SDH 1992x

pastecoolΤεχνίτη Νοημοσύνη και Ρομποτική

14 Νοε 2013 (πριν από 3 χρόνια και 10 μήνες)

79 εμφανίσεις

P
ARADIGMS OF
GIS

U
SE



Werner Kuhn

Dept. of Geoinformation

Technical University of Vienna

Gusshausstrasse 27
-
29

A
-
1040 Vienna, Austria

wkuhn@email.tuwien.ac.at




Abstract


The conceptualization of the use of geographic information systems
(GIS) is being
considered here as a question of metaphor selection.
The paper claims that understanding the use of a system as such,
independently of specific tasks, is always done in terms of some
familiar domain of experience and is consequently metaphorical. In
order
to distinguish this understanding from task
-
specific metaphors,
the term "paradigm" is used. A paradigm can be conscious or can
implicitly underlie the design and use of a GIS, but it always has
profound psychological, economical and organizational consequ
ences.
Some relevant paradigms are analyzed and their current or potential
role for GIS is discussed.



Introduction


The way we conceptualize the use of a GIS has a fundamental impact on the usability of
the system. This is true whether we are GIS designe
rs, GIS users, or decision makers
relying on other people's use of GIS. Designers can empower future users with
unprecedented possibilities for spatial problem solving, or they can preclude these,
depending on how they conceive GIS interaction. Users of a
given system can realize the
potential it offers, or they can fail to do so due to inappropriate expectations. Decision
makers, finally, can transcend existing organizational constraints by their vision of what
it means to use a GIS, or they can forego the

benefits of new technology by
misconceptions about its use. How, then, does the conceptualization of system use enter
the design of user interfaces?


The conceptual design of user interfaces is increasingly being recognized as a matter
of metaphor selecti
on [Foley et al. 1990]. As the first phase of a user interface design
process, the conceptual design consists of the definition of objects (including their
properties, relations, and operations) which a user will have to know in order to use the
system. Fo
r example, the conceptual design for the user interface of an operating system
could consist of a desk environment with documents, folders, a trash can etc. and
operations such as to move or delete documents. Clearly, these objects are used as
analogies to

structure the domain of an operating system. Their selection establishes a
mapping from a domain of experience familiar to the user (here, an office environment)
to the unfamiliar domain of abstract computer operations (here, those of an operating
system)
. Such a mapping is precisely the function of a metaphor.


The recognition of the conceptual design phase as a conscious act of metaphor
selection is more than the use of new terminology for old ideas. It allows designers to
benefit from work in cognitive
science over the last decade which has revolutionized the
way metaphors are understood [Lakoff and Johnson 1980]. In particular, it provides a
view of interaction design as an activity which goes far beyond the use of fancy graphics
and WIMPs (windows, ico
ns, menus, pointing devices). Rather, interaction design
becomes a process of establishing the concepts with which the users will have to deal in
order to master the system. The surface issues of how to structure menus or how to shade
icons become secondar
y at best.


The novelty of the approach proposed in this paper lies in breaking down the task of
metaphor selection into more manageable and understandable parts. In particular, the
paper distinguishes two levels of metaphor selection. While the commonly r
ecognized
level at which metaphors occur in user interfaces is that of task
-
related objects and
operations, it is suggested here that metaphors enter the game at another, more
fundamental level of design. This is where it is decided how system use as such
is to be
understood, what roles are played by users and system and what kind of relationship they
are to have.


In order to distinguish the generic metaphors which conceptualize system use from
task
-
related metaphors, the term "paradigm" shall be used
here for the former. This term
has been used in human
-
computer interaction literature with various imprecise meanings.
Its most frequent use (e.g., "the direct manipulation paradigm") comes closest to this
generic kind of metaphor for system use across app
lication domains. There is also
sufficient overlap with the notion of a scientific paradigm [Kuhn 1962], in the sense that
a paradigm dominates, consciously or unconsciously, thinking and acting in a certain
domain. Maybe the definition of interface paradi
gms in this sense can help to reduce the
terminological confusion existing about the difference between metaphors and paradigms
for user interfaces.


