Representation, etc.
[continued from Huw Price]
The best expose of current thinking on the problem of representations that I have come across is this position paper, Does Representation Need Reality? appearing in Understanding Representation in the Cognitive Sciences -- A. Riegler, M. Peschl & A. von Stein (eds.), 1999.
After describing the lack of relevance of computational symbol processing or expert systems, i.e. purely representational systems, to real world cognition, the authors go on to describe the other end of the spectrum found in 'connectionist' models:
Not only has symbolic cognitive science come to a dead end (in the context of the effort to explain cognitive processes). The euphoric days of the new possibilities of connectionist networks, including the simulation of cognitive activities and learning, have passed, too. It seems that cognitive science has returned back to -- in Kuhn's sense (1970) -- "normal science" or to "puzzle solving". As an implication of this development, connectionist cognitive science has almost lost itself in (technical) details, such as learning factors, minimal adjustments, and optimizations in learning algorithms or activation functions, and so on. In tackling only these technical "micro-problems", it seems to have failed to address the really interesting, pressing, qualitative, and "big" questions about cognition.
My guesstimate here is that cognitive science has fallen victim precisely to the conundrum that was presented by George Berkley in 1710. Cognitivists cannot live with representations and cannot live without them. What are we to do?
On a purely subjective level our linguistic and visual cognition is highly representational, presumably of an external world. However, when we attempt to computationally simulate symbolic processing, there is no semantic grounding for the symbols other than as mediated by external observer.
On the other hand, when we construct more biologically realistic systems of the neural connectionist variety, we find no stable objective structures corresponding to our subjective representations.
Cognitive science was supposed to be providing a conceptual bridge between the objective and subjective worlds, but after a half century of 'progress' we seem no closer to constructing that bridge. And we are not even talking here about the touchy-feely qualia of the 'hard' problem. We are just talking about the most basic abstract competencies of cognition. In fact there does not even appear to be any conceptual progress in attempting to embed a symbol system in a connectionist system (see below). This seems to be the most likely form that such a bridge might take.
There simply appears to be no piecemeal method for constructing that mainly unconscious background of natural intelligence, which the phenomenologists speak of as being the necessary platform for any particular, conscious cognitive process.
[9/18]
A good summary of the representation problem appears in this article on connectionism (see also systematicity). We see there the controversy between the symbolists and the connectionists. The symbolists can manipulate symbols computationally, but they provide no internal semantic grounding for their symbols. The connectionists could conceivably provide a semantic grounding for symbols by embedding them in a closed behavioral network, but then they are unable to systematically or logically manipulate those symbols which are effectively wired into the network. Thus there appears to be a logical impasse between the manipulating and the embedding of symbolic representations.
(The robust, distributed embedding of a symbol is resistant to its individual accessibility.) (The notion of the brain as a 'syntactic engine' is relevant here.)
To be effective in any mental sense, a symbol has to be abstractable and groundable at the same time. These twin constraints seem physically contradictory. This contradiction then argues against any physical implementation of the mind. Note that this aspect of cognition was once alleged to be the 'easy' problem. One moral is that the global unity or potency of the mind subverts even this crudest attempt at the compartmentalization of its powers.
Connectionism may also be seen as a throwback to the long discredited associationism of the original empiricists.
News flash: it is now purported that dynamical systems theory supplants both symbolicism and connectionism. But, alas, it seems only to be a subspecies of connectionism.
Chris Eliasmith provides further critiques and yet another hybrid physical model of his own.
Then, while searching on neurosemantics I came across SINBAD Neurosemantics: A theory of mental representation. This seems to be the state of the art for neural modeling of cognitive functions. I will peruse it and report back.
[9/19]
Alas, it appears that SINBAD is just a resurrection of the concept of the 'grandmother cell'. A hierarchy of pattern recognition processes in the brain culminates at its highest level in cognizable constructs and models of the world. The firing of a particular cell represents a detection of 'grandmother' and corresponds to our cognition of said individual.
This presents certain logistical problems. First, how do we come to sense the firing of any particular cell or group of cells? Secondly, how do we recognize the semantic significance of any particular cell or group thereof? It would seem that the cells would have to be labeled in some fashion, and that we would then need a label reader. (Back to homunculism.)
