Content uploaded by David D Woods
Author content
All content in this area was uploaded by David D Woods on Aug 03, 2017
Content may be subject to copyright.
PROCEEDINGS
of
the
HUMAN FACTORS SOCIETY
34th
ANNUAL MEETING-I990
NAVIGATING THROUGH LARGE DISPLAY NETWORKS IN DYNAMIC
CONTROL APPLICATIONS
David
D.
Woods
Cognitive Systems Engineering Laboratory
The Ohio State University
Columbus,
OH
Emilie
M.
Roth, William
F.
Stubler, Randall
J.
Mumaw
Westinghouse Science and Technology Center
Pittsburgh PA
ABSTRACT
There is an increasing trend to use computer display systems as the primary "window" by
which users see and interact with complex dynamic processes (e.g., air traffic control;
computerized control rooms for process control). These kinds of applications offer special
challenges to the design of computer based display systems.
In
particular, the large scope
of these applications necessitates large display structures involving thousands of displays.
Further, the dynamic nature
of
the tasks mean that users need
to
be able to move rapidly
through the display structure to keep pace with temporally evolving situations and to be
able to respond
to
new events as they occur.
As
a result; special display navigation
challenges arise
in
computer based display systems for monitoring and controlling dynamic
processes.
TRENDS
IN
INFORMATION TECHNOLOGY
There
are
a set of technological changes that are in motion
with regard to information technology in dynamic control
applications such as nuclear power control rooms, industrial
process control, surgical operating rooms (e.g., Cook,
Woods and Howie, 1990), process control systems
supporting space missions, and commercial aircraft
flightdecks. First, information technology is moving from
separated physically parallel devices to integrated computer
based systems. Second, the computer based display
technology has shifted from character-oriented graphics
systems to pixel-addressable systems and now to bit-
mapped, multi-windowed workstations. Third, we are
moving to greater intelligence in systems, especially the
combination of intelligent processing
of
data and powerful
display technologies. Fourth, we are moving from hybrid
systems, where hybrid means a mixture of hardwired and
computer based instrumentation (for example, the current
glass cockpits in commercial aviation are hybrid systems
-
Wiener, 1989), to completely computer based systems. This
is happening on a variety of scales and timing; for example,
Electricite de France has implemented the first fully computer
based nuclear power control room (Sun, 1990).
New Possibilities
The technological changes related to information systems
also lead us to new possibilities. Instead of having each
piece of plant data in one home (and in one form)
-
primary
system pressure is there, volume control tank level over
here, etc., instead of the operator, pilot, physician going to
the data, we can use the power of the computer to put each
piece of plant data in context that makes it meaningful for a
particular mode of system operation (Woods, in press). We
can bring the data they need to the practitioners and display
the data in different ways corresponding to the practitioners'
task.
Another trend is integrated smart displays. This type of
display puts together several or even tens of individual data
points into one integrated display that shows higher order
properties of the system, and it requires the use of either
heuristic or algorithmic processing power behind the surface
picture (for examples cf., Beltracchi, 1987; Woods et al.,
1987; Woods and Elias, 1988).
A
shift going on that may not be as apparent is that we
are
moving beyond the classic hierarchical organization of
computer based displays. The hierarchical display system
could work when we were dealing with a backfit system of
as many as
30
or
40
individual computer displays. But when
we
are
dealing with a fully computerized control center that
includes thousands of displays, hierarchical organization is
inadequate and does not provide the navigational tools
needed to deal with the huge space of data display that is
possible. Today people are beginning to use multi-
windowed display systems. The real question with window
managers is that they allow the designer to address questions
about how to coordinate different kinds of display frames
into a coordinated workspace (Henderson and Card, 1987;
Woods, in press). Guidelines for computer based display of
data, while extremely weak at the level of designing specific
graphic forms (Woods and Eastman, 1989), are virtually
non-existent for the design of a computer based workspace
rather than the physical workspace of a spatially distributed
hardwired control room (but cf., Woods, 1984).
Technological change is making the concept of a display
page wither away because of windows and because of
intelligent processing
so
that the display is customized to the
current data or system conditions. Furthermore, technology
is pushing into the area
of
automatic display
assembly/creation instead of laborious crafting of each
individual display (e.g., Mackinlay, 1986; Roth and Mattis,
1990). Obviously, this becomes animportant issue as the
scope of such systems expands to an entire control center.
396
PROCEEDINGS
of
the
HUMAN FACTORS
SOCIETY
34th
ANNUAL MEETING-I990
New Difficulties
As all of these changes are occurring, we need to ask
if
we
are better off. The shift to more computerization in control
centers, where the computer can do much more processing,
doesn't eliminate all of the hard problems in control center
design.