The choice of task metaphors depends on the framework for paradigms, but will not
be dealt with systemati
cally in this paper. It appears likely that selecting metaphors for
specific tasks or domains will become both easier and more effective once paradigms are
better understood. The current uncertainty about novel ideas for GIS interface metaphors
may indeed
be caused by the complexity of the metaphor selection problem if the two
levels are not distinguished.


After briefly tracing the evolution of GIS user interfaces in section two, the paper
lists some pertinent examples of paradigms in section three, discus
ses some components
of these paradigms in section four, and draws conclusions about the relevance of
paradigms in section five.



The Evolution of GIS User Interfaces


In order to provide some background to the subsequent discussion of paradigms, a brief
r
eview of the evolution of user interfaces, particularly of those for GIS, is given here.
Generalizing to a high degree, a path can be identified that leads from black boxes to
"glass boxes" to interfaces which allow the user to "break the glass" [Furness 1
991].


Black Box Systems


Interactive computer programs have traditionally been conceived as command
-
driven
black boxes. A user typed a command, the computer processed it (if it was among the
admissible commands) and provided a limited amount of typed fee
dback. Not only were
the internal workings to execute a command hidden from the user, but usually also the
fact that the machine was doing some work and sometimes the success or failure of the
operation (e.g., in programs adhering to the UNIX philosophy, w
here no news from the
system is good news).


The idea of a black box is nothing else than that of an abstraction, thus, hiding
information which is considered irrelevant. Hiding system internals from the user is, of
course, central to system design. The n
otion of a black box system, however, implies that
the wall which is erected between the system and its users is fixed. Also, the decision
where to build the wall was often reached arbitrarily rather than based on an analysis of
the user's actual informati
on needs. Most black boxes are leaking and require users at
some point to know something about their internals in order to use them appropriately. In
other words, the user is not provided with a consciously designed model of the system's
behavior.


Many G
IS in practice still exhibit this interface style to a large extent, with a set of
commands rigidly but often quite arbitrarily separating the things users are allowed to see
and manipulate from those which remain hidden.


Glass Box Systems


With the adven
t of graphical user interfaces and direct manipulation, it became possible
to replace black boxes with glass boxes. A glass box presents to the user a coherent
model of a system's behavior [DuBoulay, O'Shea, and Monk 1981], creating the illusion
that the u
ser actually sees, through "windows", everything that's happening. WYSIWYG
(what you see is what you get) systems are today's standard examples for glass boxes.
Using object
-
oriented techniques, objects on the screen and objects of an internal model
can be

made to behave as one and the same. Thereby, changes in the state of the system
become readily apparent to the user.


Of course, what is presented to the user is still an abstraction of what the system
"knows"
-

more so than ever. But it is consciously d
esigned to provide a coherent mental
model of the system's workings. To capture the whole system functionality, this requires
generally a combination of several metaphors. In fact, the term glass box can be defined
as a coherent system of metaphors fully r
epresenting the relevant system functionality
[Carroll and Olson 1988]. The significance of "glass" is twofold: it permits to see what
needs to be seen (or what the designer believes the user needs to see), but it still erects a
barrier between the model a
nd the user, limiting interaction mainly to the visual channel.


User interfaces for GIS are only now reaching the glass box stage. In fact, the market
leading systems fall in between the black box and glass box stages, using WIMPs to
visualize essentially

a black box command structure.


Breaking the Glass


The next step in advancing the state of the art in user interfaces is to "break the glass" of
the box. Virtual reality [Brooks 1988] achieves this by transforming users from outside
observers to immersed

participants. Through realistic and reactive simulations of
geographic models with multi
-
sensory feedback, a person "vicariously" interacts with the
world [Mark 1989] rather than observing a more or less static, boxed model of it through
narrow windows.


