Re: Clarification about mirror neurons

On 10 Jan 2013, at 14:33, Rami Rustom <[email protected]> wrote:

> On Thu, Jan 10, 2013 at 6:29 AM, David Deutsch <david.deutsch-LR/[email protected]> wrote:
>> On 9 Jan 2013, at 22:55, a b <[email protected]> wrote:
>> 
>>> - Does accepting specialized regions for language imply accepting
>>> condition s related to language impairment (like dyslexia) are
>>> possibly legitimate?
> 
> I have a analogous question.
> 
> - Does accepting specialized regions for facial recognition imply
> accepting conditions related to reading social cues (like
> "Autism/Aspergers") are possibly legitimate? (I'll use David's answer
> below to answer this question.)

I agree that the answer is the same.
> 
> 
>> If there are specialised regions for a computational function in the brain, the unspecialised regions are still universal and can learn to perform the function.
>> 
>> But the very term "specialized" can be misleading here because it can mean many different things. Because of universality, it is impossible for neurons to be 'specialised' in any computational sense. So a *computationally* "specialised region of the brain" can only be one in which the connections between neurons are genetically preprogrammed to enact a particular algorithm. But understanding human language, and using it, are both creative functions that are not inborn (we are not born able to read, speak etc). And therefore those functions cannot be pre-programmed. So they have to be learned creatively.
> 
> Understanding that a frown (together with all the other signals that
> one could gather from another person) and then deciding that all of
> those signals taken together means that he is sad, is a computational
> function -- hence anybody can learn it because of the universality.
> 
> But, registering that a person's face is making a frown, is not computational.

I don't see why it isn't. The only way frowns might not be detectable through general-purpose thinking about what we see, would be if the data from the eyes was only presented to the general-purpose part of the brain after automatic and lossy processing, but was presented to the supposed frown-recognition centre in raw form. Something a bit like that does happen when the eye itself does information processing, discarding what its algorithms reject as erroneous data.

(Even when the information is lost, we can still learn how to take that into account, and not be fooled by the erroneous interpretations placed on our sense data by hardware, whether defective or not. Painters have to learn to do precisely that, for instance. And anyone can detect their own blind spot with simple experiments, and learn not to make various mistakes based on of thinking that their visual field is like the image on a TV screen.)

However, none of that can be the case for frowns, because we can detect them in hugely degraded images.
> 
>> A region could be specialised in another sense, e.g. it could have exclusive access to signals coming from the optic nerves. If so, then if that region was damaged, the person might be blind or have difficulty seeing.
> 
> Do you think that there is a similar part of the brain responsible for
> registering the minute differences in facial expressions?

No.

> If so, is this computational?

If it existed it would be computational, and so could be done by general-purpose thinking too, unless the information flow was lossy in the above sense. 

>> But that does not prevent the person from performing the *computational* task associated with reading. For instance, they can learn to read Braille, using a different part of the brain.
> 
> So, if there is a specialized part of the brain that registers facial
> expressions, and if a person has some genetic difference that causes
> that part of the brain to not be functional (or brain damage to that
> part of the brain), the person can still use his universal sight
> recognition to *learn* to register facial expressions.

Yes.

-- David Deutsch