Bernard Baars, In the Theater of Consciousness. Oxford U. P. 1997. Chapters 1-2. [Comments in square brackets are my own thoughts.] Preface ^^^^^^^ Current [still!] ferment in consciousness research; Essential ideas of Global Workspace theory; relation to the work of Newell, Simon, Anderson (as much AI founders as CogSci icons!); also Schacter, Crick, Gazzaniga, Minsky Prologue (p.1-7) ^^^^^^^^^^^^^^^ Consciousness and unconsciousness; Focal consciousness; f-o-c-u-s; Immediate reportability, outside verifiablity (of correlates!) The theater metaphor; Plato's shadowy Cave (which really made a rather different point, about the dependence of our conception of the world on our perception of it), & Crick's thalamic spotlight; "Seven degrees of separation" beween the 30-70 billion neurons (hign connectivity -- thousands of dendritic inputs per neuron) Consciousness as the "publicity organ (mechanism)" of the brain, providing global access/dissemination of information and global coordination & control. P A R T I : What evidence is relevant to consciousness? Ch.1: Consciousness as a variable (p.12-35) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Scientific understanding of a concept requires that it be understood as something that can be present or absent (presumably to various degrees), and depends on various other variables in a predictable, verifiable way. Much of the book can be thought of as elaborating the idea of consciousness as a variable. In physics: gravity as a variable unified the apple and the planets; In cognitive science: the idea is to seek out mental states and phenomena that are similar in many respects but differ as to the involvement of consciousness ("contrastive phenomenogy"); and to relate the involvement of consciousness to neural processes; [I would also want to study the "logic", or "geometry" of qualia, i.e., the phenomenal experiences, themselves, even ones that are equally "conscious"; Galileo gained an understanding of motion in the Earth's gravitational field (uniform acceleration) before Newton grasped the variablity of gravity.] The early idea (Descartes, even William James) that all mental events were conscious -- this seemed like the very essence of "mental". 70 years of neglect of consciousness by behaviorists (1910 (death of James) - 1980). Their "anti-subjective" scruples caused them to ignore the elephant in the room. Much evidence of unconscious cognition (p.17-18): - priming by briefly presented words (but only single words) - sentence processing mostly unconscious; upside-down reading much more conscious than normal, trying to sort out q's from p's, etc.! PET scans (in the Prologue) show the contrast in processing activity (via capillary blood supply) for untrained and trained Tetris playing; makes plausible that more consciousness implies more neural activity, esp. in sensory projection areas (and certain other areas), and more attention to something implies greater neural activity in the associated brain regions for that "something". He wants to get at qualia (p.21-22) by contrasting present perception with past perception e.g., when a painter is at work, trying to get a blue that is "just-so" [imagine a studio artist, for whom "just-so" relates what s/he is creating to an unconscious conception/recollection, rather than present objects or scenery]. Two-channel experiments (p.22-7) ~~~~~~~~~~~~~~~~~~~~~~~ Dual-channel audio streams, one in each ear: we can shut out one from consciousness (more or less), attending to the other. Nicely illustrated with Donald Broadbent's dual-channel text, MARY paper HAD A brick cream LITTLE LAMB morning ITS FLEECE day brick WAS dot WHITE ... etc. fail to recognize repetition of "brick" when attending to upper case portions -- so less conscious of the lower-case stuff. This is still often interpretetd in terms of "selection" with no allusion to consciousness. But consciousness correlates strongly with reportability. Yet the "ignored" text is being at least lightly processed, since words like "danger!" or our name will catch our eye (p.25) Donald McKay's experiments showing that unconsciously processed words shift interpretation of consciously processed ones, "John and Mary walked by the bank" {money, water} (break-through effect). [was also discussed by Blackmore] Consciousness and coma (p. 27-31) ~~~~~~~~~~~~~~~~~~~~~~ Brain can sustain heavy damage (even hemispherectomy) without coma, but oxygen starvation of reticular formation (involved in arousal and motivation) in brainstem causes coma. ["reticular" = lattice-like] Karen Quinlan: coma caused by damage to both thalami (adjacent, egg-shaped, midbrain structures). The thalami are central relay stations. They contain centers (nuclei) each with inputs from some sensory system (eyes, ears, etc, and output to the appropriate region of the sensory cortex (e.g., visual cortex at the back, touch on top, hearing on the sides); or, inputs from the motor cortex and output to the muscles (eventually). In addition, widely projecting fibers go from the "intralaminar nuclei" (ILNs, smaller