Saturday, May 10, 2014

Watching the Watchers

Signs and theories of consciousness; review of Consciousness in the Brain, by Stanislas Dehaene.

We love it, can't do without it, and wouldn't be without it. We worship it, and warp it with drugs and crazy religion. But what is it? Consciousness is rather hard to describe. Stanislas Dehaene does an excellent job, however, in a status report on what is known about the physical correlates of consciousness, showing that consciousness can indeed be studied and conforms to the "workspace" theory elegantly elaborated by Bernard Baars almost 30 years ago.

Most of what our brain does is unconscious. When we lay eyes on something, imagery just happens. We have no introspective knowledge of the enormous processing that goes into rendering a visual scene. It was only with the invention of computers and the pursuit of artificial intelligence (and of modern animation with its farms of rendering computers) that its scale could be properly appreciated. The refined end-product just pops into consciousness, making it seem oh so easy. What is important to realize, however, is that vast and various processes are taking place in the background, all the time, whether we attend to them or not. Writing in 1988, Bernard Baars paints it as follows:
"A fourth popular metaphor may be called the "search light" or Theater Hypothesis. This idea is sometimes called "the screen of consciousness." An early version may be found in Plato's classic Allegory of the Cave. Plato compared ordinary perception to the plight of bound prisoners in a cave, who can see only the cave wall with the shadows projected on it of people moving about in front of a fire. The people projecting the shadows are themselves invisible; they cannot be seen directly. We humans, according to Plato, are like those prisoners -- we only see the shadows of reality. Modern versions of the Theater Hypothesis may be found in Lindsay & Norman, Jung, Crick, -- and throughout this book. It has been beautifully articulated by the French historian and philosopher Hyppolite Taine (1828-1893): "One can therefore compare the mind of a man to a theatre of indefinite depth whose apron is very narrow but whose stage becomes larger away from the apron. On this lighted apron there is room for one actor only. He enters, gestures for a moment, and leaves; another arrives, then another, and so on ... Among the scenery and on the far-off stage or even before the lights of the apron, unknown evolutions take place incessantly among this crowd of actors of every kind, to furnish the stars who pass before our eyes one by one, as in a magic lantern." Taine managed to combine several significant features in his theater image. First, he includes the observation that we are conscious of only one "thing" at a time, as if different mental contents drive each other from consciousness. Second, he incorporates the Tip-of-the-Iceberg Hypothesis, the idea that at any moment much more is going on than we can know. And third, his metaphor includes the rather ominous feeling that unknown events going on behind the scenes are in control of whatever happens on our subjective stage (cf. Chapters 4 and 5)."
"Thus consciousness involves a kind of a filter -- not an input filter, but a distribution filter. The nervous system seems to work like a society equipped with a television broadcasting station. The station takes in information from all the wire services, from foreign newspapers, radio, and from its own correspondents. It will analyze all this information quite completely, but does not broadcast it to the society as a whole. Therefore all the various resources of the society cannot be focused on all the incoming information, but just on whatever is broadcast through the television station. From inside the society it seems as if external information is totally filtered out, although in fact it was analyzed quite thoroughly by automatic systems. Consciousness thus gives access to internal unconscious resources."

Putting it in my own words, what is the point of consciousness? It seems to be the caboose on the train of thought, seeing things well after they occur, getting highly summarized and schematized reports, whether of perceptions, or even of actions we are doing. The key function seems to be its role of integration, memory, and broadcast. Subliminal (unconscious) data that don't get into consciousness seem to die forthwith. The parallel processing capacity of the brain seems to send the vast majority of what it does into the circular file, never to enter memory, never to be weighed and considered, never to be transformed into language, the coin of much internal thinking as well as social community and extended consciousness. Consciousness seems be the narrow, single-lane road that imposes a top-level, purpose-driven time-sequencing on our brain, so that the most important thing at any given time can be attended to in isolation and with all needed resources, from the large variety available.

The core of the book reports from Dehaene's own work, which uses the visual system to demonstrate a dramatic phase transition between unconscious and conscious mental events. His group, and many others, use a rapidly presented image, which is quickly followed by some other "masking" image, which is is a noisy kind of background that doesn't trip the same kind of consciousness (i.e. subjective recognition) that the discrete test image, of, say, a face, presents. The subject reports whether the first image was seen or not, and can be subjected to various brain imaging tests, bias tests, etc. Typically, if the first image was presented for 60 milliseconds before being superceded by something else, it is consciously perceived and reported. If it is presented for less than 40 milliseconds, subjects never report it, if it is subsequently masked. But brain scans, and other subtle responses, indicate that their visual systems did process that brief image to an extensive degree. It just didn't make it into consciousness.

Experimental method. Top shows locations of surface EEG electrodes. Bottom shows the image presentation sequence, with consciousness-tripping words shown briefly between neutral "masking" symbols. 29 milliseconds is typically too brief to register consciously, if replaced with a subsequent masking image.

