J. Vassileva
A practical Architecture for User Modeling in a Hypermedia-based
Information system.

Oct. 10, 1994, Notes by David M. Carlson

Should systems support ad hoc queries?
	This question seems to depend on the content/form
	of the underlying info., and on the users task (browsing/
	searching). Ad hoc queries could be supported through
	use of template queries.

	Browsing/Searching decisions involves certain inherent tradeoffs:
	  SQL - more complete, faster, hard.
	  Browse - easier, often get incomplete answers.
	The user's task and role may determine the appropriate approach.

	One student mentionned that the outline of the system is not
	modifiable by the users.  Using SQL may fit the into the role 
	of the administrator.

How does one support new groups of users?
	One approach is to use a default:
		Task-based approaches may be useful to determine needs
		  of new classes of users. One group's tasks may be 
		  similar to another's, thus they can serve as a starting
		  point for a new class of users.
		Superclass - Facilitator class with wide range of access may
		  be useful?

How does the group model support co-operative work?
	There seem to be fundamental incompatabilities between AUI and
	CSCW.  Meta-information about tasks are hard to quantify when
	work is collaborative.

	One student suggested that instead of adding CSCW support to
	hypertext, which is commonly the practice, we should see these
	as completely different things.  An alternative approach is to
	add hypertext to CSCW systems.

Would such an architecture support other domains?
	disappointment: Only a few ways exist to get information.
	  Want to switch views dynamically.  Architecture does not
	  scale well.
	Is this supported with user models?
	  Some people want multiple approaches some do not.
	What is the cost of this flexibility?
	  N-pigeonholes for N-presentations.

	With limited tasks, you can include the task model in the user
	model.
	The Win of doing this is in bringing myltiple models into a single
	model.  This simplifies the maintenance of these models.

	Finally, these systems are very brittle in that they do not transfer
	to other domains easily.