Essential Background: Dissolving the Question

How could we fully explain the difference between red and green to a colorblind person?

Well, we could of course draw the analogy between colors of the spectrum and tones of sound; have them learn which objects are typically green and which are typically red (or better yet, give them a video camera with a red filter to look through); explain many of the political, cultural and emotional associations of red and green, and so forth... but it seems that the actual difference between our experience of redness and our experience of greenness is something much harder to convey. If we focus in on that aspect of experience, we end up with the classic philosophical concept of qualia, and the famous thought experiment known as Mary’s Room1.

Mary is a brilliant neuroscientist who has been colorblind from birth (due to a retina problem; her visual cortex would work normally if it were given the color input). She’s an expert on the electromagnetic spectrum, optics, and the science of color vision. We can postulate, since this is a thought experiment, that she knows and fully understands every physical fact involved in color vision; she knows precisely what happens, on various levels, when the human eye sees red (and the optic nerve transmits particular types of signals, and the visual cortex processes these signals, etc).

One day, Mary gets an operation that fixes her retinas, so that she finally sees in color for the first time. And when she wakes up, she looks at an apple and exclaims, "Oh! So that's what red actually looks like."2

Now, this exclamation poses a challenge to any physical reductionist account of subjective experience. For if the qualia of seeing red could be reduced to a collection of basic facts about the physical world, then Mary would have learned those facts earlier and wouldn't learn anything extra now– but of course it seems that she really does learn something when she sees red for the first time. This is not merely the god-of-the-gaps argument that we haven't yet found a full reductionist explanation of subjective experience, but an intuitive proof that no such explanation would be complete.

The argument in academic philosophy over Mary's Room remains unsettled to this day (though it has an interesting history, including a change of mind on the part of its originator). If we ignore the topic of subjective experience, the arguments for reductionism appear to be quite overwhelming; so why does this objection, in a domain in which our ignorance is so vast3, seem so difficult for reductionists to convincingly reject?

Veterans of this blog will know where I'm going: a question like this needs to be dissolved, not merely answered.

That is, rather than just rehashing the philosophical arguments about whether and in what sense qualia exist4, as plenty of philosophers have done without reaching consensus, we might instead ask where our thoughts about qualia come from, and search for a simplified version of the cognitive algorithm behind (our expectation of) Mary's reaction. The great thing about this alternative query is that it's likely to actually have an answer, and that this answer can help us in our thinking about the original question.

Eliezer introduced this approach in his discussion of classical definitional disputes and later on in the sequence on free will, and (independently, it seems) Gary Drescher relied on it in his excellent book Good and Real to account for a number of apparent paradoxes, but it seems that academic philosophers haven't yet taken to the idea. Essentially, it brings to the philosophy of mind an approach that is standard in the mathematical sciences: if there's a phenomenon we don't understand, it usually helps to find a simpler model that exhibits the same phenomenon, and figure out how exactly it arises in that model.

Modeling Qualia

Our goal, then, is to build a model of a mind that would have an analogous reaction for a genuine reason5 when placed in a scenario like Mary's Room. We don't need this model to encapsulate the full structure of human subjective experience, just enough to see where the Mary's Room argument pulls a sleight of hand.

What kinds of features might our model require in order to qualify? Since the argument relies on the notions of learning and direct experience, we will certainly need to incorporate these. Another factor which is not immediately relevant, but which I argue is vital, is that our model must designate some smaller part of itself as the "conscious" mind, and have much of its activity take place outside of that part.

Now, why should the conscious/unconscious divide matter to the experience of qualia? Firstly, we note that our qualia feel ineffable to us: that is, it seems like we know their nature very well but could never adequately communicate or articulate it. If we're thinking like a cognitive scientist, we might hypothesize that an unconscious part of the mind knows something more fully while the conscious mind, better suited to using language, lacks access to the full knowledge6.

Secondly, there's an interesting pattern to our intuitions about qualia: we only get this feeling of ineffability about mental events that we're conscious of, but which are mostly processed subconsciously. For example, we don't experience the feeling of ineffability for something like counting, which happens consciously (above a threshold of five or six). If Mary had never counted more than 100 objects before, and today she counted 113 sheep in a field, we wouldn't expect her to exclaim "Oh, so that's what 113 looks like!"

In the other direction, there's a lot of unconscious processing that goes into the process of digestion, but unless we get sick, the intermediate steps don't generally rise to conscious awareness. If Mary had never had pineapple before, she might well extol the qualia of its taste, but not that of its properties as it navigates her small intestine. You could think of these as hidden qualia, perhaps, but it doesn't intuitively feel like there's something extra to be explained the way there is with redness.

Of course, there are plenty of other features we might nominate for inclusion in our model, but as it turns out, we can get a long way with just these two. In the next post, I'll introduce Martha, a simple model of a learning mind with a conscious/unconscious distinction, and in the third post I'll show how Martha reacts in the situation of Mary's Room, and how this reaction arises in a non-mysterious way. Even without claiming that Martha is a good analogue of the human mind, this will suffice to show why Mary's Room is not a logically valid argument against reductionism, since if it were then it would equally apply to Martha. And if we start to see a bit of ourselves in Martha after all, so much the better for our understanding of qualia...




One could reasonably ask what makes my attempt special on such a well-argued topic, given that I’m not credentialed as a philosopher. First, I'd reiterate that academic philosophers really haven’t started to use the concept of dissolving a question- I don’t think Daniel Dennett, for instance, ever explored this train of thought. And secondly, of those who do try and map cognitive algorithms within philosophy of mind, Eliezer hasn't tackled qualia in this way, while Gary Drescher gives them short shrift in Good and Real. (The latter essentially makes Dennett's argument that with enough self-knowledge qualia wouldn’t be ineffable. But in my mind this fails to really dissolve the question- see my footnote 4.)


1. The argument is called "Mary’s Room" because the original version (due to Frank Jackson) posited that Mary had perfectly normal vision but happened to be raised and educated in a perfectly grayscale environment, and one day stepped out into the colorful world like Dorothy in The Wizard of Oz. I prefer the more plausible and philosophically equivalent variant discussed above, although it drifts away from the etymology of the argument’s name.

2. Ironically, it was a green apple rather than a red one, but Mary soon realized and rectified her error. The point stands.

