What Marty Didn't Know

Great Debates
1

For those not immediately tipped of by the title, this is a takeoff on Frank Jackson’s famous thought experiment commonly known as ‘Mary’s Room’. It is one of the most well-known arguments thought to argue against physicalism, by postulating the existence of knowledge that is not contained within a purely physical description of the universe. It goes like this:

Mary has spent all her life imprisoned in a black-and-white room; she has never once seen colour (she’s probably wearing a black-and-white full body suit, as well, so she can’t see her own colour, and there are no mirrors, or just picture her permanently equipped with a pair of black-and-white VR goggles, it really doesn’t matter). But, she is a determined scientist of near-unlimited genius, and so, through careful research, she has availed herself of all physical facts about colour and colour perception in the human brain (not just all the knowledge we have now, but all the knowledge it’s possible to attain on the subject). Then, at some point, her captors show mercy, and release her from her room (or take off her goggles, whatever you like better). Outside, for the first time ever, she actually sees colour. The question is now, has she learned something new? For, if she has (or so the ostensibly unavoidable conclusion goes), there are facts about the world not contained within a purely physical description of it, and one of them is the subjective experience of colour.

Now, to most people, myself included, it is immediately obvious that Mary will learn something when first perceiving colour; how could she not? Even if she knew what colour, physically, is, and what the neural correlates of colour perception are, how could she have known what it was to her to experience those correlates? But, as I shall argue, I don’t think this disqualifies physicalism; I believe there is a hidden assumption in the argument that invalidates its conclusion, which I shall attempt to show with an analogous example.

Meet Marty. He has thrust upon him a fate similar to Mary’s – like her, he is permanently confined to his room, and like her, he’s a scientific genius determined to find out the answer to a certain question. However, his task, on first glance, seems much simpler than Mary’s – all he has to find out is which of his hands is his right (and therefore, which one is his left – more precisely, he has to find a difference between left and right, the naming is merely convention after all). To do this, he has access to a universal question answering machine – he can ask every question he wants, and the machine will answer truthfully, with one exception: everything within the room is ‘hidden’ from the answerer’s view, so he can’t just say things like ‘when you’re facing the answering machine, the wall the sofa’s standing against is to your right’ (also, by some strange quirk of nature, his own body is perfectly left-right symmetric, so that, for instance, his heart is perfectly in the centre etc.). Can he now in any way determine the answer to his problem?

I don’t believe he can. But that doesn’t imply that this answer isn’t to be found in a physical description of the universe – it’s a surprisingly hard question to answer, even then, though. As far as I know, the only manifest difference in the physical world between left and right is exhibited by parity violation in weak interactions; nevertheless, it’s a straight-forward experiment to find out the difference between the two.

Now, I believe most people would point out that Marty’s failure at distinguishing between left and right is solely a consequence of me artificially limiting his modes of inquiry – if he had a screen capable of displaying even rudimentary pictures, the task would have been trivial; even just something printing out text could simply produce the sentence ‘where you started to read this is left’. And that’s exactly right: all I’ve shown is that the information about the difference between left and right can’t be communicated in a certain way.

However, that is precisely where I believe the hidden assumption in the original thought experiment lies: when we imagine Mary researching her question, we picture her maybe reading books, or articles on the internet, looking at (black-and-white) pictures and whatnot, and we assume that through these channels, she is in fact able to gather all the relevant information. But, if some channels of information are unable to communicate some kinds of information which nevertheless is perfectly physical (as I think Marty’s example shows), then that assumption is completely unwarranted! It may well be the case that the information about the subjective experience of colour can’t be communicated via the channels we imagine Mary accessing, and yet, for it to be completely physical in nature. Hence, I believe the ostensible conclusion does not follow.

For another way to see this, imagine in Mary’s place a computer faced with the task of merely displaying colour on its screen. This computer possesses an extremely sophisticated text-parsing program, to the point as to enable a reasonable semblance of communication with it (however, it is explicitly not a conscious entity). You can ask it to display pictures, and it will comply – draw you a cat, a dog, a house, a car, even abstract shapes, limited only by your ability of describing them. However, ask it for a red rose, and it is stymied; while it ‘knows’ which commands to send to the screen to make it display arbitrary shapes, it has no knowledge of how to get it to display colour. Could you, with any amounts of explanation, now tell the computer how to display colour? Again, I don’t believe you can (after all, how would you explain ‘red’ to a blind person?), and again, I believe that this limitation is merely inherent in the channel of communication used to try to convey this information – natural language. If you, for instance, now outfit it with a colour camera, it will display the colours easily, and it would not be hard to create the architecture in such a way as to enable it to ‘learn’ from those pictures the commands needed to get the screen to display colours – if you for instance tell it that the roof of the house is red, it will then be able to draw you a red rose.

