CHI 97 Electronic Publications: Doctoral Consortium
Representation Without Taxation: What Makes GUI Good
Brian D. Ehret
George Mason University
Department of Psychology, m/s 3f5
Fairfax, VA 22030 USA
+1 703 993 4667
behret@gmu.edu
ABSTRACT
In the proposed work, research in cognitive science and display-based
HCI is synthesized and brought to bear on the question of "what
makes GUI good?". A two-phase approach is outlined. The empirical
phase will build upon a foundation laid by display-based HCI research.
The computational modeling phase will be informed by the empirical
phase and previous modeling efforts. The primary goal is to be
able to explicate conditions under which a user will rely on external
display components vs. internal knowledge structures to control
task performance.
Keywords
Display-based HCI, Cognitive modeling, ACT-R, Expertise, GUI.
© 1997 Copyright on this material is held by the authors.
INTRODUCTION
Since its introduction over a decade ago, the graphical user interface
(GUI) has become the standard means of human computer interaction.
This trend shows no indication of subsiding. If anything, it is
gaining ground as increasingly powerful processing capability
has allowed the incorporation of digital video, audio, and high
resolution graphics into interfaces. The latest of such advances
in GUI design have been driven by the popularity of the World
Wide Web.
Although GUIs have met with significant popularity and commercial
success, the obvious, and outstanding question remains, "what
makes GUI good?". Although there have been attempts to address
this question (e.g., [3], [9]), there remain important and interesting
ambiguities surrounding GUI-based interaction.
Casual observation reveals that users, even after thousands of
hours of experience, still rely in part on "tedious, hand-holding"
GUI features (e.g., menus, toolbars). Granted, some experts may
have taken the time to write macros or learn faster and more efficient
shortcut keys, but few if any of these users have ditched the
GUI altogether and reverted to older, command line versions of
the software. This begs the question: what are experts getting
out of this GUI deal that keeps them hooked?
Although research has not revealed what the experts are
getting, several interesting studies have demonstrated what they
are not getting. Mayes, Draper, McGregor & Oatley [7]
and Payne [8] have both shown that even skilled users of software
with several thousand hours of experience cannot accurately predict
the outcomes of simple command sequences or recall the names of
commonly used menu items. These experts have logged thousands
of hours using the software and yet they cannot recall properties
of commonly used features of the GUI. If experts can skillfully
use the software and prefer to stick with a GUI, then GUIs must
be a good thing. Given the aforementioned dissociation between
recall and performance, then just what is it that makes GUIs good?
My thesis work proposes to address this question. This work will
combine insights from diverse groups of researchers who have sought
to rethink traditional accounts of the control of complex cognition
(e.g., [5], [6], [10], [11]). In the course of this work, I plan
to build upon the empirical foundations of Mayes et al. [7] and
Payne [8], as well as the theory building and computational modeling
work in the area of display-based HCI (e.g., [2], [4]).
The result will be an attempt at a new synthesis that explicates
the interplay between representation and cognitive control in
display-based HCI. My goal is to do for procedural knowledge in
general, and HCI in particular, what Zhang & Norman [11] did
for our understanding of representations in distributed cognitive
tasks.
Internal vs. External representations
Zhang & Norman [11] argue that in order to understand cognition
in tasks where information is distributed across the internal
mind and external environment, one needs to decompose the task
into its associated internal and external representations. Further,
the traditional cognitive science approach to problem solving,
which emphasizes internal mechanisms in control of cognition (e.g.,
hierarchical goal structures, internal search), has systematically
underplayed the role of external representations by treating them
as mere memory aids. External representations are far richer than
this: they can provide information that can be directly perceived
and used without explicit formulation or interpretation, they
provide constraints that can anchor and structure cognitive behavior,
and they can change the nature of the task (e.g., make it less
taxing).
The importance of the external environment in the control of cognition
has also been addressed in the expertise and diagrammatic reasoning
literature (e.g., [5], [6]). This work suggests that expertise
in tasks that provide efficient, informationally rich external
representations is not necessarily manifested in complex internal
representations, but rather exists by virtue of reliance upon
these external representations.
External representations and GUI Expertise
In HCI, to the extent that the task and software support the ability
to rely on display features (external representations) for relevant
information, novice users should not need elaborate knowledge
(internal representations) of the interface in order to use its
basic features. Furthermore, these users should come to use more
advanced features without developing complex internal representations.