3. In general, an important rationalist heuristic is to not draw far-reaching conclusions from an intuitively plausible argument about a subject (like subjective experience) which you find extremely confusing.

4. Before we move on, though, one key reductionist reply to Mary’s Room is that either qualia have physical effects (like causing Mary to say "Oh!") or they don't. If they do, then either they reduce to ordinary physics or you could expect to find violations of physical law in the human brain, which few modern philosophers would dare to bet on. And if they don't have any physical effects, then somehow whatever causes her to say "Oh!" has nothing to do with her actual experience of redness, which is an exceptionally weird stance if you ponder it for a moment; read the zombie sequence if you're curious.

Furthermore, one could object (as Dennett does) that Mary’s Room, like Searle’s Chinese Room, is playing sleight of hand with impossible levels of knowledge for a human, and that an agent who could really handle such massive quantities of information really wouldn't learn anything new when finally having the experience. But to me this is an unsatisfying objection, because we don’t expect to see the effect of the experience diminish significantly as we increase her level of understanding within human bounds– and at most, this objection provides a plausible escape from the argument rather than a refutation.

5. (and not, for instance, because we programmed in that specific reaction on its own)

6. Indeed, the vast majority of visual processing- estimating distances, distinguishing objects, even identifying colors- is done subconsciously; that's why knowing that something is an optical illusion doesn't make you stop seeing the illusion. Steven Pinker's How the Mind Works contains a treasure trove of examples on this subject.

New Comment
155 comments, sorted by Click to highlight new comments since:
Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings
  1. A physically plausible scenario would involve growing up under a monochromatic light source.

  2. Growing up without sensory input actually affects the brain; see Wikipedia's article on monocular deprivation. I'm actually an example of this - I was born without the Mystic Eyes Of Depth Perception so I'll never know what stereoscopic vision "feels like".

  3. I propose that "qualia" is a word that, like "microevolution", is mainly used by people who are very confused (and dissolving the question is the appropriate approach).


I'm actually an example of this - I was born without the Mystic Eyes Of Depth Perception so I'll never know what stereoscopic vision "feels like".

If you turn something or move around it, even if you only use one eye to do this, your brain puts together the succeeding images to create a three-dimensional visual experience of the scene. Here is an example. If you're curious about "what it's like" to have stereo vision, in my opinion it is not far off from this, without the movement.

Here are two more.
The link is well worth following. Wow! Stereo vision with one eye closed!
Could you expand on this? I've seen before here the notion that the term "qualia" should be gotten rid of entirely, but I've never really understood it. For instance, asking what kinds of processes are capable of producing qualia, in order to figure out which animals are capable of feeling pain, certainly seems relevant for utilitarian ethics. (You could reword the question as "which animals can feel pain", which avoids using the term 'qualia', but you're at heart still referring to the same concept.)
It's also pretty difficult to describe phenomena like synaesthesa without terms like qualia.
Depth perception can be gained through vision therapy, even if you've never had it before. This is something I'm looking into doing, since I also grew up without depth perception.
I should have been more precise. I was born without a fully formed right eye - it has no lens and does not transmit a signal to my brain. Therefore, no "therapy" can improve my Vision Onefold. People in my situation (monocular blindness from birth) are extremely rare, so your assumption is understandable. I can get around in 3D space just fine, and I'm extremely good at first-person shooters, so I know I'm not missing much. (Coincidentally, I have no interest in physical sports.) The wiggle images do "work" for me. (On the other hand, "Possession of a single Eye is said to make the bearer equivalent to royalty.")

I'm not sure that this enterprise should be called "dissolving the question".

The question at hand is, "Is there something about red things that Mary can learn only by having certain kinds of input fed into her visual cortex?" This seems like a question that should be answered, not dissolved.

If you were trying to dissolve this question, you would probably proceed by trying to show that the concept of the things that Mary can learn about red things is itself meaningless, or at least that its meaning is too vague to give meaning to the qu... (read more)

The argument is called "Mary’s Room"...I prefer the more plausible and philosophically equivalent variant discussed above, although it drifts away from the etymology of the argument’s name

To be precise, the argument itself is called the Knowledge Argument (KA); "Mary's Room" is a name for the thought experiment Jackson used to present it.

Actually, it was one of two similar thought experiments in Jackson's original paper: the other one concerned a character named Fred who could see more colors than normal humans.

He really should have reversed the names. Colorblindness is more common in men than women, and it actually is possible for a few women to see an additional primary color. To summarize the linked article: Most people have three distinct kinds of photoreceptors that react to different frequencies of light: one receptor that is most sensitive to red, one receptor that is most sensitive to green, and one that is most sensitive to blue. (You can tell what color something is by the differences in the strength of the responses of the three different photoreceptors.) Genes for the red and green receptors are present on the X chromosome, so men have only one copy. Colorblind men have an abnormal version of one of these genes, so instead of getting a gene for seeing red and a gene for seeing green, they end up with a gene for seeing red and a gene for seeing a slightly different shade of red (which is why they can't tell the difference between red and green). On the other hand, if a woman has one copy of the "defective" gene and one copy of the "normal" gene, she could end up with four kinds of color receptors instead of the normal three: the one for red, the one for green, the one for blue, and the one for the slightly different red. This would let her see a difference between colors that look identical to people with normal color vision.

That someone knows every physical fact about gold doesn't make that person own any gold.

The Mary’s Room thought experiment explicitly claims that Mary knows every physical fact about the given phenomenon but does at the same time implicitly suggest that some information is missing.

Mary was merely able to to dissolve part of human nature by incorporating an algorithmic understanding of it. Mary wasn't able to evoke the dynamic state sequence from the human machine by computing the algorithm.

Understanding something means to assimilate a model of what is ... (read more)

In the same vein: Mary doesn't dissolve any part of human nature either. She dissolves an irregularity in her map.

FWIW, I'm satisfied with Dennett's explanation. If Mary knows everything physical about color, then there's nothing for her to be surprised about when she sees red. If your intuitions tell you otherwise, then your intuitions are wrong.

This begs the question, to be sure, but think of it more like moving to a more appropriate field of battle.


