Back to CQ Homepage

Back to Archives' Table of Contents

Qualia and Computers

by Teed Rockwell


     In my last post, I described two sets of currently respectable concepts that seemed to presuppose the existence of sense data: qualia theory and computer theory. Since then, I've had some correspondence with David Chalmers between posts which seems to indicate that in his case, at least, the connection between sense data and qualia may take some work to unravel. (and remember folks, you read it here first!). Computer theory, on the other hand, now appears to me to be actually incompatible with sense datum theory. The criticisms of my original claim to the contrary have shown me that there is another concept which I was conflating with sense data: the psychologist's concept of sensation. This latter concept is far more resilient, but may be vulnerable to more subtle critiques found in the works of psychologists J.J. and E.J. Gibson. Those who are more interested in Psychology than Philosophy might want to skip ahead to the next section. I need a lot of input from psychologists on what Sue Pockett calls "this restricted trade sense" of sensation, and how it relates to the more cognitive concept of perception. But I also think that having Chalmers clarify his position on the relationship between sense data and qualia is helpful for those of us who are interested in understanding his ideas.


      Chalmers first response to the last CQ Post was the following:

      It's one thing to believe in qualia (phenomenal properties, characterizing what it's like to be in a given mental state). it's a quite different thing to believe in sense data (considered as the foundation of all perceptual knowledge, the things we observe directly in indirectly observing the world). There are fairly strong reasons based in epistemology and the philosophy of perception for rejecting the latter, but none of these reasons provide grounds for rejecting the former.

      I suppose that sense data, if they existed, would be qualia, but it's not the case that qualia must be sense data, at least in the objectionable sense. . . (e.g. re kane's point, while i believe in qualia, i certainly don't believe that perception is the "passive reception of data".) (***)

      I gathered from this that Chalmers might think that many of the characteristics shared by sense data and qualia ought to be removed from the latter concept. So I sent him this post (the list below I got from Owen Flanagan's "Consciousness Reconsidered"):

      qualia and sense data are both usually considered to be:

      1)intrinsic 2)atomic 3) Unanalyzable 4)non-relational 5)ineffable 6)essentially private 7)immediately corrigible and apprehensible.

      I think that most concepts of sense data would include all of these. Are there any you would not include in your concept of qualia? And if Chalmersian qualia lack any of these characteristics, what do they have instead, in each case?(***)

      To which Chalmers replied:

      i think qualia are certainly intrinsic and nonrelational. they are at least partly effable (e.g. i can describe their structure), though they may be partly ineffable. i don't know whether they are atomic or unanalyzable -- these notions would have to be spelled out better -- but i am tempted to say "not necessarily". i think they are essentially private in a sense, and i think they are immediately apprehensible.

      obviously sense data and qualia have many characteristics in common. given that a sense datum, if it exists, will be a quale. . .. but the reverse doesn't hold: sense data are generally taken to have further properties, e.g. they are the foundation of all perceptual knowledge and they are what we observe directly when we observe the external world indirectly, and these are the aspects of the doctrine of sense data that are most widely attacked. there's no reason why a supporter of qualia needs to say that qualia have these properties. (***)

      To which I replied:

      1)Owen Flanagan said that even though he agreed with Dennett that Qualia don't have these characteristics, that doesn't prove that there are no such things as Qualia. It seems, however, that you really do believe in just the sort of Qualia that Dennett wants to "Quine."

      2)Lets look at the properties that you say qualia need not have but which sense data do have. . . It seems to me that once we. . . shift the discussion to epistemology, the properties you attribute to qualia make them hard to distinguish from sense data. If they are immediately apprehensible, this would mean that our knowledge of them would be directly observed which would make them a foundation for our other knowledge, which is only indirectly observed. . . It also seems that once you have bought into the idea that qualia are intrinsic and non-relational, you also have to accept that they are immediately apprehensible. If their properties are completely intrinsic, they cannot be apprended by relating them to other things, which leaves unmediated apprehension as the only way of being aware of them.

      Maybe this is not the only explanation for why qualia have the characteristics you give them, but it seems the most obvious one, and it also seems pretty closely tied up with the concept of sense data. My only real point is that separating qualia from sense data is a job that still needs to be done by any sense-dataphobic qualiaphile, which is why sense datum theory is not as dead a horse as is generally believed. Dead horses, despite their numerous other failings, do not require us to clean up after them.(***)

      Chalmers replied with:

      Sure, I tend to believe in the qualia Dennett wants to quine. I don't necessarily buy all four properties straight up (it depends on how you analyze them), but i have some sympathy for privacy, immediate apprehensibility, and intrinsicness; ineffability is tricky. I don't think Dennett has good arguments against the first three (in the relevant sense).

