In our first two dialogues, we presented the standard, or “internalist” version of how our conscious experience of the world comes about: very bluntly, it assumes that the brain receives “inputs” from the sense organs—eyes, ears, nose, etc.—and transforms them into the physical phenomenon we know as consciousness, perhaps the single most important phenomenon of our lives. We also pointed out, particularly with reference to color perception, how difficult it has been for scientists to demonstrate how, or even whether, this really happens. Neuroscientists can correlate activity in the brain with specific kinds of experience, but they cannot say this activity is the experience. In fact, the neural activity relating to one experience often seems nearly indistinguishable from the neural activity relating to another quite different experience. So we remain unsure where or how consciousness happens. All the same, the internalist model remains dominant and continues to be taught in textbooks and broadcast to a wider public in TV documentaries and popular non-fiction books. So our questions today are: Why this apparent consensus in the absence of convincing evidence? And what new ideas are internalists exploring to advance the science?
Tim Parks: Riccardo, I know I should be asking the questions, not answering them. But I’m going to suggest that one reason for this consensus is that we are in thrall to the analogy of the brain as computer. For example, a recent paper I was reading about the neural activity that correlates with the sense of smell begins, “The lateral entorhinal cortex (LEC) computes and transfers olfactory information from the olfactory bulb to the hippocampus.” Words like “input,” “output,” “code,” “encoding,” and “decoding” abound. It all sounds so familiar, as if we knew exactly what was going on.
Riccardo Manzotti: We must distinguish between internalism as an approach to the problem of consciousness (the idea that it is entirely produced in the head) and neuroscience as a discipline. The neuroscientists have made huge progress in mapping out the brain and analyzing the nitty-gritty of what goes on there, which neurons are firing impulses in which rhythms to which others, what chemical exchanges are involved, and so on. But you are right, the way they describe their experiments by way of a computer analogy—in particular of information processing and memory storage—can give the mistaken impression that they’re getting nearer to understanding what consciousness is.
When physiologists address other parts of the body—the immune system, the kidneys, our blood circulation—they don’t feel the need to use anything but the language of biology. Read a paper on, say, the liver, and it will be talking about biochemical mechanisms—metabolites, ion homeostasis, acetaminophen poisoning, sepsis, infection, fibrosis, and the like, all terms that refer to actual physical circumstances. Yet, when dealing with the brain, we suddenly find that neurons are processing “information,” rather than chemicals.
Parks: Is this because while we know what other organs are doing—I mean, which physical processes in the body each is responsible for—we’re not sure what all this neural activity is for?
Manzotti: On the contrary. We know very well that neural activity controls behavior, the nervous system having evolved to meet complex external circumstances with appropriate reactions. The question is, did it also evolve to orchestrate an internal mental theater for us—David Chalmers’s “movie-in-the-head”? Or to “process information”? Stanislas Dehaene and Jean Pierre Changeux, two leading neuroscientists, recently claimed that to explain consciousness we must show “how an external or internal piece of information goes beyond nonconscious processing and gains access to conscious processing, a transition characterized by the existence of a reportable subjective experience.” There’s barely a word here that refers to anything physical.
Parks: But is it really not possible to connect the notion of information with chemical exchanges occurring in the brain? Surely when we use a computer the information input is moved along toward the output through electrical signals. Can’t this also be the case with the brain? Hasn’t the philosopher Luciano Floridi claimed that “information is a physical phenomenon, subject to the laws of thermodynamics”?
Manzotti: Listen, when something physically exists and obeys the laws of thermodynamics, then you can find it, concretely. Electrons were predicted to exist and then found. Likewise the planet Neptune and a host of other things. But information, or data, is not a thing. It’s an idea we stipulated because it served a certain purpose, but it doesn’t exist physically, as an entity in its own right in the causal chain. Brutally, when we look inside a computer, or a brain, we don’t see or even detect information. Or data. We see physical stuff: voltage levels in a computer, chemicals in the brain.
Parks: So what you’re saying is that everything that goes on in a computer or in a brain could be fully and properly described without resorting to words like information or data?