If Mary knows everything physical about color, then there's nothing for her to be surprised about when she sees red. If your intuitions tell you otherwise, then your intuitions are wrong.

Not really; it just means that our ability to imagine sensory experiences is underpowered. There are limits to what you can imagine and call up in conscious experience, even of things you have experienced. A person could imagine what it would be like to be betrayed by a friend, and yet still not be able to experience the same "qualia" as they would in the actual situation.

So, you can know precisely which neurons should fire to create a sensation of red (or anything else), and yet not be able to make them fire as a result.

Mere knowledge isn't sufficient to recreate any experience, but that's just a fact about the structure and limitations of human brains, not evidence of some special status for qualia. (It's certainly not an argument for non-materialism.)

That more or less corresponds to the way I break it down, and I'd take it a step further by saying that thinking of the problem this way reduces Mary's room to a definitional conflict. If we classify the experiential feeling of redness under "everything physical about color" -- which is quite viable given a reductionist interpretation of the problem -- then Mary by definition knows how it feels. This is probably impossible in practice if Mary has a normal human cognitive architecture, but that's okay, since we're working in the magical world of thought experiments where anything goes. If we don't, on the other hand, then Mary can quite easily lack experiential knowledge of redness without fear of contradiction, by the process you've outlined. It's only an apparent paradox because of an ambiguity in our formulation of experiential knowledge.
That's not how reduction works. You don't just declare a problem to consist only of (known) physics, and then declare it solved.You attempt to understand it in terms of known physics, and that attempt either succeeds or fails. Reductionism is not an apriori truth, or a method guaranteed to succeed. And no reduction of qualia has succeeded. Whether that me we need new explanations, new physics, non-reductionism or dualism is an open question.
I'm not sure you understand what I'm trying to say -- or, for that matter, what pjeby was trying to say. Notice how I never used the word "qualia"? That's because I'm trying to avoid becoming entangled in issues surrounding the reduction of qualia; instead, I'm examining the results of Mary's room given two mutually exclusive possible assumptions -- that such a reduction exists or that it doesn't -- and pointing out that the thought experiment generates results consistent with known physics in either case, provided we keep that assumption consistent within it. That doesn't reduce qualia as traditionally conceived to known physics, but it does demonstrate that Mary's room doesn't provide evidence either way.
Not being able to make the neurons fire doesn't mean you don't know how it would feel if they did. I hate this whole scenario for this kind of "This knowledge is a given but wait no it is not." kind of thinking. Whether or not all the physical knowledge is enough to know qualia is the question and as such it should not be answered in the conclusion of a hypothetical story, and then taken as evidence.
Huh? That sounds confused to me. As I said, I can "know" how it would feel to be betrayed by a friend, without actually experiencing it. And that difference between "knowing" and "experiencing" is what we're talking about here.
From what you quoted I thought you were arguing that there was something for her to be surprised about.
Of course there's something for her to be surprised about. The non-materialists are merely wrong to think this means there's something mysterious or non-physical about that something.
It may be more accurate to say that when she sees a red object, that generates a feeling of surprise, because her visual cortex is doing something it has never done before. Not that there was ever any information missing -- but the surprise still happens as a fact about the brain.
We measure information in terms of surprise, so you're kind of contradicting yourself there. The entire "thought experiment" hinges on getting you to accept a false premise: that "knowledge" is of a single kind. It then encourages you to follow this premise through to the seeming contradiction that Mary shouldn't be able to be surprised. It ignores the critical role of knowledge representation, and is thus a paradox of the form, "If the barber shaves everyone who doesn't shave themselves, does the barber shave him/herself?" The paradox comes from mixing two levels of knowledge, and pretending they're the same, in precisely the same way that Mary's Room does.
I mean surprise in the sense of the feeling, which doesn't have to be justified to be felt. Perhaps a better word is "enlightenment". Seeing red feels like enlightenment because the brain is put into a state it has never been in before, as a result of which Mary gains the ability (through memory) to put her brain into that state at will.
That isn't a paradox. It is a simple logical question with the answer yes.
Hm, I guess that should probably be, "if the barber shaves only those who don't shave themselves."
"if and only if"-type language has to enter into. If the barber shaves all and only those who don't save themselves...
Cracked me up. I think you might mean "shave" here.
Oh no! The barber of Seville is coming! I'll hold him off, you save yourself!
But what if I run into the barber of Fleet Street?!
What is it that she's surprised about?
The difference between knowing what seeing red is supposed to feel like, and what it actually feels like.
I think the idea that "what it actually feels like" is knowledge beyond "every physical fact on various levels" is just asserting the conclusion. I actually think it is the posited level of knowledge that is screwing with our intuitions and/or communication here. We've never traced our own algorithms, so the idea that someone could fully expect novel qualia is alien. I suspect we're also not smart enough to actually have that level of knowledge of color vision, but that is what the thought experiment gives us. I think the chinese room has a similar problem: a human is not a reliable substrate for computation. We instinctively know that a human can choose to ignore the scribbles on paper, so the chinese speaking entity never happens.
Ah, but what conclusion? I'm saying, it doesn't matter whether you assume they're the same or different. Either way, the whole "experiment" is another stupid definitional argument. However, materialism does not require us to believe that looking at a menu can make you feel full. So, there's no reason not to accept the experiment's premise that Mary experiences something new by seeing red. That's not where the error comes from. The error is in assuming that a brain ought to be able to translate knowledge of one kind into another, independent of its physical form. If you buy that implicit premise, then you seem to run into a contradiction. However, since materialism doesn't require this premise, there's no reason to assume it. I don't, so I see no contradiction in the experiment. If you think that you can be "smart enough" then you are positing a different brain architecture than the ones human beings have. But let's assume that Mary isn't human. She's a transhuman, or posthuman, or some sort of alien being. In order for her to know what red actually feels like, she'd need to be able to create the experience -- i.e., have a neural architecture that lets her go, "ah, so it's that neuron that does 'red'... let me go ahead and trigger that." At this point, we've reduced the "experiment" to an absurdity, because now Mary has experienced "red". Neither with a plain human architecture, nor with a super-advanced alien one, do we get a place where there is some mysterious non-material thing left over. Not exactly. It's an intuition pump, drawing on your intuitive sense that the only thing in the room that could "understand" Chinese is the human... and he clearly doesn't, so there must not be any understanding going on. If you replace the room with a computer, then the same intuition pump needn't apply. For that matter, suppose you replace the chinese room with a brain filled with individual computing units... then the same "experiment" "proves" that brains can't po
