Questions? Call 800-624-6242
|
|
|
|
|
The following HTML text is provided to enhance online
readability. Many aspects of typography translate only awkwardly to HTML.
Please use the page image
as the authoritative form to ensure accuracy.
Page 411
On Key Processes
Cross-Disciplinary, Social-Context
Research
John Leslie King
University of California, Irvine
A key challenge in developing an every-citizen interface to the
national information infrastructure is in recognizing the ongoing
evolution in our concepts of what the "interface" encompasses. The
development of concern over interface issues in the past 30 years
reflects the complexities of these issues and provides some
direction for improving these interfaces.
Background
The term interface in the computing field has been
appropriated by a relatively narrow community of interest, namely
those interested in human-machine interactions at the ergonomic and
perceptual level. Most of this study, which goes under the name of
human-computer interaction (HCI or CHI) is strictly limited to
studies of individual human actors interacting with specific
packages of hardware and software. This focus of work has been very
successful, producing among other things the innovations of
"pointing" aids such as the mouse, trackball, touchscreen, and
digital pad, as well as the graphical screen interface ubiquitous
in all modern operating systems. These advancements have their
intellectual
|
|
|
|
|
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 411
Page 411
On Key Processes
Cross-Disciplinary, Social-Context
Research
John Leslie King
University of California, Irvine
A key challenge in developing an every-citizen interface to the
national information infrastructure is in recognizing the ongoing
evolution in our concepts of what the "interface" encompasses. The
development of concern over interface issues in the past 30 years
reflects the complexities of these issues and provides some
direction for improving these interfaces.
Background
The term interface in the computing field has been
appropriated by a relatively narrow community of interest, namely
those interested in human-machine interactions at the ergonomic and
perceptual level. Most of this study, which goes under the name of
human-computer interaction (HCI or CHI) is strictly limited to
studies of individual human actors interacting with specific
packages of hardware and software. This focus of work has been very
successful, producing among other things the innovations of
"pointing" aids such as the mouse, trackball, touchscreen, and
digital pad, as well as the graphical screen interface ubiquitous
in all modern operating systems. These advancements have their
intellectual
OCR for page 412
Page 412
roots in fairly circumscribed zones of disciplinary concern,
namely cognitive psychology and human factors engineering. In
essence the focus of this line of work is on the human working with
a computer-based system.
In the past several years the HCI focus has broadened somewhat
to include small groups of individuals each working with a
computer-based system but for the purpose of working
collaboratively with other members of the group. This is a
significantly different conceptual focus, with individuals working
through the computer to interact with other individuals. Prominent
developments in this domain have been technologies for
computer-mediated communication, "groupware," and
computer-supported cooperative work. The intellectual roots of this
work go beyond cognitive psychology into other realms of the social
sciences, especially social psychology, but also into anthropology,
organizational psychology and sociology, and economics. The
applications of these technologies have caught the attention of
scholars interested in fundamental questions of human discourse,
social network construction and maintenance, identity and
personality formation and expression, and the social construction
of meaning and reality. These rapidly growing areas of interest
have been stimulated by the stunning speed with which major
components of the national information infrastructure such as the
Internet and the World Wide Web have invaded social life in all
dimensions.
These developments illustrate the evolving capacity of
computer-based systems to affect basic human activities and reflect
the fact that the concept of "interface" between humans and
information technology is an elastic concept that expands to deal
with the new opportunities and problems presented by technological
change. Three observations can be made from this evolving concern
with interface.
1.
The parochial concerns of any particular group
that engages interface issues at any given moment tend to
appropriate and dominate the evolving meaning of interface-related
research. The routine disciplinary politics of research
institutions affect researchers in the interface field. Interface
research was for many years (and to a considerable degree still is)
politically marginalized within the field of academic computer
science. Even within the interface field, some researchers whose
work is fundamentally grounded in psychology feel themselves to be
marginalized by those whose work is based on traditions of
engineering in which psychology plays little part. The lesson here
is that the dominant definitions of what constitutes the "real"
issues in interface research and what constitutes the "right"
approaches to doing such research are very misleading. It is
necessary to look beyond these politically constructed definitions
of what ought to be done and focus on the broader challenges of
what emerging applications will require.
