Re: [Phys-L] multitasking

[John Denker <jsd@av8n.com>]

The charter of this list centers on teaching.  Successful teaching 
implies the students are /learning/ which has to do with /memory/
and /thinking/.  Memory itself is a thought process, comprising
both memorization and recall.

My point for today is that the vast majority of human thought is
done in parallel.  Note the contrast:
 *) the parts that are easy to talk about -- namely the parts 
  involving conscious attention -- are mostly serial, but
 *) the other 99.999%  -- including perception, learning, and
  recall -- is massively parallel.

Let's be clear:  In this thread, everything that has been said
(by me and others) about "division of attention" is misleading.
Even the parts that were 100% true and useful were misleading,
because they missed the most important point:  Even though 
/attention/ is in some sense serial, most brain function is 
massively parallel.

The cycle time of human neurons is on the order of 20 milliseconds.
Shifting conscious attention from one thing to another takes even
longer than that. The fact that we can do anything useful absolutely 
depends on doing massive amounts of stuff in parallel.

As a simple example, consider the test that is commonly performed
in the eye doctor's office, using small flashes of light to test
your peripheral vision.  If you had to consciously scan the visual
field looking for a flash here or a flash there, you would score
zero on this test.  The point is that the retina and the visual
cortex have millions of pixels and are wired to detect movement
in general and flashes in particular.  The multi-million-pixel input 
gets processed at the pre-attentive level, so that a single signal
describing a single flashed gets passed up to the conscious attentive
level.

Let's see what the the roughly 20 millisecond cycle time tells us
about memory and recall.  That's fifty cycles per second.  I'll bet 
you know more than 50 things that you can recall in less than one 
second, such as
  -- first name, middle name, last name
  -- parents' names, spouse's name, kids' names
  -- phone number, address, email address
  -- number of cm per meter
  -- et cetera

It should be obvious that you do not recall those things by consciously
running down a /sequential list/ of things you know, nor by any other 
serial process.  You recall things by sifting your memory.  Such sifting
is massively parallel.

Remember: The word that can be spoken is not the true Word.  The part 
of the thought process that we can easily describe is a very tiny part 
of the whole process.

This is profoundly important in connection with the metacognition
issues that we recently discussed in this forum.  One of the arguments
in favor of "peer instruction" is that it gives students the opportunity 
to discuss their conscious thought process with other students. That's 
fine as far as it goes ... but it is incomplete and highly misleading,
because the conscious thought process is only a very small part of the 
total thought process.  It is only the flea on the tip of the iceberg.  

As an illustration, consider the famously ill-posed "Mississippi flow" 
question.
  http://www.av8n.com/physics/thinking.htm#sec-miss-flow
  http://www.av8n.com/physics/ill-posed.htm#sec-mississippi
This is a hard question ... but almost everybody concedes _in retrospect_
that they "should" have been able to figure it out.  I mention it 
because the first time I figured it out, I used Samuel Clemens's pen
name as one of the inputs to the calculation.  That leads us to ask,
how did I know that was relevant?  Obviously I did not run down a
sequential list of all the books I had ever read, let alone all the
data about authors of books.  Rather, there was a massively parallel
query, sifting my memory for information connected in any way to 
the Mississippi.

The most important question we could possibly discuss is this:
     _How do we teach people to be smart?_

Being smart requires being able to recall and/or figure out information 
when needed.  Effective recall means being able to conduct a massively 
parallel query of your memory.  Ironically, you (mostly) cannot teach 
this by focusing on recall.  You (mostly) cannot do it by demanding 
that the student "try harder" to recall the needed information ... 
because by that time it is too late.  Recall depends on the /connections/
between ideas, and those connections are formed over time, as various
ideas are learned and used.  Therefore, even though the goal is effective
recall, this is achieved by means of effective memorization i.e.
effective learning.  It has been known for a very long time that the 
right way to learn something is to turn it over in your mind, looking 
for connections to other ideas.  This takes time and effort.

This connects to our recent discussion of how to organize the introductory
course ... but I will defer this point to another message.

I don't claim to know very much about the process of learning and 
recall ... but you don't need to know very much to know that it's
not serial!

Note that database query languages such as SQL allow for massive
parallelization of certain operations.  SQL is an example of declarative 
programming.  A spreadsheet is another example of declarative programming.
This stands in contrast to imperative programming languages such as C++,
which are much harder to make massively parallel.

All in all, I am astonished and horrified when I see things like this:
  http://www.npr.org/templates/story/story.php?storyId=95256794

which revolves around the assertion that:
> researchers say [multitasking is] still a myth — and they have the data to prove it.

Data, schmata.  These "researchers" obviously have not the slightest 
understanding of human perception, thought, or learning.