nuclei wedged between the large relay nuclei) to the cortical dome; they *trigger cortical arousal* (observable as the quicker irregular waves one can record with electrodes on the scalp). If the ILNs are damaged on both sides (as with Quinlan, because of blockage of the artery that supplies blood to both sets of ILNs) => coma. The ILNs, together with the reticular nucleus, may generate a regular waveform (40 hertz) that coordinates and "binds" activities in various areas of the cortex into a coherent conscious experience (Rudolfo Llinas). Or they create a coherent feedback (reentrant) loop that elevates its content to the conscious level (Gerald Edelman). [A 2002 article, http://www.general-anaesthesia.com/thalmretic.htm, suggests that the reticular nucleus, a thin upper layer on the thalami with thalamic projections only, serves to *inhibit*/*regulate* information flow between the thalami and cortex.] Animal consciousness p.31-3) ~~~~~~~~~~~~~~~~~~~~ Animals are very probably conscious, at least in the basic sense of having conscious states -- vertebrates all have the reticular formation, thalamus, and thalamic nuclei. However, only mammals have a cortex, which is particularly well-developed in primates and humans (who have a particularly prominent frontal cortex, crucial in abstract thought, planning, voluntary action). [I recall reading that functions similar to the cortex may be performed in birds like crows and parrots by the "hyperstriatus".] Baars also lists the many behavioral similarities between the animals and us: response to threat and pain, food, mate selection and sex, care of young, attention to new stimuli, etc. Humans have unique long-term planning, abstract thinking, imagining, and language abilities, but we are not sharply distinguished from the animals. [But it seems to me that language, theory of mind, "recursion" are very important and distinctive attributes.] He concludes (p.34) by rejecting the notion that while all of this evidence bears on the neural correlates of consciousness, it doesn't explain consciousness (the *character* of subjective experience) itself. [I think he's just not very interested in, or impressed by, the philosophical question of why all this neural activity would be accompanied by subjective experience at all. He finds it sufficient to say, WE CAN PREDICT CONSCIOUS EXPERIENCE FROM NEURAL ACTIVITY; WHAT MORE DO YOU WANT? Well, I do think there's at least the *general* problem of saying under what conditions *any* intelligent agent has particular kinds of experiences, E.g., no amount of brain science seems to tell us whether or not a robot could experience pain, and if so, just what capacities or architecture would imply this. In fact, there *are* AI systems based on blackboard architectures resembling Baars' global workspace; but no-one seems to think they have any conscious experiences; well, except perhaps Stan Franklin concerning his IDA system.] P A R T II: The theater of consciousness C.2: The theater stage has limited capacity but creates vast access ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (p.40-61). - The dependence of global visual perception on "sampling" (fixating briefly) on some spots in a scene (he considers a painting) - foveal region only 4 degrees across, with limited peripheral awareness (e.g., separated digits on a line of a page) - large amounts of knowledge potentially made available by whatever is focused on, and thereby rendered conscious (e.g., the word "guesswork" -- a 100,000 word lexicon is accessed, meanings and related words become available, hundreds of vocal tract muscles may become deployed to utter the word, etc.) The theater model (p.41-48) ~~~~~~~~~~~~~~~~~ - The STAGE represents our working memory (low capacity -- 7 or so chunks of inner speech, 4 items of visual imagery; 10 second span). [E.g., you can remember positions of major landmarks/ obstacles in a room for about that long after turning off the light.] - The SPOTLIGHT is what we are conscious of (attending to); whatever is in the spotlight is "visible" to the entire audience, as well as backstage personnel - The ACTORS are various processes vying for the spotlight; - The "sets" and backstage personnel invisibly provide an interpretive context for the "show" and direct it. - The AUDIENCE consists of numerous specialized processes or capabilities that operate unconsciously: lexical lookup, utterance or scene interpretation, episodic (autobiographical) memory, semantic memory (generalities), declarative memory (particular facts), implicit memory (skills), problem-solving specialists of all sorts, etc. - The audience members are interconnected, can form teams/alliances to do their work or get certain actors into the spotlight (make certain processes conscious), and they learn automatically somehow, merely by what is currently in the spotlight (being attended to); "Today's automatic processes emerged from yesterday's effortful and elaborately conscious projects." [One point that bothered me a bit is that the concept of an "audience" suggests relative