Discussion for Nov. 16., Notes by David Carlson

AIDA: An Adaptive System for Interactive Drafting and CAD Applications

Question 1:

Is there a range of knowledge for which this approach is appropriate?

(No discussion on this topic)

Question 2:

How do you handle differnces in user knowledge?

(No discussion on this topic)

Question 3:

Should you allow for fuzzy knowledge?

The observation was made that most observable user tasks are not fuzzy.
For example, delete and cut are fairly unambiguous.  The variable affecting
this is the granularity of the observation.  In the AIDA implementation 
focus is on tasks which share this lack of ambiguity.

Concievably, at coarser levels of granularity, the level of ambiguity would
rise.  It is unfortunately not obvious how these would be noted or represented.

Another participant noted that the number of dimensions for recording human
activity is too high.  Small increases in granularity size cause large 
increases in overhead required to recognize and represent the activity.
"The problem grows exponentially." No basis is given for this claim.

Question 4:

Can you tailor the commands to users?

The suggestion is for users to start with a generic set of tasks and 
grows with the user's needs.

The problem is that the view of the specific user remains the same.  It is
not obvious that this is an improvement.

In an unconnected thread, it is suggested that the # of representations
required before the suggestion of a macro be modifiable.

Further the presenter suggested that the user may require tutoring to use 
the available macros.  This seems to be against the point of the system.

Question 5:

Is this approach applicable to other fields?

Discussion here focused not on the suggested question but on whether the
paper followed through from the promise of the introduction.  All parties
agreed that the rest of the paper was anticlimactic.

The response to the final question was simply yes.  The principle concept
of this system is simple and transferrable.  It may be that the approach is
simply barking up the wrong tree. Use oriented feedback as provided in AIDA
is "nice" but not particularly useful.  A useful system would do as the
introduction of the paper promised which is task oriented feedback.  We had
already discussed difficulties with trying to do this.