Given this description of novice and intermediate users, what
then is the nature of GUI expertise?
The display-based account, which is really an extension of the
above, is that experts continue to rely primarily on external
representations in task performance. This explanation suggests
that experts are content to rely on the external representations
provided by the interface and don't bother to develop internalized
procedural knowledge.
There is a performance cost associated with reliance upon external
vs. internal representations, however (e.g., shortcut keys are
faster than menus), so experts may internalize procedures in order
to improve performance. Thus, an alternative account suggests
that even though a GUI may support reliance on external representations,
experts choose instead to develop more efficient internal representations.
PROPOSED WORK
The overall approach of the proposed work is to empirically investigate
the potency of these accounts and to synthesize the results into
a computational cognitive model. Based on the empirical work and
the subsequent analytical modeling, my primary goal is to explicate
conditions under which a user will rely on external display components
vs. internal knowledge structures to control task performance.
The empirical phase will extend the studies by Mayes et al. [6]
and Payne [7] by evaluating user interface knowledge in the context
of task performance. In part, this research will investigate the
effects of degrading the informational quality of a GUI's external
representations on the performance of users at different skill
levels.
The computational modeling phase will be an attempt to synthesize
a control structure of cognition capable of reproducing the data
obtained in the empirical phase. Modeling will be done using Anderson's
ACT-R architecture [1] primarily because of the recent addition
of a visual attention mechanism that allows ACT-R models to search
in and interact with a display screen. Such a capability will
likely prove essential in modeling display-based interaction.
SUMMARY
The proposed research synthesizes contemporary research in cognitive
science and sets out to address the question of "what makes
GUI good?". To this end, conceptions of internal and external
representations will be mapped onto human computer interaction,
an empirical investigation will explore the potency of this mapping,
and the sufficiency of the resulting explanation will be determined
via analytical modeling and simulation. The results of this work
will advance the theoretical understanding of the nature of skilled
HCI, and could have significant implications for interface design.
ACKNOWLEDGMENTS
This research is supported in part by ONR grant N00014-95-1-0175.
I would like to thank Wayne D. Gray for his guidance on this project.
REFERENCES
- Anderson, J. R. (1993). Rules of the Mind. Hillsdale,
NJ: Erlbaum.
- Howes, A. & Payne, S. J. (1990). Display-based competence:
towards user models for menu-driven interfaces. International
Journal of Man-Machine Studies, 33, 637-655.
- Hutchins, E. L., Hollan, J. D. & Norman, D. A. (1986).
Direct manipulation interfaces. In D. A. Norman & S. W. Draper
(Eds.), User centered system design: New perspectives on human-computer
interaction. Hillsdale, NJ: Erlbaum.
- Kitajima, M. & Polson, P. G. (1995). A Comprehension based
model of correct and errors in skilled, display-based, human-computer
interaction. International Journal of Human-Computer Studies,
43, 65-99.
- Larkin, J. H. & Simon, H. A. (1987). Why a diagram is
(sometimes) worth a thousand words. Cognitive Science, 11,
65-99.
- Larkin, J. H. (1989). Display-based problem solving. In Complex
Information Processing: the Impact of Herbert A. Simon (Klahr,
D. & Kotovsky, K., Eds.). Hillsdale, NJ: Erlbaum.
- Mayes, J. T., Draper, S. W., McGregor, M. A. & Oatley,
K. (1988). Information flow in a user interface: the effect of
experience and context on the recall of MacWrite screens. In D.
M. Jones & R. Einder, (Eds.). People and Computers IV.
Cambridge: Cambridge University Press.
- Payne, S. J. (1991). Display-based action at the user-interface.
International Journal of Man-Machine Studies, 35, 275-289.
- Shneiderman, B. (1992). Designing the user interface: Strategies
for effective human-computer interaction (2nd ed.). Reading,
MA: Addison Wesley.
- Suchman, L. A. (1987). Plans and situated action: the problems
of human-machine communication. New York: Cambridge University
Press.
- Zhang, J. & Norman, D. A. (1994). Representations in distributed
cognitive tasks. Cognitive Science, 18, 87-122.
CHI 97 Electronic Publications: Doctoral Consortium