OCR for page 413
Page 413
2.
The trend in conceptual evolution of interface
concerns is generally "upward" in the layers of social focus, from
the individual to small group, to organizational, sectoral,
institutional, and cultural. To the extent that being human
involves essential attributes of group identity, organization of
production and consumption, formation and sustainment of social
order and culture, and so on, these too require attention in
interface research. This notion is captured well in recent use of
the term usability to
describe research aimed at developing information technologies that
actually accomplish what those who use them desire. This focus has
clearly emerged in the computer-supported cooperative work (CSCW)
research community and has appeared as well in discussions of
organizational usability and even institutional usability of
information technologies. It is certain that a concern about
every-citizen interfaces to the national information infrastructure
must embody such perspectives. While it is true that at some level
all interface issues can be traced to rudimentary human-computer
interface concerns as represented by the parochial HCI community,
these broader issues of usability involve concerns that have
nothing at all to do with the narrow HCI focus and must be
addressed by research methods that traditional HCI researchers
would never consider.
3.
Although the concern with interface issues is
usually tied to the evolving HCI, CSCW, and other perspectives,
important aspects of research into group, organizational, and
institutional usability have been under way for many years.
Although largely ignored by the computer science research
community, the vast range of economically vital computing
applications in organizational information processing have drawn
much attention from researchers in management information systems,
library and information science, medical informatics, and other
fields. Transaction-processing systems, which remain among the
largest and most complex computerized information systems, were
made possible only by careful study and learning-by-doing design to
meet interface needs at the individual, work group, organizational,
and institutional levels. To pick just one case in point, designers
of the airline reservations system, which literally revolutionized
air travel, had to overcome numerous complicated problems at all
social levels, including being modified to comply with
court-ordered remedies against unfair competitive practices.
Similar stories can be told regarding credit data-reporting
systems, financial accounting and reporting systems, personnel
management systems, computer-integrated manufacturing systems, and
so on. The lesson here is that a great deal of useful information
on the development of effective interfaces at the higher social
levels is available in the applications-oriented research
communities.
OCR for page 414
Page 414
Research Needs
The point of the discussion above is not to argue against
further investment in the well-established traditions of research
in computer science, HCI, and so on. Such investments have yielded
great payoffs and will continue to do so in the future. Instead,
the point is to place in the foreground the need for focused
research on the "higher-level" issues of interface development at
the group organization, sectoral, institutional, social, and
cultural levels. Moreover, these are not merely desirable venues
for research investment, but rather are as essential to the goal of
an every-citizen interface as well-established domains of federally
supported research. Unfortunately, such research has been
comparatively underfunded. There are numerous advocates of more
traditional research needs to articulate requirements for such
research. I will focus on the needs of the higher-level
challenges.
The primary goal of research into the higher-level challenges of
interface design is to reduce the cost and increase the speed of
effective design. A great many extraordinary information
technologies have been developed that demonstrate the virtues of
good interfaces and that are highly usable in routine production.
But the cost of developing them has been quite high because most of
the systems have been built on the ruins of expensive failed
efforts. It has been estimated that as many as half of large
information systems projects fail to meet their objectives, and a
significant fraction of those fail altogether. Many examples can be
drawn from public-sector projects such as the disastrous World Wide
Military Command and Control System (more than $5 billion), the
Federal Aviation Administration's advanced technology program to
replace the aging air traffic control system (more than $2
billion), and California's write-off of its ambitious overhaul of
motor vehicle information systems (a mere $55 million). Similar
failures abound in the private sector, but they are more easily
hidden from view. The American Airlines effort to replicate its
marvelous success in airline reservations systems in the French
national railways and in its Encompass freight management system
come to mind as just two examples. Research into higher-level
challenges is aimed at learning what works and what does not and
putting that knowledge to work.
The following constitute important areas of needed research
investment at the higher level:
•
Synthetic studies that pull
together the extensive social learning already accumulated through
important development projects. Most
large-scale system development efforts occur in operational
settings, not research laboratories. Research on such systems must
be done in vivo, in the living systems.
