Re: Clarification about mirror neurons

On 2 Jan 2013, at 13:26, Alan Forrester <alan_forrester2-/[email protected]> wrote:

> On 30 Dec 2012, at 15:17, Alan Forrester <[email protected]> wrote:
> 
>> In BoI, David explicitly says that mirror neurons don't explain human imitation. However, there is another problem with the mirror neuron idea.
>> 
>> Neurons, or some set of neurons, can act as a set of universal classical gates. So mirror neurons (or some subset thereof) play a role in imitation by virtue of instantiating some algorithm. The explanation of how the imitation is done is in terms of that algorithm rather than in terms of the neurons per se. So then calling them mirror neurons is a bit like saying that a universal computer that happens to be set up to implement a chess playing algorithm is a chess playing chip when that chip could just as easily be programmed to do any other classical computation.
>> 
>> This could lead to mistakes. It could be the case that some of the mirror neurons could be cut out or destroyed without destroying the mirror functionality. They happen to go off because they're wired up to other neurons that do the mirror neuron algorithm. Worse, it could be the case that there are two sets of neurons A and B and you could destroy either one of them without destroying the mirror functionality but if we destroy both sets then we destroy the mirror functionality. In the absence of a specific explanation it's important to think of the mirror functionality as being software so we don't miss stuff.
> 
> In animals, the neurons used to do imitation may not change much since the knowledge in the animal's brain is coded for by genes and the animal creates no new knowledge. So then looking for neurons that go off when an animal does something and when it recognises another animal doing the same thing, mirror neurons, may provide a way of testing ideas about mirror algorithms. Having a term for that idea makes some sense and that seems to be David's position in BoI.

One reason I think it's vital to take that position is that the commonsense theory of how animals, especially apes, can imitate complex actions is that they do it by understanding what the actions are for -- i.e. the same way that humans do it. Indeed, that misconception was one of the motivations for biologists to search for mirror neurons in the first place. 

It is an attractive misconception because copying even a simple action does indeed require a large amount of knowledge to be present in the copying animal. So the traditional misconception that knowledge requires a knowing subject then leads directly to the conclusion that an ape must be a knowing subject.

To make the arguments in BoI about what humans are and how they evolved, one must therefore refute those misconceptions. One thing I referred to for that purpose was the excellent research of Richard Byrne, which showed that apes have no idea of the purpose of their complex meme-encoded actions, which are not directly copied but are composed of several copied actions put together in coordinated ways. And he showed how those composite actions are learned. But that learning process requires apes to be able to recognise and copy the individual actions, and Byrne's research does not tell us how they do that because it happens inside the ape's brain. So it is vital to explain how that is possible.

The answer, in short, is that the necessary knowledge is encoded in the ape's genes, and implemented in a special-purpose computational system in its brain, which translates from sensory to motor signals. But only a particular class of actions, fixed once and for all in the animal's genome, can be copied using that system.

Would it not have been perverse, given that it was essential to argue that such a system must exist in the brains of apes, for BoI not to mention that a plausible candidate for that system has recently been discovered? Or not to call that system by its conventional and perfectly reasonable name, namely mirror neurons?

-- David Deutsch