Look at the difference between a hardwired control room and
a fully computer based control room. In the hardwired case
much of the design work is directly visible in the layout of
controls, displays, status panels and annunciators in the
physically available space. But look closely at a picture of a
proposed fully computerized control room.You can
see
the
arrangement of the computer screens and workstations, but
the real design action and potential complexity is behind the
screens in the thousands of displays that an operator could
call up (for example, the Electricite de France computerized
control room has well over ten thousand computer displays,
and a new operating room integrated patient monitoring
system has well over 150 menu screens).
Before people had to navigate a large array of spatially
dedicated physically separated displays; now they will have
to navigate in a virtual space of thousands of computer
displays. Note how large display networks create new HCI
design challenges. The critical design bottleneck is shifted
from individual displays to the system of displays. Design
errors that can occur at the level of the interaction across
displays create new types of human performance problems
including getting lost in large display networks, tunnel
vision onto only a narrow subset of displays, display
thrashing, and new types of mental overhead related to
managing the display of data (e.g., Woods, 1984; Elm and
Woods, 1985; Henderson and Card, 1987; Cook et al.,
1990).
Our
experience with computer based display of data in
control centers comes from backfits of computers into
hardwired control centers, what we referred to earlier as
hybrid control centers. However, a system of thousands of
displays is not a simple evolution from a
30
display page
system organized in a simple hierarchy; it is a radical step-
change relative to information management. The challenges
in designing large display networks revolve around how to
avoid users getting lost in the large space
of
possibilities,
and how to avoid tunnel vision, or keyhole effects, where
users focus in on only a small portion
of
the display space
and are unaware of important changes in plant status that are
indicated in other parts of the display space where they are
not looking (Woods, in press). Large scale display systems
can place new mental burdens on the operator related to
information management (for specific studies showing this
see Moray, 1986; Cook et al., 1990). Given that one of the
problems in existing control centers is data overload in
rapidly changing circumstances, the shift to more computer
based systems can exacerbate this problem as well as
mitigate it (Wiener, 1989; Cook et al., 1990).
The use of designer aids that support rapid prototyping of
displays, as is necessary when one is developing large scale
display systems, can lead to a proliferation of displays
without adequate consideration of across display
organization and navigation issues. In part, this can occur
because, when it is technologically easy to create and add a
new display to the system, the solution to every problem can
end up being a new computer display (this is in part why
there is such a huge number of computer displays in the new
French nuclear control room). This is one case that illustrates
a potential for misuse of rapid prototyping: with rapid
prototyping techniques, you can make the same mistakes,
only more quickly or on a larger scale.
WORKSPACE DESIGN
At the workspace level, design activities are concerned with
grouping, organizing and coordinating forms to create a
workspace
-
the set of
viewports
and classes of
display
chunks
(content) that can be seen together in parallel or in
series. In the extreme only one viewport
is
available and
each display chunk takes up the entire viewport (the
historical default of a "display page"). The workspace can
include multiple windows and/or multiple
VDUs
as
viewports. The workspace includes the classes of display
chunks that are available and their inter-relationships.
Design of the workspace requires specification of how the
classes of display chunks are mapped into the available
viewports, i.e., a set of coordinated viewports/display
classes. Total flexibility, i.e., any display chunk can appear
in any viewport as the observer chooses, represents a failure
to design the workspace (e.g., Moray, 1986; Cook et al.,
1990).
A viewport is defined as any screen real estate that serves as
a unit where display chunks can appear. A viewport can be a
region within a single VDU
(
i.e., a window) or a whole
VDU screen. A set of viewports (Le., a workspace) can
consist of multiple window viewports, multiple VDU
viewports, and any combination of the two (Norman,
Weldon,
&
Shneiderman, 1986).
One of the critical forcing functions in the design of data
displays for the computer medium is that the set of
potentially observable display units or chunks is very much
larger than the available viewports (physical display area
or
real estate). This characteristic of computer based display
systems creates the danger of the
keyhole
effect where the
user is unable to maintain a broad overview, becomes
disoriented, fixated, or "lost" in the display structure
(Woods, 1984; Elm and Woods, 1985; Nielsen
&
Lyngboek, 1990). The main challenge in developing
computer based display systems is how to capitalize on the
computational power and display flexibility of computers
while still supporting the critical monitoring functions that
require the kind of rapid access to particular pieces of
information that traditional control center design supports at
least to some degree.
PARADIGMATIC COGNITIVE FUNCTIONS ELATED
TO
DISPLAY NAVIGATION
There are a number of paradigmatic user cognitive activities
that arise in dynamic process control applications for testing
an interface design with respect to navigation. How the
control center/workspace design supports or fails to support
these paradigmatic cognitive activities will determine the
degree of navigational trouble and whether any keyhole
effects occur. These constitute a set of "test cases" that can
provide a guide in designing or evaluating the viability of
any display navigation mechanism for dynamic control
applications.
