Molyneux's problem | Hacker News Reader

Molyneux's problem

en.wikipedia.org

44 pointsmike_esspe45 comments

>In 2003, Pawan Sinha, a professor at MIT in Boston, set up a program in India as a part of which he treated 5 patients that almost instantly took them from total congenital blindness to fully seeing.[8][9] This provided a unique opportunity or answer the Molyneux's problem experimentally. Based on this study, on April 10, 2011, he concluded that the answer, in short, to Molyneux's problem was "no". Although after restoration of sight, the subjects could distinguish between objects visually as effectively as they would do by touch alone, they were unable to form the connection between object perceived using the two different senses.

That is a really counterintuitive result for me. I always assumed that one could construct a spatial model of something by touching it, without that directly being related to vision.

I think that's why the problem is so interesting and the answer so telling about the human brain. The way I picture it is that if you closed your eyes and held a ball, you would picture a sphere... but it is because you have in internalized image of a sphere. If you had never seen a sphere, you probably wouldn't be able to imagine one... even if you could feel all it's dimensions. Your mental 'canvas' is not only lacking paints... you don't even have a brush.

You'd think, because it all boils down to coordinates, right? Or at least relative coordinates. Apparently not. It's more like pattern matching, where the patterns are much more abstract than a coordinate system, where the connection between felt and seen patterns has to be learned, and the translation of either into "coordinates" has to be consciously worked at. I don't really know, not being anything like a neuroscientist.

I'm also reminded of the way babies are born seeing upside down, and learn naturally to invert their vision. Apparently if one wears glasses that invert their vision, it eventually flips back. Is that a built-in transformation, or is it deduced from first principles, just to make a simpler mapping between seen data and other data, e.g., to more simply reconcile feeling your hand move one direction and watching it move in another? If from first principles, what else is it doing? And we know the brain adds stuff to the picture before it's presented to the conscious mind, too.

> I'm also reminded of the way babies are born seeing upside down

Do you have a source for that?

I believe he means that the light receptors on your eye are inverted like this:

http://www.northlight-images.co.uk/content_images_2/digital_...

But the idea that people "see upside-down" is a common misconception. How the inputs map onto your brain spatially don't really matter. Well they do, but it doesn't matter if it's inverted or not because your brain learns (or likely is hardwired biologically) to map signals that come from the bottom to being located at "top" and signals that come from the top as "bottom".

Right, it's precisely that learning I'm interested in. I could well be wrong in that specific example. IIRC, I first got that from one of those "science" books for kids.

I suspect the GP is thinking of this or similar studies: http://en.wikipedia.org/wiki/Perceptual_adaptation#Experimen...

You probably can. You can't construct a spatial model of something by seeing it unless you've already gotten used to the various effects (perspective, parallax, etc) and how they represent space, though.

Or, at least, that's my totally untrained opinion.

You absolutely can, and there's even been recent research into the computational mechanisms that are behind this kind of inference [1]. They build a computational model that learns spatial/visual representation from (simulated) haptic stimuli with performance similar to human subjects.

I think the reason Sinha's experiment had a negative result was, as other commenters have mentioned, that this ability is a learned behavior. Those who are blind from birth simply never developed the ability to form abstract representations from visual stimuli, therefore haptic information cannot be translated into these visual representations. The other commenters who are saying there is no pathway between the two modalities are wrong, and it's easy to set up an experiment to show it. Just blindfold someone and hand them an unfamiliar object, then, without showing it to them, have them draw it. It's clear that we can do this to some extent.

[1] http://www.bcs.rochester.edu/people/robbie/jacobslab/abstrac...

Right, there is a physical difference between the brain of a person who has been blind from birth and someone who has not. This also means that if you restore sight to someone who once had it and then lost it, they should have no trouble picking out an object they hadn't seen (only felt), even if it was an object they had neither seen nor felt prior to losing their vision.

> I always assumed that one could construct a spatial model of something by touching it, without that directly being related to vision.