OCR for page 415
Page 415
There are significant problems with
doing such research that need to be acknowledged. Among others,
such research usually presents researchers with difficulties in
gaining access to the right study sites, reluctance of
organizations to reveal their failures due to fear of
embarrassment, reluctance to reveal details of successes for
competitive reasons, and the high cost of travel and time on site.
But these are really no more troublesome than the challenges faced
by numerous other research communities that must go to the field
for their data collection. The synthetic research suggested here
must be theoretically driven, but ultimately it is empirical in
nature and akin to research traditions in engineering and
management. The questions are: What works? What doesn't work? And
why?
•
Analytical studies of the likely
evolutionary pathways of complex sociotechnical systems over
time. It is increasingly recognized that
information technologies do not merely allow efficiency and
effectiveness improvements in well-established activities. Much
more importantly, they enable fundamental changes in the nature of
the activities themselves. The whole "reengineering" movement,
which was articulated by people from organizational information
systems backgrounds, is predicated on the argument that fundamental
changes in what is to be done are required, displacing the
traditional focus on how things are to be done. Much of the
learning about what will "really" happen is inevitably going to be
empirical-we will watch and see. But advances in conceptual tools
such as game theory have made possible much more sophisticated
modeling of possible interactions among actors under different
assumptions, including enablement from new information
technologies. Such studies seldom provide real predictive power in
the sense that they can tell designers or decision makers exactly
what to do. But they have been applied with great success in
narrowing the search space around the likely outcomes under
different assumptions-a contribution that can greatly improve the
efficiency in design of complex systems that must be built through
learning by doing processes such as prototyping. Significant
increases in investment in such research are needed.
•
Collaborative research and
development projects that allow different development strategies to
be tested in real settings. There are
numerous recommended strategies for improving the design of complex
sociotechnical systems. Many of these have evolved from the
traditions of system design and software engineering and range from
structured analysis and design techniques to participatory design.
Unfortunately, our understanding of the efficacy of these
approaches amounts to little more than folklore. There have been
few systematic studies to demonstrate the utility of these
approaches in real development situations and the contingencies
under which the different approaches offer advantages.
Moreover,
OCR for page 416
Page 416
this lack of systematic study makes
it difficult to identify holes in existing knowledge and theory
that require further research attention.
These kinds of studies require multidisciplinary research
approaches, involving specialists from information and computer
sciences, management, and the social sciences. In many cases,
particularly where applications are in expert domains (e.g.,
medicine), subject matter specialists will be needed as well.
OCR for page 417
Page 417
Audio Access To The National
Information Infrastructure
John C. Thomas
NYNEX Science and Technology
Goals
There are trade-offs among the goals given below and the costs.
While each is a desirable direction, any of these goals, if pushed
to an extreme, could render other goals impossible.
•
Universality. Every citizen should have some way of gaining access to
information that has been made public. In some cases this may
require special devices, language help, training, or economic
assistance.
•
Privacy. Individuals should have the ability to create, store,
and modify information and restrict the access others have to that
information. Information created about individuals by others should
also be restricted in ways that the individual has some power over.
At a minimum, individuals should know who has what information
about them and should know when new information is being
collected.
•
Security. Information should be safe from unauthorized
destruction, alteration, or copying.
•
Usability. Information should be presented in a form that is
maximally useful. This depends on the person, the task, the
context, and their access to the system.
•
Empowerment. The individual should be free to determine how, when,
and where he or she access information. For example, some people
may prefer, for certain tasks, a very "active" system in which
agents make frequent suggestions. Others may prefer a more passive
system.
•
Responsibility. Economic or other incentives should be in place so that
limited resources (bandwidth, storage capacity, computer power,
creative human power that produces new information) are not simply
"taken" for free by whoever gets there first. There should also be
some incentive for those who create information to keep on
creating.
•
Translatability. Information entered in one medium should be capable of
being translated into another medium. Not only should documents be
easily translatable into another machine's format, but
faxes
OCR for page 418
Page 418
should be translatable into ASCII,
spoken words into ASCII, and, ultimately, even pictures and video
should be somewhat describable as text.