Of course we could program a robot using these pattern recognition processes so that it would say 'grandmother' upon any successful instance of grandmother detection. We could do that now with a simple fingerprint matching device. Would this be an instance of natural cognition? Would this justify the publication of yet another paper in 'neurophilosophy'? I would hope not. But what is the logical or philosophical significance of the quantitative difference in complexity or sophistication between two levels of computational pattern recognition? Claiming that a thermostat is conscious of the temperature does not seem to shed much light on natural cognition. It is useful only as an eliminativist gambit.
The wiring of 'grandmother' into the brain hardly seems conducive to the notion of concept formation and manipulation. Again we face the physical contradiction between symbol grounding and symbol use. There ought to be a convenient designation for this problem. I'm just not aware of it.
[Later] I am finding this paper, Radical Connectionism: Thinking With (Not In) Language, to be a good introduction to connectionist concepts of representation.
[9/21]
Neurophilosophy is currently the label of choice for the popularizers of naturalism and materialism. It is worth glancing at some of their books:
Neurophilosophy : Toward a Unified Science of the Mind-Brain -- Patricia Churchland
The Engine of Reason, The Seat of the Soul : A Philosophical Journey into the Brain -- Paul Churchland
Synaptic Self: How Our Brains Become Who We Are
The Emotional Brain: The Mysterious Underpinnings of Emotional Life
etc., etc....
The only philosophy visible in the books typical of this genre is eliminativism. Metaphysical innocence is the order of the day, as the writers launch into guided tours of the flora and fauna in the not quite tamed neuronal jungles of our brains. There are occasional rest stops were we are treated to quickie seminars on the wonderful 'illusions' of the mind such as the self and free-will. If this is the last frontier of science, Colonel Sanders would feel right at home. Speaking of which, has America been entertained by such as the Churchlands since the days of Roy Rogers and Dale Evans? Am I being catty? I spent 30 minutes spell-bound by a 56kps video of Patricia. Obviously metaphysics is lacking in charisma, so it is just too bad that Patricia is playing for the wrong team. The mini-messiah is going to need all the help she can get.
Where are the metaphysicians when we need them? This is pseudo-science raised to a fine art. Apparently, all that one has to do is throw around bits of scientific jargon to become fully absolved of all metaphysical responsibility. It is a brave new world, this neurophilosophy.
Amazon has a plethora of new books touting materialism, and I don't even see where to look for the rejoinders. It appears that the non-sectarian non-materialists have been brow beaten into silence. This is sad. And considering the state of the world today it is emphatically dangerous for there not to be a single visible intellect proposing a transcendental, ecumenical rationalism. Has God forsaken us?
On Amazon the non-materialist field is dominated by John Searle. No other author comes close. But John is very far from theism or idealism, the last I checked. He is simply a conscientious objector to the vocal horde of eliminativists. Otherwise, the intellectual ground between the secular and sectarian is a no-man's land.
However, there are other contrarians, as can be seen on these lists:
[9/22]
Listmania! The Philosophy of Consciousness. This list is thorough and even-handed.
Listmania! Academic books on consciousness and science. This brings in some of the quantum thinking.
Listmania! Consciousness in Physics. And some more contrarian physics.
And the lists go on. I'm going to check these out a bit.
The writers on the side of neurophilosophy have a coherent cosmological story to tell. They are even anticipating the logical end of science as it successfully conquers the mind, its final frontier. It is the masterful wrapping up of the story that began being told with the rise of the secular Enlightenment almost three centuries ago. The only thing missing from this story is a subject, that is the subjective self. Neurophilosophy is the end of the line for objectivity. Beyond this point you have to consider the human self, and then, almost certainly, a cosmic self.
But that last logical step is a big hurdle for anyone trying to think outside the sectarian box. Materialism has become a crutch and a conditioned reflex for anyone trying to break out of the sectarian mold. For those who wish now to think outside the box of scientific materialism, there is approximately zero guidance. The only lamp post is a trans-objective coherence. It is necessarily eschatological. If you don't believe that, then just try thinking about it for two minutes, although most people would find that extraordinarily onerous. There is a final coherence out there that is trying to find a voice. Where is the fat lady when we so desperately need her?
