Review of Radiant Cool, by Dan Lloyd
New Scientist, 13 December 2003, 36-39
“Nobody expects the Spanish Inquisition” says our philosopher hero, Miranda Sharpe, to her empty-headed cat, Holly Golightly. Indeed they don’t. And nobody expects to read an exciting crime thriller that is set in the world of cognitive science, peopled by real philosophers who care about the mysteries of existence, and claims to present a new theory of consciousness.
The author, Dan Lloyd, a neurophilosopher from Trinity College in Hartford, Connecticut, has burst onto the consciousness scene with a book that is both a gripping story and an intellectual challenge. He even appears in his own plot, portraying himself (or his namesake) as a polite, if dull, middle-aged cognitive scientist who creates the best-ever website on consciousness and lives in a barn with avocado kitchen appliances.
The story begins one day at 6 a.m. when graduate student Miranda slips unnoticed into her Professor’s office and finds, to her horror, that the gruesome Max Grue is lying there, slumped over his keyboard. Asleep? Sick? Dead? She grabs what she has come for, a bright red folder labelled “Consciousness”, and runs.
The folder contains many surprises, as does the creepy female therapist, the sinister Russian forensic data scientist, and Gordon, the nerdy fellow grad student. So does Miranda’s research on “The Thrill of Phenomenology”.
It has to be said that not many people find phenomenology thrilling. This daunting philosophical tradition is based on the work of German philosopher Edmund Husserl (1859-1938), who advocated exploring consciousness by suspending judgement and looking directly into immediate experience. It has recently become very trendy in consciousness studies, perhaps because it offers the hope of reconciling private conscious experiences with the study of the brain — but phenomenology is notorious for its impenetrable language and slippery concepts.
Yet Miranda loves it. One morning, finding herself alone in charge of Grue’s class on “The Mystery of Consciousness” she is confronted by a Barbie-faced sophomore who complains “I don’t get any of this at all” — an understandable reaction from a student faced with Husserl.
Miranda rises to the challenge. She explains to the class why appearance is reality, why meaning takes time, and why the mind is a text. She even explains that “superposition” has nothing to do with quantum physics and Schrödinger’s cat, but refers to Husserl’s idea that conscious experience is always heaped up with meanings — every moment of awareness is a pile of interpretations all in “superposition”. A single state of mind is layered with harmonics of meaning — yet somehow remains one experience.
You might dismiss all this as wordy waffle — and the book along with it. But there are good reasons not to. One is that Miranda really does understand why consciousness is such a mystery. Another is that Max Grue, and by implication Grue’s creator Lloyd, claims to have built a novel scientific theory that bridges the gulf between neuroscience and phenomenology.
So, first to the mystery. At the start of the twenty-first century, there are hundreds of books on consciousness and dozens of writers who claim to have solved it. One of the most frustrating things is how few of them appreciate just how deep the mystery is. Daniel Dennett, a philosopher at Tufts University in Massachusetts and author of Consciousness Explained, defines a mystery as something “that people don’t know how to think about — yet.” He calls consciousness “just about the last surviving mystery”.
The trouble lies with subjectivity, or “what it is like” being me now. For example, I am right now having the experience of sitting in this room, with all its many sights, sounds and feels. My experience seems to be private, fleeting, ungraspable and utterly undeniable, and this is what we mean by consciousness. But how can this subjective experience relate to the objective world around me and to the physical brain inside my head? Real, physical things like rooms and brains seem to be of a completely different order from subjective conscious experiences.
We face what Australian philosopher David Chalmers, of the University of Arizona, calls the “hard problem” — the impossibility of seeing how the activity of brain cells can possibly give rise to subjective experience. In spite of dramatic developments in neuroscience, there is still this fathomless abyss between the objective and subjective worlds. We can study the conscious experiences of “mindspace” and the neural events of “brainspace” but we never seem able to map one onto the other.
Half-baked theories of the “consciousness is a fifth dimension” or “consciousness is a spiritual force” type miss this point completely. But so do the many scientific theories that locate consciousness in one part of the brain, equate it with a particular pattern of neural firing, or reduce it down to quantum levels inside minuscule cellular structures. In every case the mystery remains because it is impossible to see why that particular process or this particular brain area is conscious while all the rest are not. In fact just about every theory we have fails utterly in this way and leaves the mystery untouched.