Manzotti: Absolutely. Imagine you’re describing a battery; you will have to refer to electricity. It is an indispensable part of the thing. But, when you describe what the brain or even a calculator does, everything can be exhaustively described in terms of causal processes, chemical releases, and voltage changes without ever using the word information.
Parks: But then what is information? How can Floridi make the claims he does? What part can information have in the consciousness debate?
Manzotti: Obviously there is the definition of the word in common use: “facts, data, communicated about something.” The bus leaves at six. Yesterday it rained. The cash machine is out of order. That meaning has been around in English since the fifteenth century.
Manzotti: Then there is the technical IT definition established by the mathematician Claude Shannon in 1949. Shannon was concerned about achieving accurate communication through technological devices and described information as an estimate of the probability that a given channel would successfully transmit words, images, or sound between a source and a receiver.
Parks: Sorry, what do you mean exactly by a channel? I’m lost.
Manzotti: A channel is the physical structure or circumstances that allow two separate events to be connected—the air pressure waves that occur when Romeo utters loving words to Juliet, the wire between a switch that you flip and a light that turns on, or everything that happens between your typing some letters on your phone and someone else reading them on theirs. Essentially, Shannon broke down any communication of data into its most basic constituents, namely a multitude of yes/no questions, that he called bits. Eight bits would make a byte. Information, in this new manifestation, is expressed as a number that tells us how many yes/no questions can be asked and answered through a given channel. A megabyte, for example, indicates capacity for around eight million such questions. Your smart phone requires a few million bits—yes/no questions—to put together, point by point, a photo on the screen. But, there is no internal semantic content, no data or image inside the device, no point along the causal chain where you can put your finger and say, Aha, information!
Parks: Could we say that there is no more information in a cell phone, than there is information in the air between my voice speaking and your ear listening? Or between a radio transmitter and a radio receiver?
Manzotti: You could indeed. Information here is simply the capacity of any channel to effect a causal coupling between two events, speaking and hearing, typing letters and reading them. It is not a thing between those events. If there is no one on the receiving end to hear the voice or read the letters then quite simply there is no information because we don’t have our two events.
Parks: So what do neuroscientists mean when they talk of information processing in relation to the brain? For example, a mouse’s brain when the animal smells a piece of cheese.
Manzotti: Honestly, it is a bit like when we say that the sun goes down. Of course, we know it doesn’t literally go down, but it is a nice expression and it saves a lot of explanation. The problem with the concept of “information” comes when we start to take it literally, as Floridi does. We start to imagine there really is a mental, non-physical stuff called information. A subtle dualism creeps in, as if the brain contained organic material on the one hand and this mysterious, immaterial “information” on the other. In fact Floridi speaks of moving from a materialist vision “in which physical objects and processes play a key role, to an informational one,” as if there were some sphere of existence that is not physical.
However, in its precise scientific usage—and certainly most neuroscientists would see it this way—“information processing” simply means that a physical system—a computer, or the human body, the brain—allows given events to pass along their causal influence to further events. When your mouse recognizes the smell of cheese and moves toward it, the cheese becomes the cause of effects in the olfactory bulb, which themselves cause effects in the lateral entorhinal cortex, which themselves cause effects in the hippocampus, and so on. But there is no immaterial “message” being passed along, no code, or coded representation of “cheese,” existing separately from these organic changes, which are very many and very, very complex. The notion of information and information processing is then built on top of all that causation. It is a kind of shorthand for describing a causal chain so complex as to be beyond any visualization or easy explanation.
Parks: But does the chain end anywhere? Is there a point where we could say, this is where everything arrives, where conscious experience happens?
Manzotti: Alas no. Rather than ending, the causal chain branches and every branch is a constant back and forth with as much feedback as input. In this sense the brain is completely different from IT devices which are always channels leading somewhere, usually to a person who reads off the message that arrives—the second of the two events we talked about.
Parks: OK, let me try to sum up so far. The neuroscientists, for the most part internalists, continue to fill us in on the brain’s exceedingly complex chemical and electronic activity. Meantime the extended computer metaphor that they almost always employ conveys the impression that what is going on is not just organic, but “mental,” that the brain is producing consciousness, storing memories, decoding representations, processing data. So there is a general feeling of promise and expectation, but actually we get no nearer to an explanation of consciousness itself, since we are simply describing, with ever greater precision, what neurons organically do.