The significance to GIS of recent developments in virtual reality appears to be very
high. Some existing virtual reality systems or prototypes outside the GIS field represent
in some sense the most advanced "GIS" user interfaces. A reason for this may be t
hat
modern interface styles, like direct manipulation and virtual reality, are characterized by
spatialization: They rely on human spatial abilities to facilitate interaction with problem
domains which are often not spatial per se, for example, the organiz
ation of data files in
an operating system. Basic human activities in space (e.g., to see, point, pick, manipulate,
and move) have become the primitives of modern, general
-
purpose interaction languages.
This situation represents the unique chance of gettin
g an important part of the technology
for geographic problem solving for free. However, GIS user interfaces will have to
evolve quite rapidly, if a medical surgery support system, for instance, is not soon to
exhibit stronger spatial problem solving capabi
lities than an average GIS.



A Catalogue of Paradigms


The following paradigms have all explicitly or implicitly been applied to conceptualize
the use of computer systems in general and of GIS in particular. The list is certainly
incomplete, but demonstra
tes some trends in our understanding of what it means to use a
system.


To use a system is to ...

... program


This notion represents the traditional understanding of human
-
computer interaction. The
system is seen as an (abstract) machine, a black box, pro
grammed and operated on by the
user. The notion originated in the early days when electronic data processing was
performed by specialists. However, it still permeates systems, design methods, and
terminology ("commands", "macros", "errors" etc.) throughout

most applications of
information technology today. The user interfaces of widely marketed commercial
systems have been designed by people whose job and devotion is to program. Since these
people are generally also the first users of their own systems, the
y naturally come to
expect similar inclinations from other users.


Underlying the paradigm that to use is to program is the philosophy that computer
systems have their own, often arcane languages in which they need to be addressed and
which users conseque
ntly need to learn. Moreover, the only mode in which a system can
be "talked to" is that of a formal, predetermined series of instructions. There are clear
advantages for system maintenance and security (not every "fool" gets access to the
sophisticated sy
stem). From today's perspective, however, such a conception severely
limits the bandwidth of interaction, particularly for domains like spatial information, as
will become clear when alternative conceptions are discussed.


... manipulate


This notion becam
e popular with the direct manipulation style of interaction. Users are
offered tools or a whole tool box. In order to solve a task with the system, they have to
select appropriate tools and to know how to use them. Application tasks are executed
metaphoric
ally in terms of manipulations of these tools. The key ingredient to successful
direct manipulation interfaces are operations and corresponding tools which are easy to
visualize, understand and use. They also should rapidly complete and be easily reversibl
e
[Shneiderman 1983]. This requires a careful selection of appropriate task metaphors,
which might for example be a typewriter metaphor for text processing or a map metaphor
for spatial analysis in GIS.


The paradigm of direct manipulation offers powerful

feedback and a much stronger
sense of control than black box programming. However, it implies a largely passive and
"dumb" system. Tools, by nature, can be complex to use and generally require training.
Not all users can be expected to be "craftsmen" know
ing how to make best use of the
tools. Help systems can be supplied by designers, but are not inherent to the paradigm (a
hammer doesn't tell you how to use it). The art of direct manipulation interface design is,
therefore, to design tools such that they
provide "affordances" [Norman 1988], i.e., that
their design makes it obvious how to use them. This seems easier to achieve for very
familiar tasks of limited complexity (such as general office manipulations) than for ill
-
structured domains like spatial da
ta handling.


The success of direct manipulation office systems has led some developers to expect
miracles from the use of tool boxes wrapped in WIMPs. However, after ten years of
experience with direct manipulation interfaces of all kinds, it remains
unclear, how well
the tool box concept scales up to complex application domains like GIS. The future role
of direct manipulation seems to lie in a combination with other paradigms discussed
below.


... communicate


This conception of system use marks a
radical departure from the previous ones,
emphasizing the conversational rather than the computational aspects of interaction. The
user is seen as a partner in a dialogue or as a participant in a group communication
process. The system is either a partner
or a communication medium or might even take
on both roles.


The idea to model human
-
computer interaction after human
-
human communication
is older than that of direct manipulation [Nickerson 1977], but it has only gained
momentum and received appropriate
technological support (e.g., speech recognition) in
the past few years [Luff, Gilbert, and Frohlich 1990]. The conversational paradigm
provides important notions like context, intentions, beliefs, or cooperation (see below),
which are lacking from traditio
nal interaction styles. The danger which has to be avoided
is that of anthropomorphism, where the user is led to believe that the system has (almost)
human capabilities and intelligence. With the advent of non
-
human communication
partners, however, the aut
omatic association of communication with human intelligence
will probably decrease.