passivity; we tend to think of the main "action" as being the events on the stage, when in reality it is the audience -- the unconscious processes -- that is getting all the organism's work done! Also it's unclear so far "who directs the director" (the "self"), who is supposedly controlling stage events from behind the scenes. I think somehow the director must correspond to planning and evaluation systems, where the latter evaluate how well the entire "show" is going (and how well it will be doing, given the current plans) and makes further planning and control decisions on that basis. But much of the planning is unconscious (e.g., think of planning what to say, or how to avoid obstacles while walking), and evaluation seems even more dependent on unconscious processes (though it, too, can be raised to consciousness in the case contemplated actions/decisions whose effects are difficult to foresee), and so again must draw on the activity of a lot of audience members (special processes). SO: there really must be a kind of integration of activities between the audience and the director (and other backstage personnel), and that notion also strains the theater metaphor.] Spontaneous problem solving (p.49-52) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Baars points out that all that seems required for problem solving is to bring the problem to consciousness (put it in the spotlight); appropriate experts in the audience will go to work on it and (somehow induce the director to?) bring the solution into the spotlight of consciousness (or perhaps bring subproblems to consciousness, so that the original problem is broken down into easier ones). - Poincare's observations about solutions springing to mind after incubation in the unconscious. [Hadamard had similar observations, and also studied imagistic vs symbolic styles of problem solving.] - Mozart's reliance on mostly (but not entirely) unconscious processes Limited capacity but vast access; the vast unconscious brain (p.54-7) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Psychologists have tended to focus on the limitations in cognitive functioning and conscious awareness; neuroscientists have tended to focus on the activity of the brain as a whole, which is extremely rich, though mostly unconscious. Discussion of cerebral architecture: - the 30-70 billion neurons, each making thousands of synaptic connections - the 6 layers of the cortex, the outermost densely networked, the innermost most closely associated with consciousness - the major organs in the midbrain, esp. the thalami (relay stations and arousal to conscious state) - the corpus callosum, and the front to back "cables running underneath the hemispheres" - 1/10th of a second to become aware of something Why is consciousness so limited? - might not multiple foci of attention be advantageous for, say, a creature drinking at a waterhole, also watching for predators, also watching its offspring? - [His answer is unclear to me; he says it's a trade-off issue, and the benefit of undivided attention is greater capacity for learning. But why is that? Perhaps because if you attend to two things at the same time, and the features of these two things are broadcast to the brain as a whole, then it's unclear how the brain, as a whole, could come up with two different coherent responses to the two different things, or learn simultaneously about those two things. Robots might have much less trouble with divided attention, because their memories and processes can more easily be compartmentalized -- they operate on the basis of addressing, not on the basis of global broadcast.] Interesting facts about biofeedback and learning: - Control of just about any neuron or neural system can be learned, as long as successful control signals are translated into consciously perceptible sights, sounds, etc. [This suggests to me a uniform learning technique that involves trial and error firing of neural subsystems, with "reward" and "punishment" of those systems that are found to be contributing positively or negatively (respectively) to an "accidentally" correct response (e,g, hearing a "click" sound indicating muscle fiber firing); i.e., *backpropagation* of some sort. Baars instead alludes again to the theater metaphor, but I think "I see and therefore I do" as a theory of what the "audience members" do doesn't quite cut it.] - The *content* of what is learned generally remains unconscious (e.g., language learning) Did evolution invent theaters before people did? (p.61) [Well, he's taking the metaphor a little too seriously for me here. But I suppose it's fair to say evolution invented complex control and learning via centralized gathering of key information and its global dissemination. One thing that's missing here is any word about ABSTRACTION processes that produce the information worthy of global dissemination. I think much of what the brain does can be viewed as abstraction and summarization of (originally) vast amounts of detail. Again, the theater metaphor is not helpful, perhaps even misleading, in this respect.]