Discussion of Adaptive Hypermedia for Support Systems, 
     by Gernoth Grunst

Discussion notes by Tonya Beers
Presentataion by David Carlson

     Discussion started out talking about the following:

               system    user      
initiation                o
proposal         o
decision                  o
execution                 o

     Matthias believes that this configuration is completely
wrong, because the system adapts the help contents.  The user
initiates the help, the system checks the request and does the
rest (decides what to provide, and executes action).

     Another view is that the user decides among several
proposals, in which case the configuration would be the
following:

               system    user
initiation                o
proposal         o
decision         o        o
execution        o
     
     With this in mind, perhaps it is best to have the expert
sitting next to the user.

     Q:  In a perfect system, what type of adaptive help system
would be desired?
     A:  Obviously, this would be one in which the system did
everything.  But more realistically, it is necessary for the user
to indicate when it is appropriate to initiate various actions. 
Perhaps this is the best configuration possible:

               system    user
initiation                o
proposal         o        
decision         o
execution        o

     There is also some interest in having all of the boxes
filled, but this is not a likely solution.


     If a human expert were to be used as a help system, he/she
would constantly need to interrupt the user.  The human expert
would only get input based on what the user typed in.  He/she
would not know much about the user's task model, and would
constantly need to probe the user to determine what he/she was
trying to accomplish.  Experts don't know when it is advisable to
interfere.  The computer on the other hand can remember long
term, and have a good knowledge base of previous actions on which
to decide on the help to be provided.

     Perhaps we should at least have the option available to
place the emphasis less and less on help systems, and more on
learning systems.  Just providing an answer does not do much for
the user.  Of course, this need will vary with circumstances.

     Process of supplying adaptive tasks has many steps:
          1) What does the user want to do? (human expert is
             probably the best model for this)
          2) How to present the help? (expert could show them
             how, but would probably not provide the best         
             example)
          3) When to present the help? (experts not good at this)

     The last question presented for which there was not much
time for discussion was regarding how human interaction should be
presented.  Is HyperText a good mimic?  If so, what form of
HyperText?  Is it even good to try to mimic this?