397
PROCEEDINGS
of
the HUMAN FACTORS SOCIETY 34th ANNUAL MEETING1990
Consider the traditional control center. When you enter one,
even from the back of the room, among the first things that
you notice is that there are annunciator panels and/or large
screen displays ("big boards") that provide some indication
of overall system status
to
a knowledgeable observer
(therefore let
us
refer to the displays that support this as
'overview displays.' Stripping away the physical
implementation, there
are
several cognitive functions that
are
being supported by overview displays. To understand some
of these
it
is important to remember that control centers
are
almost always multi-agent settings (e.g.,
2
or
3
pilots on a
flightdeck plus autopilot functions; 15-20 people in the
mission control center at the Johnson Space Center during a
mission).
First, the overview displays provide
a common frame
of
reference
for multi-agent problem solving (Roth and Woods,
1989). The representation of the state of the process in
existing control centers is held in common by virtue of its
extension in a physical space. The multiple people in the
setting use this property of the system representation to
share data, to coordinate their activities across interacting
scopes of responsibility, to collaborate in solving problems
-
in other words, the representation provides support as an
informational medium for collaborative work (Le., the
current interest in computer-mediated collaborative work,
e.g., Stefik et al., 1987; Olson, 1989).
Second, overview displays support a rapid overall
assessment
of
system state, what we will refer to as an
orienting function. For example, when a new event changes
the system (such as a safety system activation in process
control), human operators quickly scan or walk the control
board to update, revise and evaluate their situation
assessment. During this scan the human practitioner checks
automatic responses, builds partial diagnosis about what
factors are at work in the system, and detects abnormalities.
Note that during this period the skilled practitioner is able to
detect abnormalities that heishe is not specifically alerted to
(data-driven search) or looking for (knowledge-driven
search). This can be termed incidental detection of
abnormalities because detection occurs in the course of doing
another activity. This occurs in traditional control centers
because in moving from one display to another in a physical
space one must scan the displays in between allowing for the
possibility of incidental detection of unexpected or abnormal
states. Also consider a common multi-agent aspect of fault
management in control centers in various domains. When
trouble occurs, very often new people enter the scene to
provide additional monitoring or control or diagnostic
resources. An effective representation should support the
integration of these newcomers by providing a mechanism
for them to size up the situation and integrate smoothly into
the team, without interference in the ongoing activities of the
original personnel on the scene.
A third cognitive function is the ability to rapidly shift views
to track a dynamically evolving event, what we call the
attentional control function. This addresses the need for
attentional control in a multi-signal, interleaved task situation
where new signals may interrupt ongoing lines
of
reasoning
and activity (e.g., Miyata and Norman, 1986; Woods, in
press). Consider the case where some trouble has occurred
in the system; the human practitioner has detected and
focused in on evaluating the source of the trouble and
possible corrective actions. However, there is a danger of
becoming too focused on that particular trouble spot and
failing to monitor the rest of the system for potential trouble
which may require a shift in attentional and action focus (for
example, the L-101 1 Everglades crash).
An
effective
representation should support the ability to step back from
the current detailed
focus
of attention and quickly size up the
entire situation in a mentally economical way that does not
disrupt the ongoing line
of
reasoning.
A fourth cognitive function is concerned with the problem
of
how does the observer decide where to look next
.
A bane of
today's information habitats is data overload. Technological
change tends to exacerbate rather than relieve the problem.
Representation design needs to support the domain
practitioner in filtering irrelevant data and focusing in on
relevant data given the current context. Many commentators
and much research on complex dynamic control domains
point out that data overload is a critical limiting factor on
human performance in traditional control centers, and that
the success of new technology is strongly tied to whether it
helps practitioners
cope
with overload, or exacerbates
overload (e.g., Woods, in press). Hence, this is a critical
issue to measure in defining the success of a system in
avoiding the keyhole effect.
CONCLUSIONS
It
is
very important to note that the cognitive functions in
monitoring complex dynamic systems identified above
are
not necessarily well supported by existing control center
designs. The support that does exist for these functions in
general was not a deliberate or conscious act of individual
design teams but a serendipitous property of the medium of
representation
-
physically parallel displays, and the result of
a historical process of human adaptation to the interacting
constraints imposed
by
the representation, and by the task
demands (e.g., for an historical analysis for one case, see
Cook
et
al., 1990).
However, the technological shift to a computer medium is a
double-edged sword. While it provides new representational
power for supporting cognitive work, it also undermines
some partially successful adaptations that have been worked
out for the previous medium. It provides the capability to
create much worse control centers (e.g., Elm and Woods,
1985), as well
as
much better ones (e.g., Stefik et al., 1987;
Olson, 1989) than our previous baseline. The research
challenge is to better understand the nature of aided human
information processing and computer-mediated work. This
will require integrating results and generalizing across
different specific domains and different specific
technological systems to identify a deeper underlying
structure of concepts.