In practice, there are still only few systems exhibiting a conversational interaction
paradigm. The only GIS application domain currently exploring it is that of car
navig
ation systems, where the system can be considered an intelligent assistant to the
driver, to whom it communicates driving directions [White 1991].



... delegate


Delegation represents a special case of communication, where the system takes on the
role of
a subordinate agent or assistant. This has the advantage of establishing a
restricted, fairly simple, and familiar communication protocol. Delegation is often seen as
an antithesis to direct manipulation and gets naturally associated with speech
-
based
inte
raction [Negroponte 1989]. This emphasis on spoken communication is to some
extent a compensation for the previous negligence of this channel in user interfaces,
where the visual channel has been used extensively and almost exclusively. Delegation,
as oppo
sed to general communication, is also particularly suitable for spoken language
and allows for a richer language on the user side than on the system side which suits
current technology. It seems quite clear, however, that successful conversational
interfac
es will require appropriate combinations of multiple channels.


... query


Inherited from the database community, this paradigm offers a computationally simple
and powerful, but narrow notion of interaction. The system is conceived of as a database
to whic
h the user poses queries. Given that the primary function of an information system
is to provide answers to questions, a query language can cover at least a major part of
interaction needs. From a conversational perspective, on the other hand, being able o
nly
to ask questions is too limiting and it seems unnatural to use different languages for
questions and for assertions.


A practical problem with query languages is that they are generally based on a very
limited notion of what goes on in a question and
answer dialogue. For example, questions
are normally disconnected from each other and one can't use the previously established
context to ask further queries [Egenhofer 1989].


For some GIS applications, querying is much more important than, say, editing
data.
Therefore, interaction with GIS has often been considered as consisting essentially of
queries and successful user interfaces have been constructed based on this paradigm.
Nevertheless, severe limitations become apparent in standard database query la
nguages
when they are applied to GIS tasks [Egenhofer and Kuhn 1991].


... browse


Browsing is an attempt to break the limitations of querying, particularly for cases where
the users can't describe what they are looking for; a situation which is very commo
n in
GIS use, for example in remote sensing applications. The paradigm originates in
information retrieval [Salton and McGill 1983] and has great potential for applications
where a vast amount of information needs to be searched with only imprecise
specifi
cations of what to find.


Currently, very little is known about useful browsing methods. The major difficulties
are to find suitable abstractions to present to the user and to let the system make
appropriate guesses of what the user needs to see. Humans s
eem to be very effective at
browsing their living environment for things or events which might be relevant (a person
walking by, an item in a shop etc.). It is yet unclear, how this capability can be applied
for browsing in information systems. A crucial f
actor may be the combination of multiple
sensory channels which is hardly possible in practice today.


... skim


Skimming is going one step further than browsing, to cover cases where users don't even
know whether there is anything of interest to them. The

system needs a method to filter
information through various levels, so that the user can skim it like a newspaper,
preferably even a personalized version of a paper, oriented toward the user's particular
interests. This requires a model of the users to be

maintained by the system. A successful
skimming interface based on a news reporter metaphor has been implemented for a stock
market information system [Erickson and Salomon 1991].


As for GIS, again, remote sensing applications have the strongest need fo
r
skimming. Continuous or periodical monitoring tasks could benefit from better ways to
find out whether anything relevant to the task has occurred. This would not only improve
human
-
computer interaction, but has the potential to reduce the enormous storag
e
requirements of these applications. No doubt, the general trend of data quantities
increasing faster than the capacity to manage them is pushing user interfaces from rigid
query mechanisms toward more intelligent browsing and skimming facilities.


... pr
oduce and receive documents


An outcome from office automation, this is the predominant and most successful
paradigm for system use "by the masses" today (cf. the desktop metaphor and its
variants). It conceptualizes the user interface as a conventional wo
rking environment
(e.g., a desk) with a collection of documents (letters, tables etc.) and work tools (text
processors, printers etc.). It has become so predominant in all sorts of applications that it
has to be considered a paradigm by now, rather than a
task metaphor for office tasks.