      I don't think immediate apprehensibility implies epistemic foundationalism about external knowledge. The fact that we immediately apprehend qualia doesn't imply that our knowlege of the rest of the world is by virtue of our knowledge of qualia. for example, the qualia freak might simply take the nonfoundationalist's picture about external knowledge and say that picture works for the external, so qualia don't play an essential role there. and there are other subtler stories one can tell, involving qualia but not founding all knowledge of knowledge of qualia. (a qualia freak might even hold that knowledge of the external world comes before knowledge of qualia, even if the latter is more certain than the former.)(***)

      My last post to Chalmers was:

      Lets take a look at some of these logical possiblities, and see whether there are any you would want to actually accept. Do you really want to say that we have a direct awareness of the inhabitants of our inner space, but that we don't use this direct awareness in any way to help us learn about the outside world? This would strongly imply that qualia have no significant connection to the outside world that we could make use of. . . And yet surely this is false; If there such things as qualia, then most of the time when we sense red-apple shaped qualia, it is because there is an apple in front of us. So if qualia are directly grasped, they are clearly telling us a great deal about the world most of the time. How could we have another source of knowledge that was more reliable than this? It seems the only way we can have a nonfoundationalist view of knowledge is to assume that qualia are not directly grasped in this way. Otherwise our knowledge of qualia would beat out any other source of knowledge we had.

      I think most of the hard problems of consciousness that you raise would still be worthy of consideration if we cast them as bridging the gap between the third person and the first person perspective. I don't think we need to say that the first person perspective reveals a whole menagerie of entities that are distinct from those encountered from the third person perspective. Don't you find it ontologically offensive to posit the existance of both red apples and red apple shaped qualia, grey elephants and grey elephant shaped qualia, and so on ad infinitum?. . .I don't think that we need to posit a separate category for mental things that is more ontologically distinct than say the category of social things. We still have a big problem to deal with when we consider the fact that my brain/etc. states seem very different to me than they do to you. But we don't have to posit the existance of directly apprehended qualia to acknowledge that this is a problem.(***)

      I don't know whether Chalmers will acknowledge this, but it seems to me that he still has some work to do before he can claim that his qualia are completely distinct from sense data. However, I think that this separation can be accomplished, and that both Flanagan's book and the last paragraph of mine above give some indication of how it could be done. (Michael Tye's article in the On-line "Stanford Encyclopedia of Philosophy" also mentions the close relationship between sense data and qualia, and his suggestion for separating the two is quite similar to Flanagan's.)


      In the previous CQ post, I claimed that any attempt to apply computer theory to understanding human cognition would presuppose the existence of sense data. Jim Garson very effectively disposed of this claim with the following critique:

      . . .it is far from clear that the inputs {in classical computer theory} could even possibly be encountered AS sensations. For that, one would need to encounter inputs related to others in the right kinds of ways and this would involve finding the data within a cognitive level theoretical structure of some kind.... raw inputs by this very theory cannot count as the sensations, since their individuation AS sensations requires a framework of interrelations provided by the functionalist theory that undergirds the identification of mental states.

      I am reminded of a parallel moral to be drawn about computers. Imagine we have a simple caclulator set up to do addition and subtraction. The raw input might be arrays of 0s and 1s (or more carefully, switch settings). These have no status as NUMERICAL inputs until we encounter their computational roles defined by the machine. Computer states do not take on their intentional properties in a vacuum. This is the sense that inputs mathematically construed ( i.e. intentionally construed) are never "raw". This is an important Kantian theme in classical computationalism that seems to me to challenge features of "whole hog" sense data theory. Further evidence that the horse is long dead. (***)

      Andy Clark adds the following:

      At times, I felt you just might be failing to distinguish the idea of sense data from the idea of MERE INPUT. The point about sense data was that they were meant to be both RAW and yet somehow REALLY CONTENTFUL: a combination that looks ultimately incoherent. All that information theory really needs is the idea of an input to the system, and a set of actions to select from on the basis of that input. (***)

      These comments, in conjunction with some comments made by Sue Pockett and others, made me realise that I had been conflating SENSE DATA with SENSATIONS:two closely related (but crucially different) concepts, one from philosophy and the other from psychology.

