From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!jvnc.net!darwin.sura.net!ukma!memstvx1!langston Mon Mar  9 18:35:41 EST 1992
Article 4303 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!jvnc.net!darwin.sura.net!ukma!memstvx1!langston
>From: langston@memstvx1.memst.edu
Newsgroups: comp.ai.philosophy
Subject: Re: Definition of understanding
Message-ID: <1992Mar6.001859.1518@memstvx1.memst.edu>
Date: 6 Mar 92 00:18:59 -0600
References: <1992Mar4.210627.28060@psych.toronto.edu> <1992Mar5.001144.28065@beaver.cs.washington.edu> <1992Mar5.203720.4209@psych.toronto.edu> <1992Mar6.012217.25722@news.media.mit.edu>
Summary: reply to Minsky
Organization: Memphis State University
Lines: 87

In article <1992Mar6.012217.25722@news.media.mit.edu>, minsky@media.mit.edu (Marvin Minsky) writes:
> 
> ========  Understanding Understanding =============
> 
> It seems to me that this discussion has become somewhat unproductive
> because of non-agreement about the meaning of "understanding". Some
> discussants assume that whatever "to understand" means, it is an
> absolute, all or none attribute of a system.  Others say that it must
> depend on something called "semantics" but aren't very clear about
> what a semantic might be.

        --- argument for perception of understanding deleted ---
 
> Second, there is an implicit idea that there is a definition or meaning to 
> "understand" -- provided that we could only agree upon it.  My view is that 
> this word, like many other terms from commonsense psychology, have a 
> very different and very useful usage, but one that is quite different from an 
> absolute two-argument relationship.  Namely, consider the commonsense 
> assertion that if a certain person  P can tell you only a certain definition D 
> of a term W, then you are inclined to say things like "P doesn't really 
> understand W, but has only memorized the definition by rote."  
> 
> Here's my explanation, then of why you philosophers are having so much 
> trouble agreeing on what 'meaning' or 'understanding" means.  It is partly 
> because meaning is no single thing, nor is understanding a single act.  But it 
> is really a bit worse than that, because there is a subtlety implicit in what 
> "ordinary people' are doing when they use the term.  To put it in the form 
> of an epigram (that is, a definition):
> 
>    "If you understand something only one way then you scarcely understand 
> it at all."
> 
> This is in recognition that the activities of human thought engage a large 
> society of different structures and processes.   For example, to illustrate 

          --- modified script explanation deleted ---

> Now what is the value of this?  Simply, that if you "understand" something 
> only one way -- that is, using only one representation -- then when 
> something goes wrong, you have nowhere to go.  And that is the dead end 
> that we by by the vernacular complaint that "A knows B only by rote."  
> The point is that with a single definition of representation, A is very likely 
> to get stuck in a way that we describe as "not really understanding".
> 
> There's more of course.  Because one would also say that the secret of what 
> X means to us lies not only in isolated definitions but also in how our 
> representations of X connect to other things we know.  In my view, this is 
> where logical representations usually get stuck.  There's nowhere to go.  
> But brains (and computers) can use multiply-connected representations, so 
> that when one approach fails you can try another.  Like turning ideas 
> around in your mind and trying out alternative perspectives till you find 
> one that works.  And that's what we mean by thinking! 

  Perhaps it is not the case that 'understanding' is inherent in memory
structure.  I would argue that for an agent (read agent however you like...
whatever makes this most interenting...) to 'understand' a concept (or what-
have-you) the agent must be able to MAKE USE of that item in some meaningful
way.  One could say "I understand golf" without ever actually playing golf.
Does this person really understand golf?  I could learn the procedure for
baking a cake, but without ever actually baking said cake, I would be at a loss
as far as true, useful understanding of the complexities of the process.
  This has implications for AI.  Does this mean that computers and such
'understand' their internal data representations?  Not any more than the rock
'understands' heat or molecular physics as it expands in the sun.  To
understand, the agent must act on the information in some meaningful way that
would integrate both semantic and episodic knowledge of the information at a
deep level.
  When we want to judge whether or not a student understands a concept, do we
not ask the student to demonstrate this understanding in some meaningful way?
understanding is not simply being able to represent the interconnections among
various related structures, but being able to make use of these
interconnections
in some way that is beneficial (not to be read as 'not harmful' - bad
experiences sometimes contain important information) to the agent.

(I don't disagree with you, Marvin, I only think you may have missed
something.)
 
-- 

Mark C. Langston                                  "What concerns me is not the
Psychology Department                              way things are, but rather
Memphis State University                           the way people think things
LANGSTON@MEMSTVX1.MEMST.EDU                        are."     -Epictetus

     "...a brighter tomorrow?!?  How about a better TODAY?"  -me