Looking at a menu is a rather pale imitation of the level of knowledge given Mary. That is the conclusion you're asserting. I contend that she can know, that there is nothing left for her to be surprised about when that neuron does fire. She does not say "oh wow", she says "ha, nailed it" If she has enough memory to store a physical simulation of the relevant parts of her brain, and can trigger that simulation's red neurons, and can understand the chains of causality, then she already knows what red will look like when she does see it. Now you might say that in that case Mary has already experienced red, just using a different part of her brain, but I think it's an automatic consequence of knowing all the physical facts.
No matter how much information is on the menu, it's not going to make you feel full. You could watch videos of the food being prepared for days, get a complete molecular map of what will happen in your taste buds and digestive system, and still die of hunger before you actually know what the food tastes like. In which case, we're using different definitions of what it means to know what something is like. In mine, knowing what something is "like" is not the same as actually experiencing it -- which means there is room to be surprised, no matter how much specificity there is. This difference exists because in the human neural architecture, there is necessarily a difference (however slight) between remembering or imagining an experience and actually experiencing it. Otherwise, we could become frightened upon merely imagining that a bear was in the room with us. (IOW, at least some portion of our architecture has to be able to represent "this experience is imaginary".) However, none of this matters in the slightest with regard to dissolving Mary's Room. I'm simply pointing out that it isn't necessary to assume perfect knowledge in order to dissolve the paradox. It's just as easily dissolved by assuming imperfect knowledge. And all the evidence we have suggests that the knowledge is -- and possibly must -- be imperfect. But materialism doesn't require that this knowledge be perfectable, since to a true materialist, knowledge itself is not separable from a representation, and that representation is allowed (and likely) to be imperfect in any evolved biological brain.
Metaphysics is a restaurant where they give you a thirty thousand page menu, and no food. - Robert M. Pirsig
"Feeling full" and "seeing red" also jumbles up the question. It is not "would she see red" But isn't your "knowing what something is like" based on your experience of NOT having a complete map of your sensory system? My whole point this that the given level of knowledge actually would lead to knowledge of and expectation of qualia. Nor is the question "can she imagine red". The question is: Does she get new information upon seeing red? (something to surprise her.) To phrase it slightly differently: if you showed her a green apple, would she be fooled? This is a matter-of-fact question about a hypothetical agent looking at its own algorithms.
If there's a difference in the experience, then there's information about the difference, and surprise is thus possible. How, exactly? How will this knowledge be represented? If "red" is truly a material subject -- something that exists only in the form of a certain set of neurons firing (or analagous physical processes) -- then any knowledge "about" this is necessarily separate from the thing itself. The word "red" is not equal to red, no matter how precisely you define that word. (Note: my assumption here is that red is a property of brains, not reality. Human color perception is peculiar to humans, in that it allows us to see "colors" that don't correspond to specific light frequencies. There are other complications to color vision as well.) Any knowledge of red that doesn't include the experience of redness itself is missing information, in the sense that the mental state of the experiencer is different. That's because in any hypothetical state where I'm thinking "that's what red is", my mental state is not "red", but "that's what red is". Thus, there's a difference in my state, and thus, something to be surprised about. Trying to say, "yeah, but you can take that into account" is just writing more statements about red on a piece of paper, or adding more dishes to the menu, because the mental state you're in still contains the label, "this is what I think it would be like", and lacks the portion of that state containing the actual experience of red.
The information about the difference is included in Mary's education. That is what was given. Are you surprised all the time? If the change in Mary's mental state is what Mary expected it to be, then there is no surprise. How do you know? Isn't a mind that knows every fact about a process itself an analogous physical process?
This is how this question comes to resemble POAT. Some people read it as a logic puzzle, and say that Mary's knowing what it's like to see red was given in the premise. Others read it as an engineering problem, and think about how human brains actually work.
That treatment of the POAT is flawed. The question that matter is whether there is relative motion between the air and the plane. A horizontally tethered plane in a wind tunnel would rise. The treadmill is just a fancy tether.
What? That's the best treatment of the question I've seen yet, and seems to account for every possible angle. This makes no sense: The plane in the thought experiment is not in a wind tunnel. Treated realistically, the treadmill should not have any tethering ability, fancy or otherwise. Which interpretation of the problem were you going with?
A plane can move air over its own airfoils. Or why not make it a truck on a treadmill?
By the way, you may not agree with my analysis of qualia (and if so, tell me), but I hope that the way this thread derailed is at least some indication of why I think the question needed dissolving after all. As with several other topics, the answer may be obvious to many, but people tend to disagree about which is the obvious answer (or worse, have a difficult time even figuring out whether their answer agrees or disagrees with someone else's).
I definitely welcome the series, though I have not finished it yet, and will need more time to digest it in any case.
It's at least evidence about the way our minds model other minds, and as such it might be helpful to understand where that intuition comes from.
OK. Do you know that? Does Mary?
Well, through seeing red, yes ;-) Through study, no. I think the knowledge postulated is beyond what we currently have, and must include how the algorithm feels from the inside. (edit: Mary does know through study.)
That does sound fallacious. Fortunately you don't need additional evidence. An even better proposal: You should put the answer in the prologue and then not bother writing a story at all. Because we moved on from that kind of superstition years ago.
Maybe, but Mary nonetheless by hypothesis knows exactly what it would feel like if those neurons fire, since that's a physical fact about color. Like I said, that's begging the question in the direction of materialism, but assuming that fact is non-physical is begging the question in the direction of non-materialism.

Like I said, that's begging the question in the direction of materialism

Not at all. The question is only confused because the paradox confuses "knowing what would happen if neurons fire" and "having those neurons actually fire" as being the same sort of knowledge. In the human cognitive architecture, they aren't the same thing, but that doesn't mean there's any mysterious non-physical "qualia" involved. It's just that we have different neuronal firings for knowing and experiencing.

If you taboo enough words and expand enough definitions, the qualia question is reduced to "if Mary has mental-state-representing-knowledge-of-red, but does not have mental-state-representing-experience-of-red, then what new thing does she learn upon experiencing red?"