Using GIS is widely regarded as producing and interpreting maps and related
documents. This has motivated attempts to extend the desktop metaphor with mapping
concepts [Frank 1991]. There are, however, clear limitations of

a static document view of
spatial information and alternatives are being studied [Kuhn 1991]. Also, accepting a
document conception doesn't necessarily mean that mapping documents are best
manipulated in a desktop environment [Wilson 1990].


... solve pro
blems


The idea that using a system means essentially problem solving motivates the view that
interaction should really occur with a problem domain rather than a machine [Fischer and
Lemke 1988, Mark and Gould 1991]. The system may recede into the role of
a medium,
facilitating the "direct" interaction of the user with the domain, e.g. with a model of the
landscape or of a hydrological network. This way, the user becomes more directly
engaged and the barrier of mastering system commands becomes less visible
.


The problem solving paradigm, however, applies to traditional tool box conceptions
as well, even to any kind of formal notation and interaction language. Thus, the crucial
requirement with this conception is to take into account the vast body of work w
hich has
been done on the significance of choosing appropriate problem solving languages
[Newell and Simon 1972, Polya 1945]: Having the right notation available is the
prerequisite for an elegant and effective solution. Again, this highlights the importan
ce of
choosing appropriate task metaphors. In the GIS field, studies of problem solving
activities and languages have been disappointingly rare so far [Kuhn 1990]. In
discussions of GIS user interface requirements, particularly with people from other field
s,
it is often frustrating to realize how little we know what problems we actually want to
solve with these systems.


... play


Metaphors explaining system use in terms of playing attempt to relax the austere and
often intimidating connotations of computer

use. They emphasize creativity, pose a
productive challenge to users, and, most importantly, encourage users to try out things
without fearing fatal consequences. On the other hand, playing conflicts with work ethics
in many cultures and professions. Appl
e's line of Macintosh computers provides an
interesting illustration of the ambiguity of playing paradigms. The major critique against
the Mac was initially that it was a toy rather than a computer. While this perception
rooted in the limited software and
hardware capabilities, it also had a positive aspect for
those who, for the first time, felt they could use a computer without constant frustration
and intimidation.


At about the same time, designers of user interfaces in a variety of application areas
b
egan to realize that they could learn a lot from the success of video games. Since most
common games (including those on computers, for example the famous "Adventure")
have a strong spatial component, it seems tempting to further explore this kind of
parad
igm for GIS applications. In some sense, a GIS is anyway something like a toy
world
-

a model of reality simplified to the point where users can play with it.


Playing paradigms could be most appropriate for tasks which involve the discovery
of patterns a
nd trends, rather than for tasks with routine solution procedures, such as
those in planning or engineering. Thus, a combination with the browsing idea appears
promising. The playing notion, however, also occurs in power games, which lie at the
heart of so

many spatial decision processes. Thus, applications like land
-
use or traffic
planning games come to mind. They lead immediately to another important paradigm:


... cooperate


Both human
-
computer and human
-
human cooperation can be mediated through a
comput
er system. Computer
-
Supported Cooperative Work

(CSCW) is rapidly becoming
a strong paradigm for interaction [Greif 1988]. It offers the potential to collapse space
and time by holding asynchronous meetings with worldwide participation. And it
provides the
potential for access on the spot to practically unlimited information
resources.


CSCW is likely to become particularly relevant for GIS applications, due to their
interdisciplinary nature. Most decision processes where GIS are or could be used (for
example, the siting of a landfill) involve a team of various specialists rather than an
individual only. Computer support for such collaborative efforts is highly desirable, but
hardly available so far or only at the level of general clerical functions. Th
e group
dynamics of such situations, however, require a careful design and introduction of this
empowering technology, so that useful imperfections don't get eliminated. For example, a
complete trace of everything that happened during a meeting may not be
to everybody's
liking and may even inhibit decision processes. Nevertheless, the cooperation paradigm is
very likely to reach the status in computing during the 1990s that the manipulation
paradigm had in the 1980s and GIS could become one of the foremost
application areas.