ACKNOWLEDGMENTS
The ideas reported here have been stimulated in part by
research sponsored by NASA Ames Research Center to the
first author on human-automation interaction in aerospace
systems and in part by research sponsored by Westinghouse
Electric Corporation
to
the Human Sciences group of their
Science and Technology Center on the design of computer
based control centers. Portions of this work were presented
by the first author at a National Science Foundation panel
meeting on European Research and Development on Nuclear
398
PROCEEDINGS
of the
HUMAN FACTORS SOClElY
34th
ANNUAL MEETING-1990
Power Instrumentation, Controls, and Safety Technology,
December
1989.
REFERENCES
L.
Beltracchi. A direct manipulation interface for heat
engines based upon the Rankine cycle.
IEEE Systems,
Man, and Cybernetics,
SMC-
17:478-487, 1987.
history of introducing new information technology into
a dynamic high-risk environment. In
Proceedings
of
the Human Factors Society, 34th Annual Meeting,
1990.
R. I. Cook,
D.
D. Woods, and M.
B.
Howie. The natural
W.C. Elm and D.D. Woods. Getting lost: A case study in
interface design. In
Proceedings
of
the Human Factors
Society, 29th Annual Meeting,
1985.
A. Henderson and
S.
Card. Rooms: The use of multiple
virtual workspaces to reduce space contention in a
window-based graphical interface.
ACM Transactions
on Graphics,
5:211-243, 1987.
Support of Multiple Activities. In D.A. Norman and
S.W. Draper, editors,
User Centered System Design:
New Perspectives on Human-Computer Interaction,
Erlbaum, Hillsdale NJ,
1986.
Y. Miyata and D.A. Norman. Psychological Issues in
N. Moray. Modelling cognitive activities: Human limitations
in relation to computer aids. In E. Hollnagel,
G.
Mancini, and D.D. Woods, editors,
Intelligent
Decision Support,
Springer-Verlag, New York,
1986.
J. Nielsen
&
U. Lyngboek Two Field Studies
of
Hypermedia Usability. In C. Green
&
R. McALeese
(Eds.):
Hypertext: Theory Into Proactice
II,
INTELLECT,
1990.
Cognitive Layouts of Windows and Multiple Screens
for User Interfaces.
International Journal
of
Man-
Machine Studies,
1986, 25,229-248.
K.
L.
Norman,
L.
J.
Weldon and B. Shneiderman
M.H. Olson (Ed.)
Technological
Support
for Work Group
Collaboration.
Erlbaum, Hillsdale NJ,
1989.
E.M. Roth and D.D. Woods. Cognitive Task Analysis: An
approach to knowledge acquisition for intelligent
system design. In
G.
Guida and C. Tasso, editors,
Topics in Expert System Design,
North-Holland, New
York,
1989.
M. Stefik,
G.
Foster, D. Bobrow,
K.
Kahn,
S.
Lanning
and
L.
Suchman.Beyond the chalkboard: Computer
support for collaboration and problem solving in
meetings.
Communications of the ACM,
30:32-47,
1987.
W.
Sun.
A Summary Status Report of French Nuclear
Power Plants Information and Control Systems.
Assessment
of
European Nuclear Instrumentation,
Control and Safety Technology.
J.
Sackett and
P.
Planchon (eds.), National Science Foundation,
1990.
E.L. Wiener.
Human Factors
of
Advanced Technology
(”
Glass Cockpit”)
Transport Aircraft. Technical
Report1
17528,
NASA,
1989.
D.D. Woods.Visua1 Momentum: A concept to improve the
cognitive coupling of person and computer.
International Journal
of
Man-Machine Studies,
2 1 ~229-
244, 1984.
D.D. Woods. The Cognitive Engineering of Problem
Representations. In G.R.S. Weir and J.L. Alty,
editors,
Human-Computer Interaction and Complex
Systems,
Academic Press, London, in press.
integral display concept. In
Proceedings
of
the Human
Factors Society,
32nd Annual Meeting,
1988.
D.D. Woods and M.C. Eastman. Integrating principles for
human-computer interaction into the design process.
In
IEEE International Conference on Systems, Man, and
Cybernetics,
IEEE,
1989.
D.D. Woods and G. Elias. Significance Messages: An
D.D. Woods, J. OBrien, and L.F. Hanes. Human factors
challenges in process control: The case of nuclear
power plants. In
G.
Salvendy, editor,
Handbook
of
Human FactorslErgonomics,
Wiley, New York,
1987,
1724- 1770.
399