And of course the bloody obvious answer is, the mental state representing the experience of red. The question is idiotic because it basically assumes two fundamentally different things are the same, and then tries to turn the difference between them into something mysterious. It makes no more sense than saying, "if cubes are square, then why is a sphere round? some extra mysterious thing is happening!"

So, it's not begging the question for materialism, because it doesn't matter how complete Mary's state of knowledge about neurons is. The question itself is a simple confusion of definitions, like the classic tree-forest-sound question.

I think we've at least touched upon why this question needs to be dissolved. Reading the thought experiment as a logic problem, one should accept the conflation of the two putative mental states you've identified (calling them both 'knowing') and note that by hypothesis Mary 'knows' everything physical about color. Thus, the question is resolved entirely by determining whether the quale is non-physical. And so if you accept the premises of the thought experiment, it is not good for resolving disputes over materialism. Dennet, being a materialist, reads the question in this manner and simply agrees that Mary will not be surprised, since materialism is true. Personally, I'm pretty okay with mental-state-representing-experience-of-red being part of "knowledge". Even if humans don't work that way, that's kindof irrelevant to the discussion (though it might explain why we have confused intuitions about this).
Then he is quite simply wrong. Knowledge can never be fully separated from its representation, just as one can never quite untangle a mind from the body it wears. ;-) This conclusion is a requirement of actual materialism, since if you're truly materialist, you know that knowledge can't exist apart from a representation. Our applying the same label to two different representations is our own confusion, not one that exists in reality. If you start from a nonsensical premise, you can prove just about anything. In this case, the premise is begging a question: you can only conflate the relevant types of knowledge under discussion, if you already assume that knowledge is independent of physical form... an assumption that any sufficiently advanced materialism should hold false.
It really doesn't have to be a confusion though. We apply the label 'fruit' to both apples and oranges - that doesn't mean we're confused just because apples are different from oranges. I don't think either I or Dennett made that claim. You don't need it for the premise of the thought experiment. You just need to understand that any mental state is going to be represented using some configuration of brain-stuff... According to the thought experiment, Mary "knows" everything physical about the color red, and that will include any relevant sense of the word "knows". And so if the only way to "know" what experiencing the color red feels like is to have the neurons fire that actually fire when seeing red, then she's had those neurons fire. It could be by surgery, or hallucination, or divine intervention - it doesn't matter, it was given as a premise in the thought experiment that she knows what that's like. One way to make such a Mary would be to determine what the configuration of neurons in Mary's brain would be after experiencing red, then surgically alter her brain to have that configuration. The premise of the thought experiment is that she has this information, and so if that's the only way she could have gotten it, then that's what happened.
This is going way beyond what I'd consider to be a reasonable reading of the intent of the thought experiment. If you're allowed to expand the meaning of the non-specific phrase "knows everything physical" to include an exact analogue of subjective experience, then the original meaning of the thought experiment goes right out the window. My reading of this entire exchange has thomblake and JamesAndrix repeatedly begging the question in every comment, taking great license with the intent of the thought experiment, while pjeby keeps trying to ground the discussion in reality by pinning down what brain states are being compared. So the exchange as a whole is mildly illuminating, but only because the former are acting as foils for the latter. You can't keep arguing this on the verbal/definitional level. The meat is in the bit about brain states. Call the set of brain states that enable Mary to recall the subjective experience of red, Set R. If seeing red for the first time imparts an ability to recall redness that was not there before, then as far as I'm concerned that's what's meant by "surprise". We know that seeing something red with her eyes puts her brain into a state that is in Set R. The question is whether there is a body of knowledge, this irritatingly ill-defined concept of "all 'physical' knowledge about red", that places her brain into a state in Set R. It is a useless mental exercise to divorce this from how human brains and eyes actually work. Either a brain can be put into Set R without experiencing red, or it can't. It seems very unlikely that descriptive knowledge could accomplish this. If you're just going to toss direct neuronal manipulation in there with descriptive knowledge, then the whole thought experiment becomes a farce.
Then she knows things humans in their current form can't learn except by seeing red. Either she found a way to reprogram herself, or she has seen red, or the problerm is ill-posed because it equivocates between what humans can learn at all and what they can learn from reading words in textbooks.
Why does Mary need to imagine red in order to know what it looks like? If the physical understanding she already has accounts for it, then she should be able to figure it out from that, as per the Dennett response. Like several people in this thread, you are tacitly assuming that there is something special about qualia, such that they need to be imagined or instantiated in order to be known -- something that is unique about them, even though they are ultimately physical like everything else.
Oops, I realized during editing that surprise is not so much what we should look for as learning, but I forgot to remove the instances of "surprise" from this post. Done. In a day or two, I can better explain why I reject Dennett's explanation, but for now it's enough to note that it doesn't dissolve the question at all.
This does not match my experience doing things after studying them thoroughly. Unless your definition of "everything physical about color" includes neurology far beyond the state of the art.
Indeed it does. I believe a strong reading also involves knowing the position of every colored object in the universe, and the favorite food of every person with red hair.
Wouldn't this include complete physical knowledge of the universe? It is interesting to me that this is a contradiction in a finite universe. It intuitively feels like one might be able to analyze this self-reference and its source and derive a convincing argument against Mary's room, but I cannot find one right now.
Consider one's brain state on the granularity that it stores information. It contains N bits of information. What would those N bits of information be if you saw something red? It is impossible to know this information, because it would take all your available N bits. Yet if you don't know all this information, clearly you learn something new as soon as you see something red, at least, as long as your attention would then be drawn to the parts you didn't previously know about, which doesn't seem an unreasonable assumption at all. Now suppose that you have that knowledge in some compressed form, of M<N bits. But then seeing red and entering that state would be like completing a calculation, which frequently produces the "Aha! I have learned something new!" response.
Since a human mind can hold only an infinitesimal fraction of that information, Mary is now a mind quite unlike our own, and likely to have very different qualities.
The Dennetian Answer isnt based on anything that actually happens.It's based on having a really, really strong intuition of physicalism.
Exactly. It's not a refutation so much as a bullet-biting, which is unenlightening even when correct.
Since we don't actually have physical explanations of qualia, that is itself an intuition. It's one intuition against another. not one intuition against some fact that disproves it.
No, there are (or, at least can be in principle) various good reasons for thinking physicalism is true - it need not rest on mere intuition. And once you've assumed that physicalism is true, the above is a consistent, correct conclusion for the thought experiment. If you think physicalism is false, I would not think the above explanation would feel very satisfying to you - but then, I was talking about what feels satisfying to me.
We still don't have a physical explanation of qualia. So it is still a kind of guess that physicalism, which has been successful in other areas cab be extended to all areas. It's an intuition based on evidence, but the intended intuition of the Mary story is based on evidence as well, since we have all had novel experiences which went beyond their descriptions. Can't I just find it unsatisfying anyway?
wouldn't everything physical include, say, your complete brain state now and what your complete brain state would be if you saw something red? and wouldn't it be impossible for you to hold that info in your brain?
Not if you could losslessly encode it enough and also work directly with such encodings.
1. It is difficult to losslessly compress something that has already been losslessly compressed, because its entropy/bit will be very high. 2. De-compressing something you already had losslessly encoded feels like learning or discovery, depending on whether it is externally induced. Like, if you know the statement of a difficult math problem, then you know enough information to pin down the answer, but you do not know the answer. If I tell you the answer, you will feel like you just learned something. And you did, in some senses of the word, learn something.
Google keywords: logical uncertainty