When an image does make it into consciousness, the researchers see a diagnostic EEG wave from a high-density array of electrodes on the subject's head, a strong "P3" wave of electrical voltage over the parietal lobe about 300 to 500 milliseconds after the stimulus. This is proposed as a signature of consciousness, which can be studied, manipulated with TMS, and even used diagnostically to evaluate patients in comas or other vegetative states for "locked-in" consciousness. It tracks with high fidelity the subjective state of the observer, rather than their objective inputs. Likewise, when epilepsy patients getting electrodes provide access directly to their brains, the same things are observed at the cellular level. Individual cells can be recorded, generally in the frontal areas, that only fire when, say, a picture of Bill Clinton is subjectively observed, and not when it is presented and ignored, or not recognized. There is much more to be said on the topic, but it is clear that we have some very definite, if rough, physical correlates of consciousness in hand.

From his most recent paper, Dehaene reports more of the same, using EEG electrodes all over the subject's head, which produce event-related potentials, and, after computation, event-related spectral perturbations. Phase coherence also gains significantly when the perceived word comes into consciousness. In the figure below, the beta-band brain waves are tracked, about 400 ms after image presentation, and the Granger correlation presented between discrete brain areas. An enormous leap in activity and coherence happens in the experimental condition, on right.

Point-to-point Granger correlation of beta brain waves, compared between masked (left) and unmasked (right) procedures, of which the latter is subjectively conscious, and shows here dramatically increased phase coherence.

It is very heartening to see the field of consciousness studies take off and make concrete findings, even to the point of saving locked-in patients from gruesome fates. But I was expecting another chapter in the book, where Dehaene would explain how consciousness actually works! Some parts of his model are not problematic. There seem to be many processes, from all the senses, from our memory, and elsewhere, that all get computed and prepared for conscious presentation. There is a continual competition for getting onto that "stage" by way of complicated salience calculations, always weighing the importance of each stream of input. I experience this one-at-a-time limitation all the time when I try to read something at the same time I listen to a podcast or a phone meeting.. it simply can't be done. I can attend to only one language-related task at a time, period.

And, reading a bit between the lines, it is also quite reasonable to posit that our ability to think about something, such as keeping an image in mind, involves recurrent loops of data going between the executive areas (frontal and parietal) to the input module involved, such as the visual areas. It is reasonably well-understood that the lower-level processing areas get very active connections back from the frontal areas both for attention focussing, and for re-generating their inputs for continued keeping-in-mind. Thus the information is not just tranferred from A to B, and deposited. Rather, the content is always re-presented by the originating system, though it can then be remembered, dissected mentally, rotated, expressd in language, etc. This implies a strong neural activity signature in the source areas of sensation as well as frontal areas of the brain when consciousness is active, which is indeed seen, as global connectivity and correlated waves/ firing, vs fragmented, localized patterns of activition under anaesthesia. It also implies some kind of code or lingua franca going back and forth that we really have very little idea of at this point.

I would go farther to propose that autism may be a defect in this one-at-a-time system, where the subject is inundated with excess stimulation, and perhaps unregulated stimulation. In cases where some particularly ornate unconscious calculation is mistakenly injected directly, rather than summarized as is usual, one may see the savant syndromes that are so amazing and correlated with autism. (I'll note that Dehaene posits a simlar theory for schizophrenia, as a fragmented consciousness, with insufficient filtering of images, voices, etc.) To extend the study of consciousness to a serious explanation of its defects and syndromes would be extremely rewarding.

I also agree with Dehaene when he dismisses the philosophical "free will" problems of agency for a putative artificial consciousness with a page or two of rather sharp argument. On the other hand, he waves away the so-called "hard" problem of consciousness with a paragraph of his faith that it will be dissolved in further data. But it is worth grappling with more explicitly. How exactly does a stab of pain affect us subjectively when a robot would not be so affected, as it is not conscious? How do we differ from machines, or if we don't, how can machines be made conscious?

Both Dehaene and Baars have a conviction that if the phenomenon of consciousness is described in ever more detail, at some point an understanding of this issue will naturally emerge. That may be so, but it would be highly beneficial if some theoretical insight could be developed beforehand. That is the position, incidentally, of Giulio Tononi and colleagues, whose recent paper offers a blizzard of obscure theorization about information density and consciousness, which I can't begin to interpret. They offer one interesting observation, however:
"Under special circumstances, such as after split brain surgery, the main complex may split into two main complexes, both having high ΦMax. There is solid evidence that in such cases consciousness itself splits in two individual consciousnesses that are unaware of each other."

Unfortunately, neither approach seems to have really pulled into the station yet, so we can only continue to make up metaphors for what is going on, based on ever more detailed observations of the phenomenon. I would propose something like an engine turning over, based on a basic awake + conscious cortical-thalamic loop of activity, which then pulls in (with brain wave synchrony) perceptual and other loops reverberating to all the other parts of the brain whose processing is available. Dehaene does offer one vision, speaking of the workspace model of consciousness:
"The entire machine is only partially affected by external inputs. Autonomy is its motto. It generates its own goals, thanks to spontaneous activity, and these patterns in turn shape the rest of the brain's activity in a top-down manner. They induce other areas to retrieve long-term memories, genrate a mental image, and transform it according to linguistic or logical rules. A constant flux of neuronal activation circulates within the internal workspace, carefully sifting through millions of parallel processors. Each coherent result moves us one step forward in a mental algorithm thta never stops- the flux of conscious thought. Simulating such a massively parallel statistical machine, based on realistic neuronal principles, would be fascinating...."