But I promised to check out the lists.
Everyone agrees that there has been an explosion of interest in the mind, and mostly within the last two decades. I don't see a consensus explanation for this upsurge.
All the coherence is on the side of the materialists. They have a linear story to tell, albeit, one without a subject or a resolution. No? Aren't our feeble, genetically burdened organic brains about to be engulfed by a tsunami of silicon? The neurophilosophers, however, are strangely silent on the future, even though their hind-sight is 20/20. They must all be aware of the transhumanist eschatologies, but we are left to speculate as to how discomfited they might be by these nightmarish apocalypses. Their silence ought to be deafening to anyone with even half of an organic brain. But the typically glowing reviews consistently hear no, see no and speak no apocalypse. Someone ought to formally introduce the neurophilosophers to the transhumanists. They might even have to change their tune.
The conscientious objectors to neurophilosophy and transhumanism are scattered all over the map. Even individually they have no consistent story to tell. At best they can marvel in the multi-disciplinary diversity of the topic. A few make allusions to the mystical tradition or to modern phenomenology. That's about it. There is no sense of urgency or sense of any imminent new worldview. Postmodern pluralism is the consensus ethos of these semi-non-materialists. Vision, anyone? What is that??
The cream of this crop may have been Fritjof Capra and Ken Wilber, but they have had nothing new to say for decades. It is mysticism all the way down. If we experience as much as a very ill-defined Turning Point we'll have to consider ourselves lucky bums.
---------------------------
But, stop the press and take a look at Gilles Fauconnier. Is this fellow a non-reductionist or what? How about those mental spaces? And what's all this blending business? Has froggy phenomenology been unleashed upon our hallowed shores of analytical atomism? What hath God wrought? Just when we surmised forsakenness. Is this a wolf in sheep's clothing or a sheep in wolf's clothing? I can hardly wait to find out.
There may be a pot load of constructivism chez Gilles, but, modulo the constructivism, has there not got to be something more than atoms swerving in the dark, or 'neuronal group selection'? Speaking of which, I can't resist quoting an example quotation in 'Blending As A Central Process of Grammar' (1996) that Giles uses to illustrate the blending of mental spaces:
I claim that reason is a self-developing capacity. Kant disagrees with me on this point. He says it's innate, but I answer that that's begging the question, to which he counters, in Critique of Pure Reason, that only innate ideas have power. But I say to that, what about neuronal group selection? And he gives no answer.
This is getting right down to the nitty-gritty! I detect more than a tad of continental ridicule in the choice of this pretentiously reductionistic, un-continental exemplar.
I can't help feeling, despite the heavy patina of constructivism and even of mechanism, that Gilles is leaving the neurophilosophers to choke on his astutely framed phenomenological dust. We are revisiting structuralism, an idealism dressed in a mechanic's coveralls.
It seems that cognitive science is bifurcating into micro (neuronal) and macro (conceptual) components, and that there is scant to nil correlation between the components. If this split becomes formalized it will provide powerful evidence for non-reductionism.
I'm reminded of Roger Shepard's mental rotations (1971) experiment. Not easy for connectionists to explain
[9/23]
If there is going to be a metaphysical fault-line in the life sciences, one place that it might occur is between biosemiotics and neurophilosophy. The conceptualism inherent in biosemiotics is evidently irreducible, but this stark fact, if correct, is not being advertised. One has to wonder about the political considerations that may lie behind the downplaying of this incipient fault-line. I am now just attempting to document it.
This is reminiscent of the age-old debate between nominalism and realism. Abelard's conceptualism was supposed to have been the compromise position, but it is not clear that there ever can be a compromise. The ontological issue is unavoidable, and trying to shift it around is just a shell game. The political hegemony of postmodern pluralism largely hinges on its ability to downplay any and all ontological issues. It is a shell game of truly massive proportions. Postmodernists have a deathly fear of Being and Substance. They are all trying to hide behind Abelard, but the biosemioticians are finding this more difficult. Questions concerning Being strike too close to their home. In any case, the uncovering of the ontological fault-line among practicing, politically aware scientists may prove to have all the delights of a snark hunt.