... see


This conception of system use is an instance of the metaphor "understanding is seeing"
which is fundamental to human intelligent behavior [Lakoff and Johnson 1980]. The role
of visual metaphors for abstract thought is well esta
blished [Danesi 1990], but only
beginning to be exploited for human
-
computer interaction. Of course, the current
enthusiasm for visualization techniques in computing relies on it.


Deciding and acting in spatial situations is heavily guided by the visual c
hannel.
Extending visibility metaphorically beyond physically visible entities is a traditional
cartographic technique. Superimposing abstract entities like zones, soil quality, or
ownership rights on the visible aspects of landscape is just one applicatio
n of the "using
is seeing" paradigm in GIS.


... view


This variation on "using is seeing" emphasizes the active role of the user who not only
gets to see something, but actively views things. WYSIWYG becomes much more
powerful if users can control in non
-
trivial ways what they get to see, e.g., by taking
different points of view, seeing things from different angles and distances and focusing
on other parts of a scene. Special cases of this paradigm are the standard graphical
interface metaphors of zooming

and panning, although their implementations are often a
far cry away from the power of human vision [Jackson 1990, Kuhn 1991].


We are only beginning to realize how sophisticated the filtering, focusing and
abstraction mechanisms of the human visual syste
m are and what we could learn from
them for improving user interfaces. The recent Xerox PARC development of the
information visualizer [Card, Robertson, and Mackinlay 1991] represents an impressive
extension of the direct manipulation paradigm in this dire
ction.


Since spatial information is closely linked to the visual channel, GIS could benefit
greatly from an improved understanding of these mechanisms and appropriate interactive
simulations of them.


... experience


Instead of singling out additional sen
sory channels ("using is hearing", "using is
smelling" etc.) one might consider this generalization of perception
-
based paradigms,
motivated by the metaphor "understanding is being in the world" [Johnson 1987]. It lies
at the heart of virtual reality (see
above), where the user interface essentially disappears,
and using a system means experiencing a simulated reality over multiple sensory
channels. Vision is still dominating, as it is for many tasks and most people in the real
world, but it gets supplement
ed by sound and, increasingly, tactile signals (force
feedback). Also, the angle of vision becomes much wider than that in traditional visual
interfaces, which is one of the key improvements occuring when "the glass of a glass box
is broken".


For GIS use,

this idea offers a radically new paradigm: the GIS is the terrain which
the user experiences directly. While there have been some experimental designs for using
additional sensory channels (e.g., tactile or smelling maps), the most promising direction
is
that of improving the interactive use of the audio channel. However, GIS applications
in this direction have essentially been limited so far to the support of visually impaired
people (where, of course, the benefit is most immediate).


Experiencing space i
s almost impossible without motion. The importance of motion
for cognitive development, learning, and information gathering cannot be overestimated
[Sacks 1990]. Thus, adding metaphorical motion to broad band sensory input seems an
imperative for GIS. Virt
ual reality's major thrill is indeed the possibility of the user to
move through space, especially at speeds and along paths which are physically
impossible.


The integration of metaphorical perception, action, communication and motion
pushes human computer interaction to its limits. Current technology can only support it in
rudimentary ways so far. Its appeal for GIS applications, however, is very strong, becau
se
GIS and virtual reality have a common interest: To offer simulations of reality which are
both less and more constrained than reality itself.



Components of Paradigms


The catalogue of paradigms given in the last section reveals a need and potential
for
structuring. Obviously, some of the paradigms are related. For example, some paradigms
are special cases of others and some share the conceptualization of one part of the
interaction process (e.g., the system or the user) with others. Thus, the paradig
ms can be
further analyzed with respect to their components. Some of these components are:




the user's role



the system's role



additional roles (e.g., intermediaries, cooperating users)



the style of communication (e.g. commands vs. general communica
tion)



the channels used.


Structuring the paradigms along these lines might lead to a framework for interaction
design and analysis which would reduce the complexity of user interface design by
identifying components and offering specific design choices
for each component. The
major difference of such a framework to similar attempts in the literature would be its
conceptual basis, as opposed to the usual technology
-
driven methodologies focusing
primarily on surface issues like the use of WIMPs.



