From weng@cse.msu.edu  Sun Feb 27 16:32:38 2000
From: Dr John J Weng <weng@cse.msu.edu>
Date: Sun, 27 Feb 2000 16:32:28 -0500 (EST)
To: kurt_fischer@harvard.edu
Subject: WDL Dis. #6

======== WDL Dis. #6 (Bcc: all WDL participants) ======
Dear Kurt,

I have updated your abstract.
Your description about moving down to low skill levels is a very 
interesting process of development.  

When learning a new task (e.g., learning riding a bicycle), the learner 
must choose each detailed move with a lot of uncertainty (trial and error).
However, when performing a skilled task (e.g., riding bicycle everyday 
after he has learned), he does not need to pay close attention 
to it any more and he can even do another task simultaneously 
(e.g., talking while riding a bicycle), while attention is shared 
between two tasks. In other words, riding bicycle has been "moving down" 
automatically to lower skill levels through development.  If the mind 
had to take care of every subtle move for everything it controls, 
it would have been too busy to learn any new, more complex tasks.  
It seems that a developmental robot must automatically abstract its 
skilled behaviors and "move them down" to lower levels.

Are you aware of any computational model that realizes  
this kind of "moving down" process through automatic development?

Any comments from other participants?

Best regards,

John