As I read about the debate I am struck by the explicitly materialistic bias of nominalism. Evidently something does exist, but the nominalists would argue that all existence is reducible to accidental, temporary configurations of matter. But then what about those atoms? I would suggest that where nominalism is most starkly at variance with physics is with the concept (sic) of the photon. As a result of spending their lives with such entities as photons the number of nominalist physicists can be counted liberally on one finger. There are many more nominalist mathematicians, even, than there ever will be physicists. I don't think the biosemioticians are in an intellectual position where than can blithely ignore the collective wisdom of such an influential group of scientific colleagues. As the Sokal affair so amply demonstrated, there is surely a fault line between postmodernism and physics. The biosemioticians are trying to negotiate that line. Lots of luck, fellas!
It is fair to say that in modernity it is the Physicists who are the keepers of the Sacred Ontos. Postmodern iconoclasts resent and fear that office. But what they should really fear are the escaped and wandering ontologists, such as your correspondent. Some of us are not entirely satisfied with mere photons. Our sights are set a bit higher on the Chain of Being.
Quoting now from the good Dr. Ess:
BUT -- thought out more carefully, the Neoplatonic version of realism (the primary "version" of realism available during the so-called "dark ages") in fact lands its defenders squarely in a position which, judged from the standpoint of Christian "orthodoxy," must be considered heretical.
That is, Neoplatonism's insistence on the ultimate reality of "the One" (and thus, on both the ontological and epistemological primacy of "the One" -- i.e., it is both the most real entity and the entity which makes knowledge of all subordinate entities possible) further issues in pantheism. Simply, there is only "the One" and the rest of reality as an emanation of the one. While the rest of reality (including human beings) are in some sense "less real" than "the One" -- in an equally important sense, they also are identical with "the One." In short, all things are God ("pan/theism" = "all/God-ism").
This belief, as a logical consequence of the Neoplatonic version of realism, runs squarely against the doctrinal insistence in "Orthodoxy" on God's transcendence -- i.e., on the irreducible difference between God and God' creation.
Like I often say, I am definitely a heterodox Christian, but I'll wager that's a lot better than nothing at all. You can see why Christians have a deathly fear of Ideas. You let one in and it takes over the whole neighborhood. The Idea is their Eschaton, and mine. C'est la vie, and hasta la vista.
-------------------------------
When I search on biosemiotics & neurophilosophy there is but one substantive hit: Cognitive Sciences In the Perspective of a Unified Theory of Information. It gives a credible account of the conceptual divide between the naturalists and the phenomenalists, with the biosemioticians falling toward the latter. I would have to assume that Fauconnier and his followers should be included in the biosemiotic camp, but they avoid the term.
Hofkirchner makes reference to Capurro's Trilemma: (from a trialog)
[Capurro]: Your conclusion or "solution" of this trilemma is: we go back to the etymological roots (information as "giving form") and we take an evolutionary perspective where qualities can emerge. I call this solution "dialectical informatism", considering it to be a new version of dialectical materialism.
Yes, dialectical materialism is a doctrine of emergentism, and it certainly comes down on the side of realism and objectivism against nominalism. But, again, biological emergentism has to confront the Platonism of Physics. And what is the ontological nature of those emergent entities? How do they relate to and inform matter? In a fashion similar to the laws of physics?
Is there any possibility of a unified theory of information which includes "Capurro's trilemma" as a constituent element of it, and not as something to be eliminated or "solved"? Well, this is a difficult question. Maybe we should take a look at the metaphysics of Leibniz. Leibniz considers reality to have two aspects, namely "monads" and matter. There are no monads without matter (except God), and vice-versa. Monads and matter are folded into the different levels of reality in an infinitely complicated way. This means that it is not possible for us to have a "true" view of all the "steps" faced by unfolding (or "evolution"). This means, roughly speaking, that we are faced with infinite concepts of information, something which cannot be overlooked by any kind of theory.
Now there's a leap from dialectical materialism. Those Continentalists sure are light-footed. How could an Anglo hope to keep up? Or just some intramural gymnastics...[let us not forget Stoppard's 'Jumpers'.]