So it is refreshing to find that Miranda and her fellow fictional characters do appreciate the problem. When Gordon, the neural network building nerd, makes the banal observation, “It’s all neurons. That’s all”, Miranda is quick to challenge him. How can a neuron be conscious? How can a physical piece of brain be a private experience? “What is it about neurons that makes them the medium of superposition?” she cries, “What is it about the neurons that makes seeing and smelling different?”. Poor old Gordon cannot answer, but Max and Miranda have been secretly sketching out what a viable theory of consciousness would need to do.
They explain their aim like this: it may be true that “It’s all neurons” but that statement is opaque because no one can understand how it could be true. The ultimate theory of consciousness must be a transparent theory that makes it obvious how consciousness can be the activity of neurons. In such a theory every description in “mindspace” will have an equivalent description in “brainspace”. So they set about using phenomenology to map out the nature of conscious experiences and then trying to fit this with neuroscience — eventually creating the theory on which the plot depends.
So how well does this novel theory fare? Lloyd describes the act of writing a novel as “the most thorough thought experiment” implying that the story might help reveal the implications of his theory. Yet the novel itself proves insufficient even to explain what the theory is. So Lloyd adds over a hundred pages of helpful appendix to do the job. Here we are given a tutorial on some difficult concepts — recurrent neural networks, high resolution brain scanning, multi-dimensional scaling, and the practice of phenomenology, all of which are necessary for understanding his theory.
Much of the explanation deals with the experience of time — a favourite preoccupation of phenomenology. Rather than assuming that consciousness consists of a sequence of distinct, separate moments happening one after the other, Husserl stressed that every Now carries within itself the shadow of what is just past and the expectation of what comes next. So not only is conscious experience always heaped up with superimposed meanings, but each Now is a confluence of retention, presence and protention. This is Husserl’s famous tripartite structure of time, and was this phenomenological “Now” that Lloyd wanted to be able to describe in brain terms. He would then have achieved his mapping of mind to brain and thus his “transparent theory of consciousness”.
The task, then, is to find a way of describing what the brain does that fits phenomenology. Lloyd claims that conventional theories cannot do the job but that simple recurrent neural networks can. Neural networks are computer simulations of the way real brains might work. Typically they consist of several layers of highly interconnected units each of which represents a neuron; an input layer, an output layer and some hidden layers in between. The strength of the connections between the units changes with new inputs, and the state of the whole complicated network determines what its output will be. These artificial neural networks have been highly influential in cognitive science.
In 1990, Jeffrey Elman, Professor of Cognitive Science at the University of California, San Diego, proposed a new sort of “recurrent network” that has an extra “context layer”. This copies the most recent state of a hidden layer and then presents it alongside the next input. So, in effect, the network enfolds both past and present information. If brains are like this, claims Lloyd, a description of how they work might just match the phenomenological description of “now”.
To find out, Lloyd uses networks like these and measures their activity, but their behaviour is so complicated that it can only be described in terms of a vast multidimensional space. So he uses a mathematical technique known as multi-dimensional scaling. This reduces the mathematical space to just two or three dimensions, making it much easier to understand what is going on.
Playing with multi-dimensional spaces like these provides fun for the novel. Through her humble laptop, Miranda is drawn into the fictional Dan Lloyd’s “best-ever” consciousness website, the Labyrinth of Cognition, swooping through a virtual world of spaces of the mind. “This is what I am.” she cries, “Miranda Sharpe, an ever-shifting pattern in the dark.” She falls into multiple states of consciousness until she becomes “the brain from the brain’s own point of view.” This is where brain-map and phenomenology finally fit together. This is where “brainspace is also mindspace.”
It’s a fun story, but does the theory really apply to the brain? To find out Lloyd would need to apply the same mathematical analysis to data from human brains — and he would need lots of data. Fortunately this is available. At the fMRI Data Center at Dartmouth College in Hanover, New Hampshire, researchers have been depositing vast amounts of data from brain-scanning experiments and this is what he used. The simple artificial net was complicated enough but this meant another leap in complexity – Lloyd compares it to swapping a pair of binoculars for the Hubble Space Telescope. He took the masses of fMRI data, applied the same multi-dimensional scaling techniques, and hoped that Husserl’s phenomenology would predict what he found.