I was sort of toying with an idea a while ago; it is somewhat old though I've tweaked it a bit for presentation here. I don't like it so much anymore, but I still think it has some potential so I'll go ahead and share it anyway:

Suppose we alter the Mary's room scenario so that Mary isn't human, but rather comes from a race of philosophical 'Empaths' that have the ability to perfectly convey subjective experience due to some ability to transmit and read off another each other's neural patterns as well as hack their own CNS.

In this altered scenario, Mary can... (read more)

This is somewhat circular. There isn't anyone who knows everything about the visual system. Thus, we're hypothesizing that knowing everything about the visual system is insufficient to understand what red looks like... prove that knowing everything about the visual system is insufficient to understand what red looks like.

Even given this, the obvious solution seems to be that "What red looks like" is a fact about Mary's brain. She needn't have seen red light to see red; properly stimulating some neurons would result in the same effect. That the e... (read more)

A better analogy would be: you have a DVD and a complete set of schematics for a DVD player, and the ability to understand both, but still can't figure out what the DVD would look like when viewed.
I think your analogy betrays you: an AI wouldn't need to have an actual DVD player to turn the ones and zeroes into an experience of the film, it would just need to know the right algorithm. Let's be clear here- you're advocating an epistemically non-reductionist position, which should seem at least a little weird: if brains are made of atoms, why should the hanging questions of what an experience feels like be unanswerable from knowledge of the brain structure?
Let's be clear here - I'm advocating no such thing. My position is firmly reductionist. Also, we're talking about Mary, not an AI. That counterexample is completely immaterial and is basically shifting the goalposts, at least as I understand it. Any experience is, basically, a firing of neurons. It's not something that "emerges" from the firing of neurons; it is the firing of neurons, followed by the firing of other neurons that record the experience in one's memory. What it feels like to be a bat is a fact about a bat brain. You neither have a bat brain nor have the capacity to simulate one; therefore, you cannot know what it feels like to be a bat. Mary has never had her red-seeing neurons fired; therefore, she does not know what red looks like. If Mary were an advanced AI, she could reason as follows: "I understand the physics of red light. And I fully understand my visual apparatus. And I know that red would stimulate my visual censors by activiating neurons 2.839,834,843 and 1,2345. But I'm an AI, so I can just fire those neurons on my own. Aha! That's what red looks like!" Mary obviously has no such capacity. Even if she knows everything about the visual system and the physics of red light, even if she knows precisely which neurons control seeing red, she cannot fire them manually. Neither can she modify her memory neurons to reflect an experience she has not had. Knowing what red looks like is a fact about Mary's brain, and she cannot make her brain work that way without actually seeing red or having an electrode stimulate specific neurons. She's only human. Of course, she could rig some apparatus to her brain that would fire them for her. If we give her that option, it follows that knowing enough about red would in fact allow her to understand what red looks like without ever seeing it.
Doesn't it follow that Mary, since she knows everything about color, must have both electrodes and the desire and ability to perform of a brain surgery on herself? There is a truly fabulous story, rkunyngvba ol grq puvnat in which the protagonist does this, but since it only happens half way through, I don't want to spoil it.
Once gain, Mary knows everything knowable by description only. Whether that amounts to everything simpliciter is the puzzle.
But you still haven't explained why she would need to fire her own neurons. She doens;t need to photosynthesise to understand photosynthesis.

This thought experiment always seemed silly to me. As if somehow the experience of the visual cortex reacting to "color" input was not a piece of knowledge.

If someone has a poor ability to mentally visualize 3-dimensional objects, and is shown a set of formula that will draw a specific and very odd object (learning everything but what the object actually looks like), and is only ever allowed to graph on paper, then of course when we finally hand them a physical model of the object we have given them new information.

I don't see this as any differe... (read more)

I think you're mistaking the conclusion that the non-reductionist philosophers draw from the thought experiment. They're not generally substance dualists like Descartes. Instead, they claim that reductionism is false because it is epistemically incomplete, even within a purely physical world: a human (or an AI, or anything other than a bat) cannot ever understand the experience of being a bat, and therefore not all knowledge reduces to mathematical patterns of physical objects.
The knowledge Mary has is all physical knowledge, where physical knowledge means the kind of thing that can be found in books. You deam the further, experiential knowledge she gains to be physical because sensory processing is physical, but that is a different sense of physical. If you think she learns something on exiting the room, and it seems you do, then you are conceding part of the claim, the part about the incompleteness of physical explanation, even if you insist that the epistemic problem doesn't lead to an dualistic metaphysics.
Only insofar as the definition of physical is limited to things you can find in books. I wholly reject such a definition. @ Orthonormal. The conclusion seems to me to come very naturally from the thought experiment, if you allow for its assumptions. But that is what I think is silly, its assumptions. The thought experiment tries to define "all knowledge" in two different and contradictory ways. If Mary has all knowledge, then there is nothing left for her to learn about red. If upon seeing red she learns something new, then she did not have all knowledge prior to seeing red. It is their definition of knowledge, which is inconsistent, that leads to the entire thought experiment being silly.
If you want to call someone's attention, replying directly to their comment is a great way of doing that.
The practical point is that, if not all knowledge reduces to mathematical patterns of physical objects (the sort of thing that we can organize and learn from textbooks), then the actual project of reductionists becomes futile at a really early stage- we'd have to give up on fully understanding even a worm brain, since we could never have the knowledge of its worm-qualia. I want to respond to your claim more thoroughly, but my response essentially consists of the second and third posts here. If you want to pick up this conversation on those threads, I'm all for it. Also, welcome to Less Wrong!
Your later posts do a better job of describing your position here. I don't think we disagree.

