People have asked me to comment on David Gelernter’s essay on minds and computers in the January issue of Commentary. It’s written with Gelernter’s characteristic brio and clarity, and naturally I agree with the overall thrust of it. But it seems to me that Gelernter does not quite get to the heart of the problem with the computer model of the mind. What he identifies, I would argue, are rather symptomsof the deeper problems. Those deeper problems are three, and longtime readers of this blog will recognize them. The first two have more to do with the computationalist’s notion of matter than with his conception of mind.
As I have emphasized many times, most participants in the debate between materialism and dualism, on both sides, simply take for granted a conception of matter inherited from Galileo, Descartes, Newton, and the other early moderns. On that conception, matter is essentially devoid both of teleology and of the qualitative features that common sense attributes to it. That is to say, there is, on this view of matter, nothing inherent to it that corresponds to the “directedness” toward an end (or “finality,” to use the Scholastic jargon) that the Aristotelian attributes to all natural substances. Nor are secondary qualities like color, sound, odor, etc. as common sense understands them (that is to say, as we “feel” them in conscious awareness) really out there in matter itself. What is there, on this view, is only color as redefinedfor purposes of physics (in terms of the surface reflectance properties of objects), sound as redefined (in terms of compression waves), and so forth. Matter on this conception is exhaustively describable in terms of the quantifiable categories to which physics confines itself.
Now for the Aristotelian, the point isn’t that the moderns’ conception of matter is wrong so much as that it is incomplete. The trouble is not with thinking of matter the way Galileo, Descartes, and their successors have, but with taking this to be an exhaustive conception, as something other than a mere abstraction from a much richer concrete reality. And if it is taken as an exhaustive conception, then a Cartesian form of dualism is hard to avoid. For to say that matter is essentially devoid of qualitative features like color, sound, taste, etc. and that these exist only as the qualia of conscious experience just is to make of qualia something essentially immaterial. And to say that matter is essentially devoid of anything like “directedness” or “finality” is ipso facto to make of the “directedness” or “intentionality” of desires, fears, and other such states also something essentially immaterial. Cartesian dualism was not a rearguard reaction against the early moderns’ new conception of matter, but on the contrary a direct consequence of that conception. (I addressed this issue at length in my series of posts on the critics of Thomas Nagel’s Mind and Cosmos, and it is a point Nagel himself has also emphasized.)
This brings us to the first of the three deep problems with computationalism. Computationalists, like materialists in general but also like Cartesians (though unlike us Aristotelians), take for granted the broadly Galilean or Cartesian conception of matter. Hence they will never in principle be able to fit the qualia definitive of consciousness into their account of the mind, since they are operating with a conception of matter that implicitly excludes the qualitative from material processes from the get go. Their accounts of the qualitative will therefore always in effect either change the subject or amount to a disguised eliminativism. This result can be avoided by giving up the assumption that the Galilean/Cartesian conception of matter really captures all there is to matter, but this would amount to abandoning materialism in favor of one of its rivals (if not Aristotelianism, then neutral monism, panpsychism, or the like).
The second deep problem with computationalism is that on a Galilean/Cartesian conception of matter, there can be nothing inherentto the material world that corresponds to notions like “information,” “algorithm,” “symbols,” and the like. For these notions smack of the “directedness” or intentionality that the Galilean/Cartesian conception denies to matter. The notion of a material “symbol” could be relevant to explaining mental phenomena only if it is aboutsomething; the notion of “information” could be relevant to explaining thought only if it entails semantic content; and so forth. Yet there can be no such thing as aboutness, semantic content, or the like in a material world utterly devoid of “directedness” or “finality.” Hence the key notions of computationalism can on a Galilean/Cartesian conception of matter at best be regarded only as observer-relativefeatures of the material world rather than intrinsic to the material world, and are, accordingly, not available as ingredients in a scientific account of the mind. This is a point John Searle made in his 1990 article “Is the Brain a Digital Computer?” (in a line of argument that is distinct from, and deeper than, his better known “Chinese Room” argument). Similar arguments have been made by Saul Kripke, Karl Popper, and others. (I develop the point in The Last Superstition, and have discussed Searle’s and Popper’s arguments here, and Kripke’s argument here.)
Here too the problem can be avoided by abandoning the Galilean/Cartesian conception of matter. But to regard something like information and algorithmic processes as intrinsic to material substances is precisely to return to something like an Aristotelian conception of the world and its commitment to formal and final causes. (Rightly understood, that is -- not the crude caricatures of formal and final causality usually attacked in discussions of these matters.) As the neuroscientist Valentino Braitenberg once put it:
The concept of information, properly understood, is fully sufficient to do away with popular dualistic schemes invoking spiritual substances distinct from anything in physics. This is Aristotle redivivus, the concept of matter and form united in every object of this world, body and soul, where the latter is nothing but the formal aspect of the former. The very term ‘information’ clearly demonstrates its Aristotelian origin in its linguistic root.
Indeed, I would say that something at least likecomputationalism, if conjoined with an Aristotelian conception of matter, might shed considerable light on the mental lives of non-human animals. However, this would still leave untouched what is distinctive about human beings -- our capacity to form abstract concepts (as when we form the concepts man and mortal), to put them together into judgments (as when we judge that all men are mortal), and to reason from one judgment to another in a logical way (as when we conclude from all men are mortal and Socrates is a man that Socrates is mortal). This brings us to the third problem with computationalism, which is that the most a computational system can ever do is simulate conceptual thought, and never in principle actually carry it out.