Conclusi
ons


The main thesis of this paper has been that the task
-
independent conception of system use
underlying a design is essential for the usability of the system. This conception is first of
all a matter of paradigms selection. The selection of task
-
specific

metaphors (e.g., a map
metaphor for certain spatial analysis functions) depends to some extent on this choice and
occurs later in the conceptual design of a user interface. Current interface design
methodologies, however, do not deal at all with the forme
r selection and hardly with the
latter [Kuhn and Frank 1991].


This paper could only suggest to consider paradigms and illustrate some of them
with their origins and consequences. A lot more work, including experimental studies, is
needed to refine our und
erstanding of these concepts and to advance the theory of user
interface metaphors to a point where it becomes directly applicable to system design.
What this paper should have achieved, is a clarification of some terminological issues
around interface met
aphors. It has also attempted a preliminary answer to the frequently
asked question of how to select appropriate metaphors for GIS [Gould and McGranaghan
1990], by distinguishing two levels of metaphors in user interfaces: paradigms and task
metaphors.


Th
e use of information systems, particularly of GIS, can rarely be conceived of by
only one of the listed paradigms. In fact, one of the reasons for the fairly limited usability
of current systems may well be that they are built around a single paradigm and
force
their users into a certain role (mostly that of a programmer or query language user). This
role and the corresponding style of user
-
system communication may be suitable for some
users and some tasks, but hardly for all users and all tasks they want t
o use a GIS for.


Future GIS will have to offer multiple ways of using them and smooth combinations
or transformations between these. For example, some tasks, like placing a label on a map,
are best done by direct manipulation, while others, like retrievin
g all building permits
issued for a certain area, are cases for delegation. Humans are accustomed to taking
different roles at different times, and complex tasks require not only a collaboration
among several users, but also a possibility for users to swit
ch between different roles and
communication styles.


Acknowledgment


Discussions with the leaders and participants of the Specialist Meeting of NCGIA's
research initiative on user interfaces for GIS (Initiative 13), as well as presentations by
Bob Jacobs
on and Tom Furness have directly influenced some ideas expressed here. The
support of NSF through the NCGIA grant is gratefully acknowledged.



References


Brooks, F. 1988. Grasping Reality Through Illusion: Interactive Graphics Serving
Science.
Proceedings, ACM SIGCHI Conference on Human Factors in Computing
Systems (CHI'88).
1
-
11.

Card, S.K., G.G. Robertson, and J.D. Mackinlay. 1991. The Information Visualizer, An
Information Workspace.
Proceedings, Human Factors in Computing Systems
(CHI'91).
E
dited by S. P. Robertson, G. M. Olson and J. S. Olson. 181
-
188.

Carroll, J.M., and J.R. Olson. 1988. Mental Models in Human
-
Computer Interaction. In
Handbook of Human
-
Computer Interaction.
Edited by M. Helander. 45
-
65. North
-
Holland: Elsevier Science Publi
shers.

Danesi, M. 1990. Thinking is seeing: Visual metaphors and the nature of abstract
thought.
Semiotica

80 (3/4): 221
-
237.

DuBoulay, B., T. O'Shea, and J. Monk. 1981. The black box inside the glass box:
Presenting computing concepts to novices.
Internat
ional Journal of Man Machine
Studies,

14 237
-
249.

Egenhofer, M.J. 1989. Spatial Query Languages. Ph.D., University of Maine.

Egenhofer, M.J., and W. Kuhn. 1991. Visualizing Spatial Query Results: The Limitations
of SQL.
Proceedings, IFIP Workshop on Visual

Databases.
Edited by E. Knuth.

Erickson, T., and G. Salomon. 1991. Designing a Desktop Information System:
Observations and Issues.
Proceedings, Human Factors in Computing Systems
(CHI'91).
Edited by S. P. Robertson, G. M. Olson and J. S. Olson. 49
-
54.

F
ischer, G., and A.C. Lemke. 1988. Construction Kits and Design Environments: Steps
Toward Human Problem
-
Domain Communication. 3 (3): 179
-
222.

