Project Notes
Learning on Demand, Incremental Learning and High-Functionality Applications
relevant computational substrates:
1. Microsoft Word and Office
2. Did you know (e.g., Tip of the Day)
3. Office Assistant and other assistance programs
relevant theoretical dimensions and background
information:
- papers:
- Knowledge-Based Help Systems ---> see: G.
Fischer, A.C. Lemke, and T. Schwab, Knowledge-Based Help Systems,
in Human Factors in Computing Systems, CHI'85 Conference Proceedings
(San Francisco, CA), New York, pp. 161-167
- ICM ----> see: G. Fischer, Enhancing Incremental
Learning Processes with Knowledge-Based Systems, in H. Mandl and
A. Lesgold (eds.), Learning Issues for Intelligent Tutoring Systems,
Springer-Verlag, New York, pp. 138-163.
- Making Information Relevant to the Task at Hand
---> see: G. Fischer, and K. Nakakoji, Making Design Objects
Relevant to the Task at Hand, in Proceedings of AAAI-91, Ninth
National Conference on Artificial Intelligence, AAAI Press/The
MIT Press, Cambridge, MA, pp. 67-73.
- Co-adaptivity ---> W.E. Mackay, Co-adaptive
Systems: Users as Innovators, in CHI'92 Basic Research Symposium
and G. Fischer, Shared Knowledge in Cooperative Problem-Solving
Systems - Integrating Adaptive and Adaptable Components, in M.
Schneider-Hufschmidt, T. Kuehme and U. Malinowski (eds.), Adaptive
User Interfaces - Principles and Practice, Elsevier Science Publishers,
Amsterdam, pp. 49-68.
- brief analysis of an example: Office Assistant
1.0 (integrated with Microsoft Office 97). The system selects
potientially relevant help pages based on the user's input words
as well as an analysis of the user's recent actions. Action analysis
is performed by a Baysian Network whose nodes either represent
user actions, user plans or user needs. User needs are then linked
with relevant help pages. Office Assistant 1.0 is not the first
commercial software with user-adaptive technology. It is however
the first commercial software that gives user monitoring and user-adaptation
a very prominent position. The fact that Microsoft bundled this
technology with one of its most visible products is both a good
sign for Microsoft's trust in user adaptation as well as a good
omen for the impending adoption of this technology by other software
producers. The usage of Bayesian networks for user modeling purposes
has been extensively discussed in UMUAI, as early as 1992 and
very recently in several articles in the special issue on numerical
uncertainty management.
relevant previous work:
analysis of Word (Jim Sullivan) ---> see: J.
Sullivan, A Proactive Computational Approach for Learning While
Working, Ph.D. Thesis, Department of Computer Science, University
of Colorado
Bob Gatewood's project in Information Society
class
analysis of cognitive tools (Norman, Landauer)
McGuckin study ---> see: G. Fischer, and B.N.
Reeves, Beyond Intelligent Interfaces: Exploring, Analyzing and
Creating Success Models of Cooperative Problem Solving, in E.
Rich and D. Wroblewski (eds.), Applied Intelligence, Special Issue
Intelligent Interfaces, Kluwer Academic Publishers, pp. 311-332.
interesting embedded topics:
Motivation
- engagement in self-directed, authentic problems
- make information relevant to the task at hand
- create interesting and exciting products
- provide multiple learning opportunities
- provide challenges matched to skill levels
- create communities (among peers, over the net)
- provide access to real practitioners and experts
- challenge: reinterpreting motivation at an organizational
level
- who is the beneficiary and who has to do the
work?
- memories: what will make employees want to share?
- people need to make explicit what they know
and take the trouble to enter it into the system
Revisiting Critiquing
- theoretical grounding: reflection-in-action, breakdowns,
making argumentation serve design, information retrieval by context
rather than query (making the context of the query known to the
system)
- tool (LC, technical editing) versus domain (kitchen
design, LAN design)
- embedded critiquing
- conversation with the materials (see Winograd,
p 206) ---> our claim: the back-talk of the materials is often
not good enough
- Winograd, p 44: "the difficulties with these
taxonomies and rules is that a design that serves well both the
particular material and the particular audience cannot be adduced
from principles alone: it requires a leap of invention"
- see story of tutoring versus simulation in Schoen/Winograd,
p 180 (and Schank in Aspen article)
- critiquing = "making the good better and
the bad more difficult"
- "once you know what you don't know: that
is the opening of all learning to occur"
Design Frameworks addressed:
production paradox
- people are not interested in learning per se,
but in
- ---> see: M. Carroll, and M.B. Rosson, Paradox
of the Active User, in J.M. Carroll (ed.), Interfacing Thought:
Cognitive Aspects of Human-Computer Interaction, The MIT Press,
Cambridge, MA, pp. 80-111
Designing for Letting "People Get By"
- main street and side street metaphor ---> see:
G. Fischer, and H. Nieper, Personalized Intelligent Information
Systems, Workshop Report (Breckenridge, CO), Institute of Cognitive
Science, University of Colorado, Boulder, CO
- suboptimal plateaus
- too much information in the abstract (e.g., WORD),
but not enough information in specific situations (e.g., eliminating
an erroneously included word from the spelling corrector) --->
address this problem with "programmable / scriptable /end-user
modifiable environments"; but this again adds additional
complexity
- examples: users do not switch from
- Word 5.1 ----> Word 6.0
- Word ----> Powerpoint
- Word ----> Canvas
- family of systems emerging to address this problem:
- Word ----> generate HTML code
- Pagespinner ----> Pagespinner Assistant
- Office'97 ----> user modelling to provide
contextualized help
Claims to be evaluated:
claim: "for most important mental tasks,
acquiring the knowledge needed a the time of is not feasible"
claim: "why learn on demand, if use on demand
is possible?"
Back to Top
Back to Top