From weng@cse.msu.edu  Sun Mar  5 15:25:12 2000
From: Dr John J Weng <weng@cse.msu.edu>
Date: Sun, 5 Mar 2000 15:25:09 -0500 (EST)
To: wdl@cse.msu.edu
Subject: WDL Dis. #17

======== WDL Dis. #17 (Bcc: all WDL participants) ======

Dear Susan, 

I understand your points in emphasizing genetics in development.
They are consistent to my views: Genetics has "designed a lot"
in the first four levels (timing, architecture, representation
and inborn behavior) including the representation level that you 
emphasized.  It is a major mandate of this workshop to discuss
what they are and how to understand and design them. 

However, I think the term "prior knowledge" is too vague and 
misleading.  Some papers that were arguing against each other were so 
vague that the same terms were used for different meanings.  Often,
they are not really not incompatible: what they argue is mainly
emphasis. It is more beneficial to identify precisely what the 
term means. That is why I listed the five levels.  I think it is 
better to use the term "knowledge" for information that is acquired 
after the birth.  But I am not interested in arguing just for terms.   

Why ``ad hoc?'' Currently, the main paradigm in psychology is to 
study phenomena of human cognition and behavior.  This is the beginning 
stage of a young scientific field.  Early studies of a field tend to 
be ad hoc in nature since they can only explain a special case under 
a specific condition.  This is what is called phenomelogist stage of a 
field. When the field matures, systematic knowledge is discovered and all 
the known phenomena that were observed (including those that could not 
be explained clearly and were subject to heated debates) can be 
explained clearly by newly discovered systematic natural laws.  This 
maturation process has happened in physics, astronomy, chemistry, 
and more recently in biology, thanks to the advance in genetics.  
For example, genetics gives a more systematic picture of biology 
than the early phenomelogist studies of biology.

I am not saying that the entire field of psychology is ad hoc.
Some researchers in psychology have begun to move in the right 
direction, which marks a move to maturation of a field.  They have started 
to study the developmental root (timing, architecture, representation) 
and developmental models of human cognition and behavior.  
I just mention a few of them here: J. McClelland, J. Elman, K. Plunkett,
G. M. Edelman, and E. Thelen, just to keep the list short.  Yes, there 
are decisions made in these levels, but these levels systematically 
determine, in conjunction with experience, what comes later after birth.  
It is easier to attribute something unknown or unclear now to "innate" 
than to explain exactly what "innate" is, how it works, and exactly 
what degree of freedoms there are. etc. Unless the field is serious 
about clearly explaining its subject of study, the phenomelogist stage 
of a field will continue.  

I am not saying that AI is not ad hoc either.  AI is also at a 
phenomelogist stage, due to the lack of knowledge about intelligence 
development.  AI people should study the results in neuroscience and 
psychology, especially the recent ones.  (E.g., I wish that David 
Marr studied his theory in the context of development, since what he 
was talking about was a time snapshot of the later stage of continuous 
development of early visual processing.  I am not saying this as a joke.)  

I predict that the biologically motivated computational models of 
automatic development will also have a positive impact on neuroscience 
and psychology.  This is the synergism that this workshop is aiming 
at.  

Folks, please feel free to air your views. 

John