Let's take an analogy:

You're writing a video game that will run on hardware that doesn't do floating-point arithmetic (like the Nintendo DS), so you have to write a library to emulate floating-point arithmetic. You then port the game to a very similar system whose hardware does handle floats, so you replace your library with simple operations.

Mary's situation is similar: Mary is perfectly capable of anticipating any experience related to color, and "seeing red" doesn't change that, but allows her to use the functionality built in her hardware... (read more)

"Anticipating colour" is vague. She may be able to anticipate "I will see red" without anticipating what it looks like. Since she knows everything about neurology, why can't she figure out how the refactoring will change her phenomenology? Or can she?

Looks really interesting.

I especially like the notion that the interplay between the conscious and unconscious mind is important and the observation that some sensory inputs to our body feels ineffable and some do not.


From just reading the definition of the Mary's Room problem my knee-jerk reaction was "this seems plausible." It is a textbook example of how an algorithm feels from the inside.

You might know everything there is to know intellectually about colours, but that does not induce sensations in your visual cortex. Humans don't work that way.

A bayesian AI might go "that is about what I expected" when you switch it from a black and white camera to a colour one, solely on the basis of the production parameters of the camera, a physics paper on optics and perhaps a single colour photo.

(nods) This is likely true, but a lot of work is being done by the word "intellectually". If the conclusion of the thought experiment is that there is information to be obtained by experience that is not captured by whatever we're calling "intellectual" knowledge, or more generally that there's information in-principle-extractable from an event that isn't ordinarily extracted by particular cognitive systems, that's not really all that remarkable. Using this thought experiment the way it is traditionally used requires a bit of sleight of hand, wherein we are encouraged to apply our intuitions about the kinds of knowledge our brains extract to all knowledge.
it is not remarkable from many perspectives, but it is contradictory to some forms of physicalism, which argue that everything is understandable by the methods of physical science, ie from the outside.
Can you clarify the contradiction? It seems to me there could easily be information in-principle-extractable from an event, that isn't ordinarily extracted by particular cognitive systems, that is understandable by the methods of physical science. I mean, I certainly agree that if we posit the existence of information that is not in-principle understandable by the methods of physical science but is ordinarily extracted by particular cognitive systems, then that does contradict the forms of physicalism you're talking about here. My point, though, is that the scenario described in Mary's Room does not give us any reason to believe that such information exists.
Several people on this site have responded to M's R with the claim that (in effect) such information does exist. The claim is usually expressed on the lines of an individual needing to be in a brain state, to personally instantiate it, in order to understand it. For instance: "You might know everything there is to know intellectually about colours, but that does not induce sensations in your visual cortex".
So, to repeat myself, I agree that if such information exists, then the forms of physicalism you're talking about here are false. And I agree that several people have responded to Mary's Room with the claim that such information does exist, and I don't deny that they've done so, nor have I even denied that they've done so. What I do deny is that Mary's Room demonstrates that such information exists, or that they are justified in believing anything different after being exposed to MR than before. MR just invites me to generalize from my limited experience of knowing some things about color to a hypothesized state of intellectually knowing everything there is to know about color, and anticipates that I will ignorantly imagine keeping other aspects of my limited experience fixed. If I instead ignorantly imagine other aspects of my limited experience varying, it completely fails to demonstrate what it's claimed to demonstrate. For example, if I start out believing that sensations of color are in-principle unavailable to a healthy human brain solely by virtue of being in that hypothesized state, then MR might feel like a compelling demonstration of that claim. "Oh look, there's Mary," I might say, "and I know she's never had such sensations, so clearly seeing a yellow banana is new information to her, therefore...etc. etc. etc.". Conversely, if I don't believe that to start with, it might not. "Oh look," I might say, "there's Mary, who knows everything there is to know about color, and has probably therefore had vivid dreams of seeing color as her brain has made various connections with that information, so clearly seeing a yellow banana is not new information to her, therefore etc. etc." There might be good reasons to reject that second intuition and embrace the first, or vice-versa, but thinking about Mary's Room is not a good reason to do either. It's just a question-begging invitation to visualize my preconceptions and treat them as confirming data. All of tha

I like it so far! One question, though:

Our goal, then, is to build a model of a mind that would express analogous reactions in Mary's Room for a genuine reason

Are you referring the question of how to model a mind that would have analogous reactions to:

  • being told about the Mary's Room argument, or to
  • being in the position of a real-world Mary?

I assume the first, but I couldn't tell. In any case, both are worthwhile targets for reduction.

Edit: I had previous sketched out a reduction of qualia, but it seems more focused on the first issue -- why we ... (read more)

I actually mean the second question; let me see if I can make that less ambiguous in the post. The first is important too, and will come up by the end. And thanks! UPDATE: Edited the paragraph slightly.

I just had a thought. If Mary was presented with a red, a blue, and a green tile on a white background could she identify which was which without additional visual context clues like comparing them to her nails? If not, I would expect a p-zombie to have the same issue implying that that failure isn't to do with consciousness.

If Mary had never counted more than 100 objects before, and today she counted 113 sheep in a field, we wouldn't expect her to exclaim "Oh, so that's what 113 looks like!"

I'd think that partly this is because 113 is an abstract internal representation - it can't be surprising, because we have to build it for ourselves. There is no "experience of 113" in normal human cognition. If I could just look at a pile of objects and go "That's 113" the way I go "that's 3!" or "that's 4!", I could imagine very well b... (read more)

Say I jam an electrode into Mary's optical nerve and send a pulse down it, causing the nerve to report to the brain that she sees red. In this hypothetical, her field of vision fills with red. Does this count as her experiencing the qualia of red?