You can do that if you are able to see, close your eyes and touch it. What you'll be doing is seeing "in the mind's eye". Someone might have researched that, but you'll probably be exercising some cortex area related to vision.

Apparently, for people born blind, they can't make that connection since they don't have the experience to draw from.

I would think it's like singing two songs one octave apart and asking someone who was deaf which had a higher pitch. Even though he can hear, he wouldn't have this intuition developed yet.

The problem is fascinating in itself, but I don't see why it is classified as an unsolved problem in philosophy.

From the article:

"The resolution of this problem is in some sense provided by the study of human subjects who gain vision after extended congenital blindness."

The solution is purely empirical, not philosophical.

https://en.wikipedia.org/wiki/Unsolved_problems_in_philosoph...

Philosophical problems tend to turn into empirical problems with time. That’s completely normal. At some point in time you can’t actually investigate something empirically, so it’s a philosophical problem. As soon as you can, the philosophical problem turns into an empirical one.

I don't understand how it's a philosophical question. It's a question about the state of reality. Even before anyone tried to actually test it, it was still a question about the way things are. It's like asking what's inside a room, and calling it a philosophical question. Then you open the door and actually look and suddenly it's not?

The Schroedinger's Cat thought experiment is definitely a philosophical problem, yes.

Yes, that’s how it works, basically. Only that at the time you are asking the question, it has to actually be impossible to find out what’s inside the room.

Philosophy is for questions that cannot be investigated empirically. And that includes those that cannot yet be investigated empirically. That’s definitely where I see the role of philosophy.

Aren't all (interesting) philosophical questions about reality, or at least some extended or modified version of reality as we know it?

"What can be known?" "Do we have free will?" "Does God exist?"

(hint, any question that starts "Do we..." or references questions of existence is asking about the state of reality)

Possibly, though questions about morality or free-will, or the nature of knowledge or whatever, are not.

But what I meant was that almost any question can be considered philosophical under that definition. Until you actually test it scientifically at least. And if that's the case, then the meaning of "philosophical questions" becomes worthless. A word that can describe anything is useless. The value of a word is that it can be used to differentiate between things.

No, that’s not all! Philosophical questions also have to be relevant. That’s a squishy condition but I would really say it’s that simple. There is nothing special about philosophical questions per se, it’s whatever humans find intensely interesting and cannot (yet or ever) be investigated empirically.

(I’m pretty sure I agree with you that something like morality can, at its core, not be empirically investigated, though empirical investigation can help create clarity in arguments about morality, though I’m not sure whether that’s an absolute truth that cannot ever be changed. Free will? Nature of knowledge? Those obviously are ripe for empirical investigation. I don’t see why we should never be able to answer those questions conclusively and empirically.)

I don't know about free will because it depends how you define it and it's not something I've ever understood, though it doesn't seem like a concept that is dependent on how the universe actually is in a physical sense.

The nature of knowledge is pretty vague too, but if you mean the concept of how we can ever know things, that also doesn't depend on how the universe physically is. In another universe with different laws of physics, it would still apply.

The same is true of mathematics, for example.

Empiricism is a philosophical position.

(Furthermore, before finding a way to test this empirically, it was solely a thought experiment, and therefore, even more directly in the realm of philosophy.)

Because one of the fundamental questions in philosophy has always been whether knowledge comes from empirical sense experience or whether it comes from somewhere else.

Having a definitive answer to this question could be thought to potentially shed light on that.

Plato would probably argue that shape concepts are innately available to consciousness, and that when the blind person touches a sphere (for example), he connects that with his innate conception of a sphere. So Plato probably would have expected the suddenly non-blind person to be able to immediately connect his tactile experience of a sphere with his visual experience of a sphere.

British empiricist philosophers would probably have argued along the lines that there is no such thing as an internal conception of a sphere. So they would disagree with Plato.

Of course, the true answer is that neither Plato's team nor the empiricists were completely right. There are concepts, but they are formed based on sense data, not based on some kind of innate knowledge.