Flowing with time
Simplifying drastically, the argument goes like this. According to phenomenology, temporality is in everything experienced. So if it can be observed in the fMRI data it should be seen in a wide variety of tasks and across many regions of the brain. Phenomenology also describes experience as in continuous flux and temporality as monotonic — that is, it always goes in one direction. So change in the brain should also be seen always going in one direction. Put more concretely, a person’s fMRI data should show that as their brain states change, the difference between successive brain states should increase continuously over time rather than the brain returning again and again to a similar state. This is, indeed, what Lloyd found. The brains, he said, were “flowing with time”.
Success? Maybe not. David Rose, a Psychologist at the University of Surrey, says it’s “a nice idea but it’s not going to work in practice”. The fMRI data come from changes in blood flow which have a resolution of a few seconds and so are much too slow to track the phenomenological structure of time. As Rose explains “Lloyd has not shown the ‘tripartite’ nature of phenomenality reflected in the brain – he has just shown that taking an instant in the brain’s flow of activity enables you to predict the brain’s state 2 or 3 seconds into the future or the past — a trivial point to make.”
“Scientifically it’s nonsense” says Chris Frith, of the Functional Imaging Laboratory at London’s Institute of Neurology, because Lloyd has not done the necessary controls. “He has shown that the conscious person has a brain with certain temporal properties, but he hasn’t shown that an unconscious person has a brain that doesn’t have these temporal properties.” Without such a comparison we can draw no conclusions about consciousness.
For John McCrone, author of a book on consciousness called Going Inside, the ideas are not wrong, but just vague and naive. “The way Lloyd uses multi-dimensional scaling may be new and interesting but I don’t think it confirms Husserl’s phenomenology any more than many other ideas.”
This is the crunch. Lloyd may appear to have achieved a scientific success — that is, to have made a prediction from his theory and confirmed it with real data — but this same prediction could have been made in countless other ways without recourse to phenomenology. In essence Lloyd claims a symmetry between mind and brain — that you can’t step in the same stream of consciousness twice, and you can’t have exactly the same brain twice.
But surely these are obvious. On the mind side, ordinary introspection is enough to reveal this simple truth. On the brain side, the notion of continuous change is inherent in everything we know about this real live, wet, and growing organ. Its synapses are always changing and its tiny dendrites are always slightly growing and shrinking. Change in brains really is one way. So Lloyd’s findings are not surprising.
The fictitious philosophers were after a transparent theory. “It’ll be about neurons,” explained Miranda, “but once we get it, you’ll be able to interpret the neurons asconscious states.” If the real Lloyd’s theory does this, it should be obvious how the ever-changing brain gives rise to ever-changing experience. Yet it isn’t. The transparent theory is as far off as ever.
Could it be a step in the right direction? Philosophers such as Dennett and Patricia Churchland of the University of California at San Diego believe that once we really understand the brain there will be no “mystery of consciousness” left over. On this view, Lloyd’s discoveries may be a useful step toward such a theory. But others disagree. For Chalmers all this detailed study of the workings of the brain is part of the “easy problems” and does not touch on the real mystery of consciousness — the “hard problem” of why there is experience at all.
Whichever side you are on, Radiant Cool is fun, and philosophers and neuroscientists will love the jokes and allusions. There are flitting bats and brains in vats. There are “really awful” simulations that, like the real world, appear convincing. And there is a terrifying brain-machine that produces “blindsight” to order. With one zap the helpless, strapped-in victim claims to be blind yet she can still guess what is in front of her.
If the in-jokes pass you by, don’t worry. Radiant Cool is a terrific story and a great read. So do enjoy it, but don’t expect that by the end you will understand the mystery of consciousness.
Radiant Cool: A Novel Theory of Consciousness by Dan Lloyd is published this month by MIT Press
Susan Blackmore is a psychologist, writer and lecturer based in Bristol. Her latest book, Consciousness: An Introduction, was published in June by Hodder Arnold H&S