Foley, J.D., A. van Dam, S. Feiner, and J. Hughes. 1990.
Computer Graphics: Principles
and Practice.
2nd ed. The
Systems Programming Series. Addison
-
Wesley.

Frank, A.U. 1991. Beyond Query Languages For Geographic Databases: Data Cubes and
Maps.
Proceedings, Int. Workshop on Database Management Systems for
geographical applications.
Edited by G. Ausiello. 74
-
86.

Furne
ss, T. 1991. Closing Address on Virtual Reality.
CHI'91: Conference on Human
Factors in Computing Systems.

Gould, M.D., and M. McGranaghan. 1990. Metaphor in Geographic Information
Systems.
Proceedings, Fourth International Symposium on Spatial Data Handl
ing.
Edited by K. Brassel and H. Kishimoto. 433
-
442.

Greif, I. 1988.
Computer
-
Supported Cooperative Work: A Book of Readings
. Morgan
Kaufmann Publishers.

Jackson, J.P. 1990. Developing an effective human interface for geographical information
systems using

metaphors.
Proceedings, ACSM/ASPRS Annual Convention.
117
-
125.

Johnson, M. 1987.
The Body in the Mind: The Bodily Basis of Meaning, Imagination, and
Reason.
The University of Chicago Press.

Kuhn, T.S. 1962.
The Structure of Scientific Revolutions.

Univers
ity of Chicago Press.

Kuhn, W. 1990. Editing Spatial Relations.
Proceedings, Fourth International Symposium
on Spatial Data Handling.
Edited by K. Brassel and H. Kishimoto. 423
-
432.

Kuhn, W. 1991. Are Displays Maps or Views?
Proceedings, Tenth International
Symposium on Computer
-
Assisted Cartography (Auto Carto 10).
Edited by D.
Mark and D. White. 261
-
274.

Kuhn, W., and A.U. Frank. 1991. A Formalization of Metaphors and Image
-
Schemas in
User Interfaces. In
Cognitive and Lingui
stic Aspects of Geographic Space.
Edited
by D. M. Mark and A. U. Frank. Kluwer Academic Publishers.

Lakoff, G., and M. Johnson. 1980.
Metaphors We Live By.
The University of Chicago
Press.

Luff, P., N. Gilbert, and D. Frohlich. 1990.
Computers and Conversa
tion
. Academic
Press.

Mark, D. 1989. Cognitive Image
-
Schemata for Geographic Information: Relations to
User Views and GIS Interfaces.
Proceedings, GIS/LIS'89.
551
-
560.

Mark, D.M., and M.D. Gould. 1991. Interacting with Geographic Information: A
Commentary.

Photogrammetric Engineering and Remote Sensing
57 (11): 1427
-
1430.

Negroponte, N. 1989. An Iconoclastic View Beyond the Desktop Metaphor.
International
Journal of Human
-
Computer Interaction

1 (1): 109
-

113.

Newell, A., and H.A. Simon. 1972.
Human problem solving.
Prentice
-
Hall.

Nickerson, R.S. 1977. On Conversational Interaction with Computers. In
User
-
Oriented
Design of Interactive Graphics Systems.
101
-
113. ACM.

Norman, D.A. 1988.
The Psychology of Everyday Things.
Basic Books.

Polya, G. 19
45.
How to solve it.
Princeton University Press.

Sacks, O. 1990. Neurology and the Soul.
New York Review,
November 22, 1990, 44
-
50.

Salton, G., and M. McGill. 1983.
Introduction to Modern Information Retrieval.
McGraw
-
Hill.

Shneiderman, B. 1983. Direct Ma
nipulation: A Step Beyond Programming Languages.
IEEE Computer

16 (8): 57
-
68.

White, M. 1991. Car Navigation Systems. In
Geographic Information Systems:
Principles and Applications.
Edited by D. Maguire, M. Goodchild and D. Rhind.
115
-
125. Longman Publishi
ng Co.

Wilson, P.M. 1990. Get your desktop metaphor off my drafting table: User interface
design for spatial data handling.
Proceedings, Fourth International Symposium on
Spatial Data Handling.
Edited by K. Brassel

and

. Kishimoto. 455
-
462.