With respect to Molyneux and Locke, while I believe they get the general idea right, I believe they get the actual answer wrong. The answer to Molyneux's problem should be "yes" --- the newly sighted can correctly associate the shapes.

A sufficiently educated and intelligent person with congenital blindness would be able to correctly differentiate the shapes. Say, for instance, had Locke been born blind and still educated as rigorously. Of the many deductions that could be performed, the motion of the head with eyes fixed looking around the border of an object in space approximates the motion of the hand tracing the same object. Many other deductions are possible. All you need are reasonable assumptions about the acclimatization process, e.g., the patient can have a suitable amount of time to adjust to being sighted but cannot touch anything during the process.

It is fascinating to think about the idea of disconnected senses, but as so commonly happens, fascination with an idea leads to sloppy thinking, especially among the educated.

The reason why the empirical studies of the patients from India are wrong, despite being interpreted as giving a concurring answer, is that the blind have traditionally been given feeble educations, lacking in the type of rigorous thinking necessary to solve the problem. Historically it was challenging to teach abstract reasoning to the blind, and for this reason many were not taught. See Herzog's documentary "Handicapped Future" for tragic examples in relatively wealthy 1960s West Germany. http://en.wikipedia.org/wiki/Handicapped_Future

The five patients in India who could not afford simple but life-altering surgery were sure to be poor in addition to being blind, and so surely must have received awful educations. Compounding this are their young ages.

A sufficiently motivated, newly-sighted Locke would've gotten the answer correct.

I'm confused...Locke and Molyneux weren't saying the newly sighted person wouldn't be able to distinguish between the shapes, but that they wouldn't be able to synchronize that distinction with their touch sensations. In other words, they wouldn't be able to identify which was the one they identified, by touch, as spherical and which as square.

I do think that's what the poster is referring to:

> the motion of the head with eyes fixed looking around the border of an object in space approximates the motion of the hand tracing the same object

His argument is that this kind of abstract reasoning would be mostly developed by education, so people with very little education would not be able to associate the shapes they see with those they touched.

Education here is not an overriding factor. The way a congenitally blind person "imagines" abstract representations of shapes is substantially different, primarily because the physical wiring of the visual cortex and indeed a lot of the brain would be very different when compared to a person who could see. The dimensionality of their world, despite being the same as that of the sighted ones, may not involve the same descriptive system - ie maybe "polar" (just an example) coordinates would be more "natural" to them than a Cartesian system.

It's simply a problem of what gets mapped to what in the absence of mapping channels we take for granted.

Education is probably not a significant variable here.

I find it curious you would not consider education a significant variable.

Education and intelligence mean there's a significant variation in people's abilities to infer, to the point that particularly talented people get write-ups in 'Guinness Book of World Records.' and human-angle stories at the end of nightly news bulletins.

So the assumption that it's a binary answer to Molyneux's problem seems to be the first error. There would be certain individuals, who when adjusted to sight enough to work out the ratios of this color to that color could find enough data to make a choice that's better than a random guess.

However the fascination with this question isn't around those individuals who'd pass the test. It's fascination with the idea that most of us wouldn't, because as you outline the absence of input through the visual cortex mean the brain would not be able to make simple mappings sighted people feel are inherently 'natural'.

Calculus in Chinese is still calculus. The sphere is smooth, the cube is non-smooth. That's all it takes.

Yes, but what is a non-differentiable curve if you and I don't even have the same definition of a curve? What about a curve which is differentiable in one coordinate system but not so in another?

EDIT: My point is - mathematics is axiomatic in its very basis - the axioms have to be agreed upon by people who agree upon a conclusion derived from those axioms.

" The answer to Molyneux's problem should be "yes" --- the newly sighted can correctly associate the shapes."

In actual fact, the newly sighted can't correctly associate the shapes right away. They need time to learn how to negotiate sight.

There were two somewhat recent reports of this very thing[1], I remember reading about one where the man was perpetually overwhelmed and depressed after regaining his sight. It can be very difficult if not impossible to fully adjust, the theory is that the tactile areas of the brain simply overtake the visual processing centers, which are otherwise underused. It does take time for the brain to begin associating familiar but abstract concepts with their visual representations.

