times it takes.


* I personally don’t think qualia are the mystery everyone makes them out to be, but this doesn’t mean I think the hard problem is solved – far from it. The question of why we should have these informatically dumbmute qualia at all remains as much as burning mystery as ever.