The reason is that material symbols and processes cannot in principle have the universality of reference and determinacy of content that are characteristic of concepts. For example, the concept triangle of its nature applies to every single triangle without exception, whereas a material symbol either has no inherent connection to triangles at all (as in the case of the English word “triangle,” which applies to triangles merely by convention) or has an inherent connection but strictly applies only to some triangles but not all (as in a drawing of a triangle, which will always be either equilateral, isosceles, or scalene and thus strictly apply only to some of these but not all; will be of a specific color that not all triangles have; and so forth). There is also nothing in the material properties of any symbol or system of symbols that entails any determinate or exact representational content. For example, there is nothing in the symbol “triangle” that entails that it represents a specific triangle, or triangles in general, or the word “triangle” itself, or a person who calls himself “triangle,” or what have you. And merely adding further material symbols as interpretations of the first just kicks the problem up to a higher level, since these symbols themselves are, qua purely material, as indeterminate in their representational content as the one we started out with.
The basic point is as old as Plato and Aristotle and has been in recent years developed by James Ross, and I develop and defend Ross’s argument at length in my American Catholic Philosophical Quarterlyarticle “Kripke, Ross, and the Immaterial Aspects of Thought.” (I briefly sketched the argument in my review of Ray Kurzweil’s book How to Create a Mind in First Things, and had reason to discuss it in a recent series of blog posts, here, here, here, and here.)
Hence, from an Aristotelian point of view, even if qualia and some kinds of intentionality could be regarded as corporeal features of organisms on a “beefed up,” neo-Aristotelian conception of matter, conceptual thought could not be, and thus could not be captured even by a suitably updated computationalism. Conceptual thought -- which is distinctive of our rational or intellectual powers as Aristotelians understand them -- is essentially incorporeal.
When Gelernter rightly complains of the inability of computationalism to account for the subjectivity of conscious experience, then, I would argue that this inability is a symptom of the computationalist’s implicit commitment to a Galilean/Cartesian conception of matter, from which the qualitative has been extruded. It isn’t computationalism per se that is the problem, at least if the computationalist categories could be reinterpreted (as perhaps Braitenberg would interpret them) in an Aristotelian fashion. Subjectivity, in any case, isn’t for the Aristotelian the mark of the corporeal/incorporeal divide, since non-human animals (like Nagel’s famous bat) are entirely corporeal but nevertheless have subjective experiences of a sort.
Similarly, when Gelernter points out that “computers can be made to operate precisely as we choose; minds cannot,” I would argue that this is a consequence of the deeper point that the conceptual content of thought cannot be reduced to any set of relations between material symbols. There can in principle never be anything more than a very rough and general correlation between, on the one hand, the structure of corporeal states (whether in the brain, the organism as a whole, or the organism together with its environment), and, on the other hand, the conceptual content of our thoughts. Hence, even if we had total technological mastery of the relevant corporeal features of a human being, we would still never be able, even in principle, to predict and control the content of human thought with precision.
Some of Gelernter’s other points (such as that “computers can be erased; minds cannot”) are also important and deserve closer analysis than I have time to provide here. Still, they seem to me less fundamental than what I take to be the most basic problems with computationalism.
Gelernter also makes some suggestive remarks about emotions and experiences. He writes, for example, that “feelings are not information! Feelings are states of being… To experience is to be some way, not to do some thing.” This cries out for elaboration, which he does not have space to give in the article. What exactly does Gelernter have in mind here by the distinction between “being” and “doing”?
One way to read this might be in terms of what in recent analytic philosophy has been called the distinction between “categorical” and “dispositional” properties. A dispositional property would be one that a thing has when a certain conditional statement is true of it, namely a statement to the effect that if a certain kind of stimulus is present to the thing, then a certain kind of manifestation will follow. For example, brittleness is a dispositional property insofar as it involves the truth of a conditional to the effect that if a brittle thing is struck with a hard object, then it will shatter. Categorical properties, by contrast, would be those a thing simply has, unconditionally as it were. Shape is sometimes given as an example insofar as (it is sometimes held) a thing’s shape is something it simply has, unconditionally rather than merely as a manifestation in the presence of an appropriate stimulus.
Now whether the distinction holds up -- as opposed to purported categorical properties being ultimately reducible to dispositional ones, or purportedly dispositional ones being reducible to categorical ones -- is a matter of controversy in recent analytic metaphysics. (I discuss the controversy, and the relationship between the categorical/dispositional distinction and the Scholastic theory of act and potency, in Scholastic Metaphysics: A Contemporary Introduction.) But it is certainly relevant to the dispute in recent philosophy of mind over whether the qualia characteristic of emotional states and other conscious experiences can be explained in materialist terms. Functionalism -- of which computationalism is a variety -- essentially takes all mental phenomena to be describable in dispositional terms. The belief that it is raining or the experience of pain in one’s back, for example, would be characterized as internal states that will tend under appropriate conditions to manifest themselves by generating certain further internal states and/or certain kinds of overt behavior (grabbing an umbrella in the first case, say, or moaning in the second case). The critic of functionalism then objects that the feel or “quale” of the pain is something that the person or animal having it has, or at least might have, categorically, apart from any associated disposition. For the feel or quale of the pain could in principle exist (so the argument goes) even if it were associated instead with a disposition to laugh rather than moan, and the disposition to moan could exist even if it were associated with some quale other than the one we associate with pain, or indeed with no qualia at all (as in the case of a “zombie”).
Yet from an Aristotelian-Scholastic point of view such arguments are deficient, both because the categorical/dispositional distinction is too crude and fails to capture all the nuances enshrined in the theory of act and potency (as I discuss in Scholastic Metaphysics), and because “zombies” and related notions are dubious insofar as they reflect a Galilean/Cartesian conception of matter of the sort the Aristotelian would reject (a topic I discussed in a recent post).
But Gelernter might have something else in mind, and there may in any event be another way to elaborate his point.