[1] http://news.psu.edu/story/141360/2006/04/17/research/probing...

If you could solve the vision problem through deduction, then a good decision tree would be enough to implement artificial vision. The fact that it isn't suggests that the problem is much harder than it seems. https://en.wikipedia.org/wiki/Moravec%27s_paradox

Well it may be that the brain is hardwired biologically to represent sight to some extent. But otherwise, to a newly sighted person that had not learned to associate visual input with anything, it would just be completely random noise. Neurons are firing but they aren't connected to anything yet.

It would be like asking a person if they could understand text written in a foreign language they don't know anything about. It might be possible, but you would first have to train their brain extensively to understand the format and map the different symbols and words to actual concepts. Otherwise it's all gibberish, random noise.

While I think you're correct that it may take some time for the newly sighted to be able to process the visual signal at all, the question as stated mentions:

  "... the blind man, at first sight, would not be able with
   certainty to say which was the globe, which the cube, 
   whilst he only saw them..."

The phrasing of "which was the globe, which the cube" presupposes that he is able to visually distinguish the two shapes sufficiently. If he can distinguish the shapes, this implies that he is able to process the visual input already enough to understand that there are two shapes. Given this, I can't understand how a sufficiently intelligent person would be unable to tell which is which through logical deduction.

Besides which, it would be a pretty lame philosophical question if the man couldn't even process the visual input to know that there are any shapes at all.

My misunderstanding.

I need to update my citation:

"Land of Silence and Darkness" is the better source, though both may cover the subject. The two films are closely related. I think I saw them together, hence the confusion.

http://en.wikipedia.org/wiki/Land_of_Silence_and_Darkness

Wow, that's really cool.

Now I have an additional question. Assume at one point you were able to see and touch things, then lost your sight, then regained it. I'd think that while blind, you'd still try to form images of things you touched, and then be able to recognize them after your sight was recovered.

If that's true, then the connection between the senses would not be fundamentally impossible, but rather a learned behavior.

I expected to read about Peter Molyneux's problem of producing and selling games he dreams up around strange concepts, which I thought became so well known that it got an official name and put in Wikipedia.

But this was more interesting !

I think that if the previously blind person analyses the shapes in the right way, they'd be able to tell which one is the sphere and which one is the cube:

The only things the person would be able to gain from just touching the objects is that the sphere is "the same all over" and the cube is "not the same all over", to dumb it down considerably. If the newly unblind man looks at the objects in a similar way and considers symmetry, I think he would be able to tell that one is infinitely more symmetrical than the other and would be able to deduce that the shape that looks the same all over would be the sphere.

Does that make sense?

This reminded me of a feature of Esref Armagan, a blind Turkish painter, who was never sighted.

Skip to 7 minutes to see him paint a building he's never been to before, with 100% correct perspective.

http://www.dailymotion.com/video/x9ig6p_esref-armagan-blind-...

I wonder if he was given sight and at the same given two paintings of a scene, one his and another by someone else if he would be able to tell the difference.

Surely this is not a hypothetical anymore or one with merely one test case - see http://en.wikipedia.org/wiki/Visual_prosthesis for example.

It seems obvious to me that it takes some time for people to get used to using a new sense - particularly for higher-order abilities. Differentiating 'round' from 'hard edged' is probably a fairly simplistic function which your brain should be able to quickly figure out, but I don't see how it can just be instant.

I think this is exactly right. Tactile visual substitution systems provide a sensory modality that we may as well call sight, just not via the eyeballs, and it takes time for the blind to become capable with them. Why wouldn't it take time for them to become capable with a sensory modality that is causally associated with the eyeballs? And in fact, it does take time.

http://plato.stanford.edu/entries/molyneux-problem/

I say yes.

If you asked a blind person to draw a sphere they would draw a circle, if you asked them to draw a cube they would draw a square.

Reply on news.ycombinator.com