However, with such a computer, there is another possibility to get it to perform the task of displaying colour: you simply code the required commands into it, in such a way that the term ‘red’ correlates with the sequence of commands needed to display red pixels on the screen. (The computer’s self-programming capabilities are, by the way, limited – you couldn’t just tell it to program itself to be able to display colours, even if you could explain how the whole process works; it must ‘experience’ it – by displaying colour pictures – to know how to do it. This may seem like to arbitrary a limitation, but we generally don’t have easy access to our brain’s ‘programming’, either, yet we can influence it by generating the right experiences; so I think the model is reasonable in a first-approximation kind of way.) Thus, there exists a channel of communication apart from direct ‘experience’ able to convey the needed information, something which might conceivably also exist in Mary’s case, despite our (current) ignorance of such a possibility (however, even if there isn’t such a channel, I don’t believe this is of any significance to my conclusions).

I think everybody agrees that, despite the computer’s failure to learn how to display colour except after ‘experiencing’ it directly, there is no strangely non-physical quality to the ‘knowledge’ the computer needs to be able to do so; after all, you can also hack it in ‘by hand’. That it is hard – or even, as in Mary’s case, appears impossible – to communicate this knowledge has no bearing on the nature of the knowledge; it merely tells us something about the nature of certain channels of communication.

2

Agreed, when it comes to Mary. “Learning something” is a strange way to phrase it, but there is clearly a difference between understanding the anatomical response to a stimulus in general and the “understanding” generated in the response of the anatomy you yourself are made of. This need not require any epiphenomenon, disembodied qualia or any such nonsense. There is nothing non-physical about privacy: The anatomical response is effectively “encrypted” such that the anatomy actually responding does effectively “see a greater picture” than external third parties.

This is the part I do not accept about Mary. I propose that, in order to “know all there is to know” you have to not only have the “physical facts” (synapse-firings at a quantum level of detail, say – which is impossible itself but never mind). Those fact also have to be “fed in” at precisely the correct rate, in order. If you could engineer this, it would constitute a “sim program” which did indeed recreate the actual experience of Mary. But that is a long way off from her simply reading a few books on neuroanatomy.

But let’s get back to my “encryption” model (which I thought I’d come up with myself, only to see Dennett using something infuriatingly similar in Freedom Evolves, and here you go and independently come up with it yourself!) You can see that Marty falls into this model very nicely, since some information has been rendered private by the rules of your game, yet physicalism remains unimpugned.

Nice examples, sir. (From your user name, I suspect you might be impressed by the fact that I actually do own a Dukla Prague Away Kit.)

3

I don’t think this is correct. There is the issue of chirality. For example in chemistry there are molecules which have different left-handed and right-handed versions.

4

With Mary, there is only logical and intrapersonal intelligence. Just being able to experience color and the world for the first time opens new worlds of intelligence; naturalistic, visual-spatial, even existential intelligence, etc. Of course, this is just based on Howard Gardner’s theory of multiple intelligences, disregarding the fact that the first encounter with color and an outside world would generate such an emotional response that could never be measured by any IQ test.

As for Marty, if the machine is truly universal, it could answer whether the hand he’s raising is his right or left, but I’m assuming you mean the machine can’t see his actions.

5

In what way? It seems natural, to me: if Mary knows all the physical facts about colour vision (or even, all the physical facts about the whole world – if she has knowledge of the complete physical description of the world) and she learns something, gains new knowledge, this means that this physical description is not exhaustive.

Just as an aside, I maintain that the notion of epiphenomena is in and of itself physically inconsistent – to expect that an event can be caused with whatever it is that has caused it staying in exactly the same state it would be in if said event never occurred seems to me like expecting a billiard ball to continue along a straight line with unaltered speed after a contact with a second ball, with that second ball nevertheless gaining momentum from the collision; in other words I don’t think the physical world could remain unaffected by interacting with an epiphenomenal one, even though epiphenomena do not themselves have physical effects.

Yes, something like that is more or less what I had in mind (in analogy to the computer getting the information he needs to depict colour directly coded in) in speaking about a possible ‘channel’ that is able to convey the information about subjective colour perception that not even wikipedia can present you with.

Why, thanks! And Dennett is not such bad company to be in, and I can be fun too as long as the beer doesn’t run out.

Heh, I actually didn’t take my name from the band, which I don’t know much better than just by name, unfortunately. You know, it’s one of those times where you come up with an idea, only to find out that other people had it, too… :wink:

Well, I’m not sure if you can actually synthesize molecules of a given chirality on request without knowing the chirality of what goes into the mixture; and if you have some solution of just one given enantiomer, what then? You can shine light through it, and then what comes out the other side may be either left- or right-polarized, but that only helps you if you know what to expect. At any rate, I just needed for the difference between left and right to enter the physical description of the world at some point; what exactly that point is, is rather immaterial to my argument.

Hmm I don’t really think that the question of different intelligences really enters into the picture – after all, having other intelligences respond upon first seeing colour would seem to me to be equivalent to saying that Mary learns something new, and that’s where the problem lies: whether or not she does learn something new, and thereby whether or not all information about the world can be encapsulated in a physical description of it. (Or at least, that’s how the classical argument would have you believe the dichotomy goes – as I’ve elaborated, I think that she does learn something new, but only because she could not have amassed all the physical facts about the world by the means she is assumed to utilize.)