Dear Rafael [Capurro], in a somewhat pejorative manner you have described the concept of information which Wolfgang and I discussed in "Informatio Revisited" as "dialectical informatism".
...a friendly slap of the text.
I think it constructive to look for information from this perspective, and to work out the differences/similarities between the causality of information processes and that of physical processes. The advantage of this perspective is the unfolding of larger range of analysis than is permitted by physics. It allows for the inclusion of completely different causal relations which cannot be viewed by the natural sciences.
...and then a left hook into the Scientific spleen. Ouch! Do we catch the scent of Blood in the Water? Where is Sokal when we need him?
Information, on the other hand, is able to mediate between cause and effect without having to obey the Laws of Conservation as such. There is no need for uniqueness of the effect. More than that, not even the type or quality of the effect needs to be predictable. One of the main differences between physics and information sciences can thus be seen here. While physics is oriented towards forecasting effects based on some explicated premise, information sciences have to deal with the structure of the unpredictable. At first glance this seems to be a very fuzzy enterprise, but it is in fact the complement of the unique predictability of the physical world. The new cannot be grasped theoretically by physics. If everything were uniquely predictable in a physical environment, no development of the world would be possible.
A second difference can be found between physical and informational processes. Physical units (in particular in classical mechanics) are based on the metaphysical principle of some factor (like mass, force, energy) while the substantialism gets lost in the realm of information. Information in its reified aspect is like a symbol, a metaphor. It is more a pointer to some kind of interpretation process than an object per se. It does not always point to any object of so-called reality. If there is some conscious process needed to interpret the pointer, and thus to create a new piece of reality, the substantiality of information is no longer needed. By the same argument, uniqueness of interpretation is no longer assured; on the contrary, in information processes there must be freedom of choice, on some level there must even be unpredictability, there must be the opportunity for creativity, and producing something new. Here Leibniz's monads could come into the picture. They could be interpreted as an early concept to explain the half of the world which was left out by mechanical materialism, such as processes of the human mind, the soul and creativity. In my opinion there is no need to stick to Leibniz's interpretation of the monads as being divinely created. The concept of information processes would allow for a materialistic [sic? naturalistic? or maybe you're smoking something? Leibniz is turning over in his grave!] explanation.
I've been following the informatics scene for some time, but have not heard them speak quite so forthrightly. Is this something new under the Sun? Does it go back to Johnny von Neumann?
[Capurro] Dear Peter, I am glad that the pejorative undertone of my criticism has become an incentive for, not an obstacle to, our dialogue.
...what a gentleman!
My question is "Where are the limits of such an information science?"
.....
Metaphysics also made a distinction between two forms of immanent causality: a "transitive" immanent causality, where a cause can change or even disappear in the process of (in) formation, and another immanent causality where the cause remains the same although changes may take place. The latter is the case for e.g. human or animal souls. The transcendent cause must be "higher" than immanent causality.
And this trialog progresses from there. It was an email discussion dating back to 1997. Capurro is openly theistic in his comments. This is my first experience of a philosophy of science discussion where the God question is treated matter-of-factly. I need to get a better handle on where Capurro is coming from, and where he might be headed? None of his metaphysics sources are contemporary, nor do I see evidence of any comrades in arms. It is difficult for me to imagine a similar discussion taking place on these shores.
R. Capurro, What is Angeletics? (2000):
Angeletics is different from angelology in that its purpose is to study the phenomenon of messages, independent of their divine origin, or, in other words, it studies this phenomenon within the boundaries of the condition humaine. This does not mean that studies relating to the fields of religion and natural sciences are excluded, but its specific focuses are on messages and human messengers.
...there is another neologism for you.
German philosopher Peter Sloterdijk pointed out that we live in a "time of empty angels" or "mediatic nihilism", in which we forget what message is to be sent while the transmission media multiply: "This is the very 'de-angeling' of current times" (Sloterdijk 1997). The word "de-angeling" stands out, in contrast to "evangeling", the empty nature of the messages disseminated by the mass media
Information science has come a long way, Cherubim.
[9/24]
Biosemiotics confronts materialism:
Levels, Emergence, and Three Versions of Downward Causation by Claus Emmeche, Simo Køppe and Frederik Stjernfelt (2000).
9/17/02