      A sense datum is a single moment in sensory experience which is supposedly directly given to the perceiver, from which he indirectly infers the existence of objects. It is sense datum theory that prompted one analytic philosopher (I think it was C.D. Broad) to claim that we directly perceive an eliptical penny, and from this infer the existence of a round, three dimensional one. The best way to understand a sense datum is to conceive of it as a single frame of film in the moving picture which is our sensory experience. This is why I think that sense datum theory inspired Edison to invent the moving picture, which is a pretty impressive accomplishment for a false theory. (It would be interesting to research Edison's letters or diaries to see whether he consciously acknowledged this.)

      Ullin (U.T.) Place also contributed another effective criticism of the philosopher's sense datum:

      After filling a particularly deep cavity in one of my teeth recently, my dentist asked me to check any pain I might subsequently have to see whether it was caused equally by hot and cold stimuli (good) or only by hot (bad, particulary if throbbing). This, however, is a rather sophisticated form of observation which we learn only AFTER we have already learned to observe what is going on in the world around us. When I say I rejected the doctrine of sense-data more than fifty years ago, what I rejected was the idea that in observing what is going on around us, we begin by observing our sensory experience, formulate those observations in the form of a sentence in a private sense datum language and then use those private observation sentences as evidence for the existence and nature of what we NEVER observe, namely the objects and situations in the world around us. (***)

      (Note: Place also makes this same point in his classic paper "Is consciousness a brain process?", in which he calls this mistake "the phenomenological fallacy". In his CQ post, he gave a detailed analysis of why sense data have to be more like sentences, rather than pictures, in order to do the job that is required by the classical theory. He may be right about this, but I think that most believers in sense datum theory see them as more like pictures, even if they have no right to do so. Also, as Place points out in the post, the fact that they can't be sentences for other reasons is one of the weaknessess of sense datum theory. {note: Place's complete post is in the Commentators Respond--WTR.)

      This kind of sense datum theory is, I now think, a completely dead horse, and well it should be, for these and many other reasons. However, Psychologists have apparently dealt with these objections by making a distinction between PERCEPTIONS and SENSATIONS. I have yet to find a specific definition for these two terms and would be grateful for any references. From what I can tell from reading between the lines in various texts, however, perceptions are structured from sensations by various kinds of processing. In retrospect, I should have realized that there is a tremendous difference between a tabula rasa and an information processor, even though both of them need input to have content. The input to a tabula rasa is conceived of as a complete picture, but the input to an information processor is a bit or unit which is shaped into a picture (or a sentence etc.) by computational processes. The perceived picture is thus constructed out of bit-like sensations, but it is the sensations that give us contact with the outside world.

      Once I became aware of this distinction, I realized that computer science is in fact more Kantian than empiricist, despite my earlier claims that implied the contrary. But Kant was in his own way a kind of empiricist, for he was a firm disbeliever in what he called "intellectual intuition". He thought that we got all of our information about our world from the sense organs, except for those a priori forms that were hard-wired in. I'm tempted to say something here like "information without computation is blind, computation without information is empty.", to paraphrase Kant's famous statement about the relationship between concepts and intuitions.

      The following selections are from a post by Peter Lloyd, which I had described in the last CQ post as "a post from a Computer person, (lost during the great data disaster), which defended sense datum theory because it was essential to computer science." Now that Lloyd has re-sent that post, it seems to me that what he was really defending was sensations, not sense data:

      1. I can't see how a non-sensationalistic epistemology can ever get off the ground. The mind is (amongst other things) an information-processing system that uses information successfully in interacting with its environment. Therefore the mind possesses information. But it is a fundamental characteristic of information that it is *not* endlessly reducible. The analysis of information comes to an end with raw data. Therefore the mind possesses basic units of information, which we might call 'raw data' or 'sensations'. (***)

      Lloyd then adds some qualifiers that blur the analogy with computers.

