Notes from 3-OCT-94   
Walter Patterson   walter.patterson@ce.gatech.edu

Questions posed for discussion (actually a fairly loose discussion
  after the first few questions were posed by Robin)


[1]  (with regards to adaptive design) Isn't this just AI?

   a.  No.
   b.  Similar to Expert Systems.
   c.  Not a Turing test.

        Iterative changes made by designers (at expense) could be avoided
   by allowing design decisions to be deferred - but then only the
   later implementations will be successful.  Who will buy the earl
   products?

        Taxonomy - we are still building "reflexive" systems - at the
   biologic level of the bacteria discussed in the article.  There is
   a large overhead in more complex, introspective or self-modelling systems
   that haven't yet been achieved.  The reflextive ("run and twiddle") 
   behavior can be very successful, but is it enough for adaptive HCI?


[2] What's new in this paper?

   a.  The idea of delaying design decisions to allow them to become
       the adaptive portions of the interface.
   b.  A tendency away from Usability Testing?  perhaps, the implication
       is that these systems should somehow "test themselves" is 
       probably overreaching.
   c.  Should the user be left out of these choices?  No, a system
       should not alone "decide" and implement what is "best" for the
       user, but if the user has control, is it still an adaptive interface?
   d.  Should User/Task be combined as one idea?  (from the perspective
       of the system).  Yes, it is useful to model users in the context
       of the task domain.
   e.  The paper states that the designer should not be responsible for
       making decisions about user preferences; that task should be left
       up to the (run-time) system.  But the designer is the one who 
       designs the system, so this argument presently involves some hand waving.
       Instead, perhaps the users should have some simple means of programming
       the environment.


[3]  Evolving systems such as E. Coli - Adaptation by allowing the users
     to create new responses to specific situations - these responses 
     are learned and become preferred or accepted responses for future
     situations.


[4]  Jim Pitkow:  The Domain of Adaptibility should become evident very
     early in the system design using prototypes, etc., so that objects
     requiring adaptibility can be identified. Typically, designers 
     select which objects can and cannot be manipulated by the user
     early in the design process.  If these selections are inappropriate,
     the interface suffers at the user's expense.  Early, rapid user-centered
     design can alleviate this problem of users' tasks falling outside
     the domain of adaptibility.