Yes, that’s exactly supposed to be the case – everything in the room is hidden from whatever intelligence gives the answers; everything else, it is allowed to access.

6

When reading the ‘Mary’ argument as you presented it the first thing that struck me is that there’s obviously a difference between the knowledge that red has a frequency 400–484 THz, which will stimulate your eye and send a signal to your brain, and the first-person experience of that actually happening. It’s the difference between knowing that water makes things wet and actually being wet - different sensory apparatues are stimulated, resulting in different signals being sent to the brain, which is different “knowledge”. Or put another way, it’s the difference between a perfect blueprint and a building - both describe the thing in question, but the arrangement and presentation of the data is itself data.

On reflection though, it occured to me that saying “she has availed herself of all physical facts about colour and colour perception in the human brain (not just all the knowledge we have now, but all the knowledge it’s possible to attain on the subject)” might imply that she has found out how to simulate the experience of seeing color, perhaps by manaully entering electrons into her optic nerve, and had done that to herself. I mean, she’d have to do this to have learned all physical facts about color.

And the thing is, if she had done this, then she really wouldn’t learn anything when let out - from her own perspective she’d have already experienced it! So, either the hypothetical is wrong about the premise (her having learned everything), or it’s wrong about the conclusion (the actual experience is new knowledge). One or the other, but either way it’s wrong.

7

I’m not actually sure this would convince somebody who holds the ‘Mary’s Room’ argument to be sound – after all, what you are proposing would appear to amount merely to exchanging the subjective experience of colour with the subjective experience of electrons trickling down the optical nerve, or, perhaps more generally, the neural correlates of colour perception. It doesn’t seem different in principle from her just biting down on her tongue and spitting the blood against the wall.

Of course, I’m convinced that this ‘subjective experience’ itself isn’t anything different in substance from the processes triggered by seeing colour/creating the illusion of seeing colour within oneself, so they themselves must have certain neural correlates, which, if those could somehow be artificially induced in Mary’s brain, would fully account for the fundamentally private experience of seeing colour. But this is a debate on its own, and it’s not easy to see how such a thing would be possible, if actually it ever becomes possible; hence, I stopped short of attempting to construct a channel with which to get the required information into Mary’s brain (and mind), and contented myself with attempting to show that such a channel can conceivably exist, and that it is a flawed assumption in the original argument that she doesn’t utilize one.

8

I can’t recommend it because it kinda sucked, but there’s a sci-fi novel with a vaguely relevant premise: The Planet of the Blind by Paul Corey.

9

I guess I’m trying to avoid saying “What she learns, in addition to all the physical facts, is what it is like to see.”

At which point a David-Chalmers type goes “Gotcha! There is something ineffable that transcends the physical facts.”

At which point I’d have to clarify, “No. I’m saying that there are some physical facts you cannot learn without actually being the seer, not just looking at seers in general.”

Those physical facts are, effectively, what those synapses are doing, precisely - not just what they are doing in general. But we don’t tend to say that when we experience things we “learn what are synapses are doing”. I’m just trying to bypass these linguistical roundabouts at the first junction.

10

Well, we’re definitely on the same page there; however, I want to keep my arguments open for the possibility that the cannot is a bit too strict, i.e. that there might be ways in which you can (or at least could, in a ‘possible in principle’ way) learn (or communicate) these physical facts, even though they are not presently accessible to us. For instance, I can use images to communicate ideas to you that are barely expressible in natural language (as, for instance, which way is left), and I can use mathematics in a similar way, I think. Facts in some language are inexpressible in another – using a language representing numbers in a decimal way, I can easily tell you the exact value of one divided by ten (0.1, there, see?), but in one using a binary representation, I would never finish reading off digits, though I would soon repeat myself (0.000110011…; though whether or not the repetition counts as me having told you, in a way, the exact value is probably again some matter of discussion, and I’d say again depends on the language you’re using – if you do some computation, and use 1/10 to whenever I got bored reciting digits, your result will be inexact, but there surely are computational representations that get around this problem just fine).

And of course, we don’t know – and can’t know – how general or fundamental the languages we so far possess actually are. An idea expressed for instance in this blog post is that it’s possible that “there exists an even more fundamental language that relates to today’s math in a way that today’s math relates to narrative”. So I guess that I think it’s at least possible that there exists some language we might one day invent/discover that is able to convey all the physical facts about the world, though I can’t conceive of what it might be like.

However, that there exists such a language is in no way necessary for my argumentation – Mary in all of her years of study could not have learned the exact value of π, in a manner of speaking; yet, most people would probably nevertheless accept that fact to be perfectly physical.

11

So, I guess this means there are no qualia enthusiasts around here? Well, I’ll give this a bump, just in case…

12

There was a panpsychist called II Gyan II and a pantheist called Aeschines back in the day (both involved in Camcorder Doom, Death Match!), and Kimmy Gibbler made some vaguely dualist noises in my Creationism thread. Not sure if there’s any genuine Dan Quales around anymore.