      2. Talking to pragmatists in the past, I've got the impression that they think of sensations only as 'naive' sensations. For instance, if you think of your visual field as consisting of pixels of visual sensation like a TV screen, then that's what I'm calling a naive view of sensations. In fact, the sensations that make up the visual field must be more complex and subtle than that. Nevertheless, the visual field still consists of sensations. (I am thinking here of Hubel & Wiesel's experiments, and my own experiences of migraine 'castellations', which suggest larger visual elements as the raw units.) (***)

      (Anyone know anything about Hubel and Wiesel's experiments, and how they apply to this problem?) {Note:I received these posts from Bickle and Pockett in response to this question.--WTR}

      some questions: If the fundamental sensationalistic "atoms" are more complex than pixels, does this complexity come from the world or is it "cooked up" by our perceptual-cognitive processes? And if the latter, what right do we have to call them raw units? ( The pun has real significance, I think.) A little further down in his post, Lloyd says that when we are "hearing phonemes belonging to a language with which we are familiar, those phonemes *are* the raw conscious sensations." So apparently even a history of learning to recognize patterns does not stop those patterns from being "raw". The thing that makes them raw is apparently the fact that "What the brain then delivers to the conscious mind are indivisible, unanalysable units of raw sensation." What happens in the wings of the Cartesian theater is apparently irrelevant; if we experience a sensation as unanalyzable, it is unanalyzable. Lloyd also says that "sensations do not come *into* the mind from outside. . . They are constructed internally in response to incoming stimuli." This rules out Clark's view that sensations are themselves input, if they are only responses to input.

      So it looks from this description that sensations do not actually produce the results they are posited to explain: They do not present us with any aspect whatsoever of what is going on right now in unprocessed raw form, and they do not get us in contact with the outside world. What Lloyd is describing is very like Kant's view of a conscious being that creates a world in response to a thing-in-itself, without ever being able to actually experience the thing-in-itself. One significant difference between Lloyd's view and Kant's is that the former permits us to *know* things about the world even though we can't *experience* it. Our science, after all, enables us to make inferences about the outside world, even though, according to this view, we never encounter the world face to face. And it is also acknowledged that somehow the outside world affects our phenomenal experience, and positing the existance of sensations seems to be a way of trying to acknowledge this. But it is not all clear to me exactly what sensations are, and how they differ from the other processes that consitute perceptual experience. I seem to remember that Kant had a great deal to say about this subject, and if anyone can cite or paraphrase him on this issue, I would be grateful (Mark Steidle?) But because Kant said that the outside world was not only inexperienceable but unknowable, I don't think that most modern scientific realists would be willing to follow his arguments to their conclusions.

      None of this is a criticism of what Lloyd has written. In fact, I think he has done a beautiful job of capturing the tensions and inconsistencies in the perception-sensation distinction. Although Lloyd refers to himself humbly as " a professional software developer and an amateur philosopher" I think he has given this position about as good a defense as it can get (which is why I want to abandon that position and look for something better.) And I think I can see something on the horizon in the works of the pragmatists, in J.J. Gibson, and in the works of Andy Clark and Ruth Millikan.

      This post is already longer than it should be, so I will only mention that J.J. Gibson says on p.319 of "the Senses Considered as Perceptual Systems" that "Sensations are not, as we have always taken for granted, the basis of perception." His arguments for this position are scattered throughout his writings, and like most scientists he is more interested in facts than in teasing out their philosophical implications. The Millikan paper I mentioned at the end of the last CQ post, and Andy Clark's book "Being There", seem to build on Gibson's scientific work to some degree, so perhaps they will also end up agreeing with him on this point. In the meantime, if anyone has either defenses or criticisms of Gibson's position on sensations, I would love to hear them. From what I understand, Gibson is a controversial figure in Psychology, and I want to be able to hear both sides of the controversy. And when we start to talk about Gibson's concept of affordances, I think we will probably start to discover connections to Pragmatism and the Pragmatist concept of experience.

      I also have a couple of other very nice long posts. John Bickle sent me a detailed critique of Sergio Chaigneau's post, which I am forwarding to Sergio. John, Sergio, and I may have a side conversation that is more neurologically oriented, so anyone who wants to be in on that should let me know. Markate Daly also pointed out some significant differences between the pragmatist concept of experience and the knowing-how/knowing-that distinction, and proposed a very ingenious explanation for some of the problems of the sensory given. Markate's post will probably be important when we are considering Ruth Millikan's paper on knowing-how/knowing-that, as will the second half of Clark's post. In the meantime, keep those cards and letters coming in, Folks. As I said before, I read everything I get several times, and things that haven't been posted yet may be posted later.

      Thanks for joining me in the pursuit of truth, for want of a better word.

      Teed Rockwell