The Design of an Information-Exploration Interface
Michelle Q Wang Baldonado
Gates Building 3B
Computer Science Department
Stanford University
Stanford, CA 94305
(415) 723-7784
michelle@cs.stanford.edu
Digital library interface designers agree that information-seeking
tasks can have divergent characteristics. Three tasks illustrate some
of these differences. In Task 1, a user hopes to find the most recent
book written by Italo Calvino. He will consider his task successfully
completed once he locates this exact book. In Task 2, a user is
interested in having fun by surfing through "cool" pages on the World
Wide Web. Her criterion for success is the quality of the experience.
Finally, in Task 3, a user needs to write a term paper for a class on
Greek art. He will consider his task successfully completed if he can
find a good topic and a collection of resources on that topic.
SenseMaker is a user interface designed to support users who are
engaged in tasks like Task 3. These tasks are information-exploration
tasks in which users look for new information within defined
conceptual areas. The design of SenseMaker builds upon the following
assumptions:
- Accumulating a collection of references helps users to
accomplish the end goal of discovering new information.
- Users often need to consult multiple, heterogeneous sources. For
example, the Greek art student might consult indices of museum
holdings as well as multiple book and article indices.
- Users can benefit from looking at collections of information from
different perspectives and at different levels of granularity.
- Users' interests evolve contextually during the task.
SenseMaker's goal is to help users contextually evolve the direction
of their search processes. No interface can encode all possible aspects
of a user's context. SenseMaker works by approximating the current
context as the current collection of accumulated information
references. We believe that this approximation facilitates:
- Examining the current context.
In SenseMaker, users examine
a context by experimenting with different views of the accumulated
information references. Within a view, "similar" references are
bundled together, while "identical" references are merged together.
Many criteria for "similarity" and "identity" are possible in
SenseMaker, and users may move fluidly from one to another to gain
different perspectives.
- Progressing from one context to the next.
In SenseMaker, users
progress from one context to the next by building upon, taking away
from, or replacing the current collection of information references.
They can also return to previous contexts and continue exploring
from there.
References
SenseMaker is the subject of my PhD dissertation.
My advisor is Professor Terry
Winograd in the Stanford University Computer Science Department.
SenseMaker is one of many ongoing projects in the Stanford Digital Library
Project. For more information on SenseMaker, please see:
Goals for Allerton 1996
I am looking forward to hearing the different points of view that will
be represented at Allerton 1996. My work has already been influenced
by the research of several Allerton participants. I am excited by the
opportunity for sharing ideas and discussion at Allerton 1996.
In particular, I hope to:
- Discuss the current SenseMaker design, implementation, and evaluation with interested participants.
- Learn about related projects.
- Find more pointers to previous research and data on how users traditionally do information-exploration
and make sense of search results.
- Gain some insight into how to continue my evaluation of SenseMaker.