But how is that going to work, and are there current systems that provide a kernel for such a dream? I think there are, in the form of the UNIX operating system. It runs countless parallel or near-parallel processes, with its capability of sending jobs to multiple processors and splitting processor time in very fine slices among many users and sub-processes. It can even share results between processes, though they have to be carefully structured. What is needed is a separate computer of a substantially different stucture that interacts with such a unix-type machine, picking off its most significant results to a single constantly running thread of very general rumination on its own conditions, and the most significant happenings internally and externally. It would transmute this into language when possible, internally, if not to external listeners.

Now, one has to realize that all of our mental experiences are the firings of neurons. This is equally true for the plainest visual scene as for the sharpest bite of pain or sweetest apple. Nothing is direct. Yet it feels like there is a watcher inside- a soul, or homunculus, which is the "real" perceiver and target of all the complex data processing. From all we now know, this can not possibly be true. And it would make little sense anyhow, since that homunculus would have to have some mechanism of its own to feel qualia, plus mechanisms to communicate out and back to the rest of the brain system which we know with certainty does at least part of the data processing.

No, the perceiver is part of the system itself, somehow, transmuting all those qualia from electrical buzz into poetry. This is the hard problem- how some portion of the brain system, i.e. consciousness, is the perceiver while most of the rest of our brain and body goes about its business in the dark. And this perceiver is also a powerful director of other activities and agent, interacting with the unconscious systems, getting data, setting alarms, applying focus and attention, getting inundated with streams while trying desperately, during meditation, to get nothing at all.

One gets the distinct sense that is one of the brain's great illusions, like making a pain in the toe feel like it really is in the toe, not in the brain. Or like re-setting time so that we don't notice the ~400 milliseconds it takes for any perception to come into consciousness. Indeed, consciousness is the master illusion, enabling all the others. But how does the magic work? It remains a work in progress.

  • "We find that completely closing the HS to fishing would simultaneously give rise to large gains in fisheries profit (>100%), fisheries yields (>30%), and fish stock conservation (>150%)"
  • Even more lame than the creationists hating on Cosmos: the recent film God's not dead, about a totally realistic philosophy class.
  • But sectarian prayer.. now OK in government, for those grievously oppressed Christians.
  • Speaking of oppressed, some billboards.
  • That is some kind of god they've got...
  • Pakistan, as usual, is a threat to the entire world.
  • Which computer languages are hot, which not?
  • Yes, Virginia, there really is junk DNA.
  • "Not a single U.S. airport is among the top 100 airports in the world."
  • Is Obama for real on climate heating? I hope so.
  • Stanford divests from coal.
  • New England is being set up for an Enron-tastrophe.
  •  Some technical notes on savings gluts. And stagnation.
  • Martin Wolf is turning against capitalism. Maybe the shareholders are not the be-all and end-all. Who really bears the risk of corporate idiocy? The employees do, in very large part. "All those who have stakes in the company that they are unable to hedge bear risks. The most obvious such risk-bearers are employees with firm-specific skills. Human capital is perhaps the least diversifiable and insurable of all our valuable assets. Among all forms of human capital, the least hedgeable are firm-specific skills."
  • A little honesty on the right: The plutocrats are in charge, and thank god!
  • Quote of the week, from Answers in Genesis:
"This episode of “Cosmos” offers a lot of beautiful graphics and special effects, but in the end it should be lumped with fairy tales like Tinker Bell or Shrek. Yet Bible-believing creation scientists who are willing to look at the world through the history provided in God’s Word without the prejudice and blind ignorance can “read” in earth’s geology a true and exciting account of our history. And biblical truth is a much better account than fiction."


  1. This comment has been removed by the author.

  2. Burk

    Really enjoying your posts.

    re: consciousness-what if it doesn't exist per se, but rather is an emergent property? Like, say, life:


  3. Hi, CC- Thank you for your comment! I don't think the essay you point to really implies that life doesn't exist, only that it can not be easily defined or differentiated from non-life, so we live in a more holistic reality than our tendency to compartmentalize would have us believe. Likewise, consciousness may be impossible to pin down to the point where the hard problem disappears entirely. But that wouldn't mean that consciousness doesn't exist.

    Anyhow, the piece is very nice- very appropriate. The only thing I would add is that with a bit more care, I think we can define life. A computer program could be called alive, but only in its context of a hosting computer, and electricity, and all the other paraphernalia required to make it work. When we are looking for life on Mars, we are looking for material entities that can reproduce and go through Darwinian evolution, using the materials (and energy) at hand, in the wild. If we found life forms from Jupiter, which had to use some breathing apparatus to survive on Mars, we wouldn't complain, but you get the idea. Computer programs are not going to make this cut. They, like viruses, are entirely dependent on a context of other pre-existing life, which is what performs that magical transmutation from inanimate to animate.