If no: your concept of qualia is magical.

If yes: qualia doesn't do the work you want it to. With the word qualia, you're drawing a distinction between knowledge of the event and the actual event of the event - the event happening as distinct from a complete understanding of the event happening. This distinction is... (read more)

Yes. Who are you addressing and how do you know? The intended distinction is between complete theoretical or descriptive knowledge and knowledge by acquaintance. The distinction is worthless in that case. However, that does not mean it is worthless in other cases. The intended conclusion is that qualia are a unique case. It's possible that her ability to predict her own surprise won't involve an ability to predict what she is surprised at, how red looks. You are saying that Mary would know what red looks like while on the room? Why doesn't that solve the qualia problem for everybody?
Give me a case where this is a useful distinction. This is exactly the useless distinction, right here. You're drawing a distinction between knowing how red looks, and being able to act as if you know how red looks. (Being able to predict her surprise vs being able to predict the look of red). In practice, everywhere, there is no difference. The only use of knowing how red looks is to check that your model of neuroscience is accurate, and we've already specified Mary's knowledge is complete. I'm not sure if I am. The way I phrased the question, I think that saying "yes, she experiences the qualia of red" means that she knows what red looks like while she's in the room, and saying "no, she does not experience the qualia of red" means she doesn't know what red looks like while inside the room. I would be greatly honored if you could quickly outline what work you feel that qualia does - what theories does it support, what does it count as evidence against, what does it rule out and what does it prove, what impact does it have? What does it mean, if qualia existed?
The intended use is to tell us whether neuroscience is complete. That is a bold claim. Surely there is a subjective difference between predicting the surpise, and predicting the red? Or are you looking at it behaviouristically, from the outside? No. We have specified that it is descpriptively complete. The whole point of the story is to explore whether complete descriptive knowledge is complete knowledge. It would be helpful if you were sure. So which is it? That qualia exists means there is something lemons taste like, and saxophones sound like and sunsets look like. Beyond that...why do you need to know? Are you planning to deny that there omething lemons taste like, and saxophones sound like and sunsets look like if it leads to implications you don't like?

For example, we don't experience the feeling of ineffability for something like counting, which happens consciously (above a threshold of five or six). If Mary had never counted more than 100 objects before, and today she counted 113 sheep in a field, we wouldn't expect her to exclaim "Oh, so that's what 113 looks like!"

What about the case of 3 sheep? Are small numbers understood both analytically, and as qualia?

That's a really good question, and I just wanted to avoid that extra complication in this example. (As noted in the last post, I couldn't even fully do that. Oh well.)

I think I have long since "dissolved the problem quite elegantly" ...,_A_Philosopher%27s_Whore_of_Color_and_Her_Knowledge_Problem

Basically Yes Mary can "know all there is to know about color ...." before being exposed. Yes she does learn something new when exposed to color.

But basically the knowledge of how her brain reacts to color is information that does not exist in the universe prior to her exposure. She is physically changed and the new information is thus created.

If her knowl... (read more)

That's a fairly standard response which has been posed and answered several times in the comnents. For instance: That's not the point: the point is why is it necessary, in the case of experience, and only in the case of experience to instantiate it in order to fully understand it. Obviously, it is true a that a descirption of a brain state won't put you into that brain state. But that doesn't show that there is nothing unusual about qualia. The problem is that there in no other case does it seem necessary to instantiate a brain state in otder to undertstand something. If another version of Mary were shut up to learn everything about, say, nuclear fusion, the question "would she actually know about nuclear fusion" could only be answered "yes, of course....didn't you just say she knows everything"? The idea that she would have to instantiate a fusion reaction within her own body. That doesn't apply, any more than a description of photosynthesis will make you photosynthesise. We expect that the description of photosynthesis is complete, so that actually being able to photosynthesise would not add anything to our knowledge. The list of things which the standard Mary's Room intution doesn't apply to is a long one. There seem to be some edge cases.: for instance, would an alternative Mary know everything about heart attacks without having one herself? Well, she would know everything except what a heart attack feels like —and what it feels like is a quale. the edge cases, like that one, are cases are just cases where an element of knowledge-by-acquaintance is needed for complete knowledge. Even other mental phenomena don't suffer from this peculiarity. Thoughts and memories are straightforwardly expressible in words — so long as they don't involve qualia. So: is the response "well, she has never actually instantiated colour vision in her own brain" one that lays to rest and the challenge posed by the Knowledge argument, leaving physicalism undisturbed? The fact that th

Well, we could of course draw the analogy between colors of the spectrum and tones of sound

Puzzle: We sense colors, which exist on a continuum, by how near one color is to each of the only 3 colors our retinas can sense directly, plus intensity. We sense tones, which exist on a continuum, directly - we can sense each separate wavelength directly. Yet we have the impression that there are more colors than sounds - we draw sounds on a line, but colors in a plane.

If you're talking about only a single frequency of light or sound, a 2-dimensional point is enough to represent human perception—one dimension for frequency and another for intensity. However, if you're talking about the full range of colors and sounds that humans can distinguish, colors can be described with only 3 dimensions, while an ideal perceptual representation of sound would need a separate dimension for every functioning hair cell.
I figured it out. An ideal perceptual representation of sound would only need 2 hair cells - if hair cells, like cones, reported a distance from the stimulus. A cone cell gives a signal whose intensity indicates how far the wavelength of the light it sensed is from its preferred frequency. 1 cone cell lets you order colors along a ray. 2 cone cells lets you order them along a line. 3 cone cells lets you order them on a plane. A hair cell is specific to a frequency, so you can't combine the output from n hair cells to give an n-1 dimensional picture.
That's true if you're talking about a stimulus that only contains a single frequency at a time, but real sounds and colors are mixtures of an entire spectrum of frequencies, each frequency having its own distinct amplitude. For example, 2 hair cells, even if they had a wider frequency response, would not be enough to understand speech; for that you need at least 4 to 8 frequency bands.
There are more colors than tones, but there are more dimensions to sound than just tone.