•
Ubiquity. People should have access from anywhere to information
(perhaps at a premium). They should not have to physically move
very far to gain this access.
Research Approach
In 13 years at IBM I saw many excellent results (in
human-computer interaction and effective software development
process) of research in our labs; other labs and universities have
a minuscule impact on what was actually done at IBM. In starting
the artificial intelligence laboratory at NYNEX, I led each group
to have a long-term vision of where to push the technology but to
choose a portfolio of short- and medium-term projects that pushed
toward that vision but also provided a real benefit for NYNEX. This
was not motivated primarily by a desire to stay funded-though that
was a consideration too-but because I believe it is far too easy to
''partial out" the really difficult issues of artificial
intelligence (AI) if one works on "toy systems." I do not believe
there are "frictionless planes" when it comes to human psychology
or building complex systems. Only through applied work is real
progress made.
The ultimate goal of speech synthesis, for example, might be to
read text as a good human actor might. However, a little reflection
will show this (and all other similar problems!) to be coextensive
with the "general AI" problem. While this is a worthy vision, a
mundane step we took in this direction was to work on a better
synthesizer for names and addresses, focusing on improving the
prosody. Here the approach of the main investigators (Kim Silverman
and Ashok Kalymanswamy) was not to "prove" that a particular
approach to prosody was the "right" one but to use everything that
worked and to do their own original research when they hit unsolved
problems.
None of this philosophy should be taken as meaning that there is
no place for theory or no place for university research. We have
had a number of good collaborations with universities, including
MIT and the University of Colorado, where we work on real problems
for a while and then spend time theorizing. But the theorizing is
based on experiences with real problems, and the theory is then
applied to the next real problem.
I also believe that working on complex, real-world problems
requires the cooperation, and perhaps the friendly competition, of
numerous research groups. The speech community has shown remarkable
progress working together through the Defense Advanced Research
Projects Agency in collecting and sharing data, trying various
approaches, and publishing results. The dawning commercial success
of this technology is a further
OCR for page 419
Page 419
driver toward progress. I believe, therefore, that the way to
develop the research agenda for issues about providing an
every-citizen interface to the national information infrastructure
is to start doing it. As we run into problems, we should use those
problems to define the research agenda.
Research Issues
I propose that we develop an audio access to the Internet. Such
a system would allow anyone access to the Internet over an ordinary
POTS telephone line by using spoken commands. Quite obviously, this
would not provide the identical experience that sitting in front of
a keyboard (for an experienced typist) and a huge full-color
monitor might. Nonetheless, such a system will have several major
practical benefits and would serve as a focal point for pushing
some important technologies.
First, the penetration of personal computers into the lowest
socioeconomic status quartile is low and staying low. This would
allow people without the financial resources immediate access to
the Internet. It would also allow people who are reluctant to buy a
computer because they don't see the value of Internet access (or
other computer applications) to "try out" Internet access without a
significant up-front investment. Individuals could use such access
to listen to speeches of political candidates on particular issues,
find out about their benefits and bills, find out about community
events and safety messages, find transportation schedules, get
sports results, and so on. Some of this information currently does
exist in various audio forms, but typically it is not updated very
frequently, and users must dial a different number for each type of
information.
Second, there are families and work groups where most of the
group has Internet access but a few do not. An audio access to the
Internet would allow such a group or family to communicate much
more effectively.
Third, even people who do have computer access to the Internet
find themselves in situations (e.g., in their cars, at payphones,
at hotels without modems) where they do not have computer access
but could use a phone to find important financial and trip
information, listen to their e-mail, etc.
Fourth, audio input/output is already becoming an important
enhancing medium for the Internet. Welsh lessons that include audio
are available from Brown University. One can listen to music,
people's voices, and sounds. Ideally, one might well navigate more
easily through voice commands. In addition, speaker verification
could add another level of security to Internet transactions.
Fifth, there are people with special needs (blind or paralyzed
users) for whom audio access to the Internet would be crucial. In
addition to these special needs, there is a huge population of
people in the United
OCR for page 420
Page 420
States who can speak and listen to English but cannot read or
write it. In a stroke, such an interface would help enfranchise and
educate them.