Manzotti: I’d agree with that. And perhaps add that maybe people are not unhappy with the situation: we get regular, often melodramatic updates on how marvelously complex we are and how clever scientists have become, while consciousness remains blissfully mysterious. In short, we get to feel very special all round.
Parks: Let’s stick to substance. Aware of this situation, some internalists have made other suggestions. David Chalmers, if I’m not mistaken, has suggested a sort of second and secret life of information—hopefully you can explain. Giulio Tononi has developed an elaborate theory of “integrated information” and “emergence.”
Manzotti: Both Chalmers and Tononi seem to see information processing as a sort of intermediate step toward the conscious mind. I’m not sure this is very enlightening, because if it is hard to imagine how consciousness might “emerge” from neurons, it is even harder to conceive how it might “emerge” from information, which, as we said, is not a physical thing, not “a thing” at all in fact. To put it another way, information can hardly form the basis for a natural phenomenon like conscious experience, which—and we must always remember this—is a thing, a physical phenomenon that we all experience at every waking moment.
Parks: Let’s take the positions one by one. What is this dual aspect of information that Chalmers proposes?
Manzotti: Chalmers agrees that information, as Shannon construes it, lacks any phenomenal character (colors, smells, feelings), or indeed intrinsic meaning, in that a string of zeros and ones in a computer might mean anything. Yet he believes that the brain is basically a computational device crammed with information. So how do all those zeroes and ones, or some neuronal version of the same, become colors, sounds, pains, and pleasures? His solution is that information has a dual aspect—the functional aspect (the zeroes and ones that govern our behavior) and the phenomenal aspect that constitutes conscious experience (colors, sound, itches, whatever). He does not explain why or how this should be and admits himself that his position is basically dualist: information has two sides, one that science can deal with, neurons controlling behavior, and another that is, simply, consciousness.
Parks: At which point we’re back with Descartes deciding what belongs to science and what doesn’t.
Manzotti: Pretty much. Tononi also distinguishes between two kinds of information. Standard information, of the IT variety, and “integrated information,” which we find in the brain and which, like Chalmers’s second, “phenomenal” aspect of information, gives rise to consciousness.
Parks: But what is “integrated information”?
Manzotti: It’s a model for quantifying how much any system brings together, or integrates within itself, the causal influences of the external world. For instance, in a starfish, none of the separate arms knows what the others are up to or what is happening to them. There is no integration at a neural level. Or consider an image on your computer screen; each pixel is quite independent from the pixels around it; you can change one without altering the others. Human beings are very different. Change one neuron and changes will occur in hundreds if not thousands of others. Read about Joyce’s Stephen Dedalus in Ulysses, or Proust’s Marcel in La Recherche and you’ll see that everything that happens to them is immediately mixed up with everything else. Everything connects. A human being is the ultimate causal Gordian knot. You can’t disentangle it. So Tononi’s integrated information is a formula that expresses quantitatively the extent of such integration in different creatures and systems.
Parks: Does that mean it can calculate how those creatures or systems react to a given stimulus?
Manzotti: Perhaps potentially yes, but at the moment no. Tononi’s formula is so complex to compute that even if you posit an unrealistically simple nervous system, it is still beyond the capacity of the most powerful computers to handle. Aside from that, the formula does not explain how or why this super integration might transform itself into the things we experience, for example the color red.
Parks: It seems whichever way internalism turns, however exhilarating its interim discoveries, when it comes to consciousness it reaches an impasse. We have the impression—or simply we’re used to believing—that consciousness is in our heads, that memories are stored in our brains, that there is a world outside and a representation of the world inside, and so on. Yet nothing we have found in the brain warrants this. In our next dialogue, then, I propose that we break out of our skulls and see if there is any other approach to this question that offers more promise.
Manzotti: Very good, but this time I’m going to have the last word! Internalism, like dualism, is, if you’ll allow me the joke, a monster with many heads. We’re going to have to come back to it again and again, to look at dreams, visualizations, hallucinations, and all kinds of other exciting creatures. And some of them will be harder to tackle than the basic premise in itself.
This is the third in a series of conversations on consciousness between Riccardo Manzotti and Tim Parks.