In addition to providing practical benefits, a workable audio
access to the Internet will force us to improve several important
technologies.
•
Speech
recognition. While some access to the
Internet could begin immediately, access could be improved via
larger, speaker-independent vocabulary. Another improvement could
be made if the system "adapted" on-line to the dialect and vocal
tract configuration of individual speakers. Performance in some
conditions will also be enhanced by better noise cancellation
techniques.
•
Speech synthesis. While current levels of intelligibility would be useful,
certainly further enhancements can and should be made in the areas
of aesthetic appeal, proper name and special symbol pronunciation,
and prosody. In some cases, the Internet may provide additional
clues for the synthesizer. For instance, in e-mail, the name and
subject fields give potential clues about how the contents should
be pronounced. Hypertext markup language (HTML) tags in Web sites
may give additional structural information about pauses and
emphasis.
•
Picture and video
understanding. To provide some (not
complete) audio information about pictures, graphics, and videos
available on the Internet, an automatic scene describer would
provide a cheaper solution than having human beings try to keep up
with the exploding information by describing each scene. By having
HTML tags and verbal materials to provide some structure and
context, the task for machine scene understanding could be made
more tractable, but there are still unsolved research issues here.
The solutions would have other applications as well, including
digital movie making and editing and security.
•
Natural language
processing. Much of the information on
the Internet is in the form of text. Being able to do a better job
of indexing, summarizing, and locating text would drive better
natural language processing. Again, the HTML tags of Web sites
provide a potential additional source for natural language
understanding systems.
•
User interface
design. How should a dialogue over the
phone be structured? Under what conditions should it be all speech
and when should DTMF also be used? When is explicit confirmation
needed? How can speech/audio be used as an adjunct to screen-based
systems?
•
System
integration. Perhaps a mobile phone user
would like to browse for certain kinds of information and bookmark
Web pages for later perusal on a screen. How can the various
networks be internetworked?
•
Speaker
verification. Perhaps the user could be
presented with a phrase to repeat in his or her own voice. This
could be used to help
OCR for page 421
Page 421
authenticate transactions. Clearly,
better verification techniques would help make the system more
secure and user friendly.
•
Media
translation. Speech recognition and
synthesis are special cases, but it would also be nice to be able
to translate a fax document into an e-mail document that could then
be read over the phone. In general, information should be capable
of being input in any medium and output in any other medium. While
the (typed) fax to e-mail issue is nearly solved, handwriting still
provides a challenge. We still do not see systems adapting to
handwriting as well as humans do.
•
Adaptive
recognition. The human being seems to be
very good at "speaker-independent" speech recognition. Yet a
selection made by "pasting together" words from a random speaker
seems very difficult for a human to understand compared to one from
any single speaker. Today's speech recognition systems don't seem
to care. Apparently, humans are doing something in the way of rapid
speaker adaptation that we do not understand well enough to
incorporate into a machine. Similarly, we are quite good at
"adapting" to a particular individual's handwriting. Again, we
don't know how to program a computer to do this very well.
Similarly, if we view a graph, photo, or movie on a particular
device in a particular lighting context, we make a fairly quick
adaptation to the style and other aspects of context. And, again,
we don't know how to do this very well in a machine. I believe a
key to significant improvement in a number of the technologies
listed here is a better understanding of how humans do adaptive
recognition. Rather than studying it in a toy domain, however, I
believe we should observe and test how people do this in a real
context doing a real task.
•
Intelligent
searching. Today's search engines on the
Internet are not very precise. They typically return very many
false-positives. Audio input/output with natural language
processing gives the possibility of more selective searches and
also provides strong motivation since audio scanning is more
onerous than visual scanning. One potential source of information
is to use the user's current task and past history to help focus a
search. Such an approach forces us to examine privacy and security
issues.
All of these technologies could be explored in their own right,
but I think that exploring them in the context of trying to provide
a real-world system will produce the best research results as well
as a practical benefit.
Appendixes
OCR for page 422
OCR for page 423
APPENDIXES
OCR for page 424
Representative terms from entire chapter:
national information
