I-17 Specialist Meeting Report, Appendix D
Go back to previous section
1.2 Data Models for Geographic Information
1.5 Knowledge Representation
2.2 Exploratory Spatial Analysis
2.4 Spatial Models
3.1.1 Human-Computer Interaction
3.1.3 Spatial Decision Support Systems
3.3.2 Visualization Tools
Paul Densham (Geography, SUNY-Buffalo)
Marc Armstrong (Geography, Iowa)
Frank Davis (Geography, UC-Santa Barbara)
Core Planning Group
Mike Batty *
Britton Harris *
Joe Ferreira *
(* indicates that we have approached this person and that they have given their consent.)
to be Involved
Geography, Computer Science, Operations Research, Management Science, Planning,
Batty (Geography, SUNY-Buffalo)
Buttenfield (Geography, SUNY-Buffalo),
Calkins (Geography, SUNY-Buffalo),
Mark (Geography, SUNY-Buffalo),
Church (Geography, UC-Santa Barbara),
Couclelis (Geography, UC-Santa Barbara),
Golledge (Geography, UC-Santa Barbara),
Lanter (Geography, UC-Santa Barbara)
Onsrud (Survey Engineering, Maine)
Pinto (Survey Engineering, Maine)
Specialist Meeting: To be held immediately before GIS/LIS '94 (Phoenix,
Closing Session: To be held at GIS/LIS '96.
1. INTRODUCTION The broad adoption of GIS technology has been fueled, in part, by its ability to support interdisciplinary approaches to spatial problem-solving. The traditional layered view of spatial data supported by GIS provides a means through which thematic data coverages can be integrated in a common spatial framework to support analyses conducted from different disciplinary perspectives. With such a repository of layered information in place, a host of powerful analytical operations can be brought to bear on spatially referenced data (e.g., Tomlin, 1990). GIS technology has been especially successful when these operations are applied to problems that are well understood -- problems with clearly defined questions and measurable outcomes. Many spatial problems, however, are not so straightforward. Consequently, spatial decision support systems (SDSS) have been developed to address ill-structured problems with spatial query, modelling and analysis, and display capabilities (Densham, 1991; Guariso and Werthner, 1989).
A mismatch exists, however, between the widespread single-user model of GIS and SDSS use and the group-based approach to decision-making that is often adopted when ill-structured public policy issues are addressed. SDSS-based spatial analysis and display methods must be expanded to encompass group decision-making processes, and new tools must be developed that will enable group members to generate, evaluate, and illustrate the strong and weak points of alternative scenarios and come to a consensus about how to proceed toward a decision.
Collective decision-making activities can be supported by enabling each member of a group to share a common view of a problem as they would when looking at a diagram on a chalkboard. Though this process is especially effective in decision room environments in which the participants are in close proximity (Desanctis and Gallupe, 1985), it can be extrapolated to environments in which the participants are dispersed if high bandwidth communication technology is available (see Newton, Zwart and Cavill, 1992).
Stefik et al. (1987) describe a decision room environment in which each
participant in a group views the current state of the problem they are
attempting to solve; in such WYSIWIS (What You See
Is What I See) environments, each group member has
a display (or views a collective display) that can be altered by other group
members to reflect different views of a problem, and these modifications are
then propagated to other displays. Several steps in the decision process have
been observed when a WYSIWIS interface is supported (Stefik et al.,
1987). The first is a free-form stage in which rough ideas are put before the
group using a variety of computer tools (e.g. typing, drawing). In the second
stage, ideas are sorted and evaluated as the plan begins to take shape. During
this stage there is greater chance for conflict because specific aspects of
alternatives are discussed, evaluated and possibly discarded. During the final
stage, a plan is formalized and articulated through the system. These stages
parallel closely those observed when an SDSS is used to solve a locational
Group use of spatial decision support systems
When decision-makers use an SDSS they become involved in the process of
seeking a solution, and they often generate and evaluate several scenarios that
result from the application of models that employ different criteria or
constraints. In this iterative process, they typically pass through several
In CSDM, a scenario can be supported or objected to by other members of the group using statistical evidence (e.g. the value of an objective function), maps that illustrate the advantages or limitations of a scenario, knowledge about the problem domain and study area, or even intuition. Since haggling and discussion about the merits of scenarios are important aspects of decision-making, especially when there is no clear single "optimal" solution, the system must provide a way for decision-makers to interact with, and to redesign, alternatives.
In addition to a standard set of mapping and report generation tools, a free-form sketching facility would enable users to annotate and highlight specific aspects of a map or table to bring out salient problems or advantages of specific locational configurations in a scenario. When users are so empowered, solutions are no longer viewed with suspicion. At present, however, there is insufficient knowledge about the kinds of drawing tools that are best applied to highlight the salient characteristics of scenarios and differences among them.
Ultimately, when different positions are articulated by group members, an
agreed upon process of resolving deadlocks must be implemented. This process
itself may occur in stages and Nunamaker et al. (1991) describe several
tools that can be used. Electronic questionnaires, for example, can be used to
determine the degree and nature of disagreement among group members. A group
matrix tool can be used to build consensus by enabling individuals to place
their questionnaire responses in a group context. In a shared workspace, users
are able to alter their responses in a matrix; these alterations are broadcast
to other group members. Because these alterations are visible, group matrices
are useful for promoting discussion about different positions. A person who
holds a position contrary to others, but who may not have good support for it,
also may decide to move with the consensus view when group matrix information
is made available to them. Finally, different types of voting strategies (e.g.
majority, plurality) can be used in cases where complete consensus cannot be
Several impediments must be overcome before effective CSDM environments can be
implemented. Research must be conducted into the design of user interfaces for
the groupware tools described in the previous section. New technology and
algorithms also must be developed and applied to support interaction and to
meet the increased computational demands that will be created by group-based
modeling, communication and display of spatial information.
The design of user interfaces that will effectively support group
decision-making and enable individuals to resolve conflicts promises to be a
challenging task. Designers concerned with single-user systems are finding
that appropriate metaphors for interfaces (e.g. desktop, rooms) are difficult
to specify for geographical domains (e.g. Kuhn, 1991; Kuhn, 1992; Mark, 1992).
In addition to metaphors, interface designers must concern themselves with the
set of tasks that must be accomplished by users. Task analysis (Rasmussen,
1986) and knowledge elicitation procedures (Greenwell, 1988) can be used to
determine these required activities so that appropriate options are made
available in a logical and consistent sequence for system users. The ultimate
goal of these research and development activities is to provide interfaces that
enable users with no prior collaborative experience to work together to solve
complex locational problems.
For example, specific facility locations could be selected by pointing, and their positions could be dragged to show how an alternative configuration of supply would affect reassignment of demand. The magnitude and location of demand also could be altered, thus permitting decision-makers to evaluate the impact of development plans under different assumptions of growth or decline of demand for services. Other structures or metaphors will need to be developed for group SDSS applications.
A WYSIWIS system holds considerable promise in locational decision-making
contexts, because maps are an essential and often requested SDSS decision aid
(Armstrong et al., 1991; 1992). Decision-makers, for example, often
wish to create maps that show an existing service system and the relationship
between supply locations and demand for service. Solutions provided by spatial
models (e.g. location-allocation) also are more easily interpreted when viewed
as maps (Harris, 1988; Armstrong et al., 1992). Furthermore, maps serve
as an effective and data-dense mechanism for exchange of locational scenarios,
and consequently, they can serve as the basic "token" of interchange among
locational decision-makers as they evaluate and compare scenarios and serve as
an effective mechanism for promoting discussion of alternative results. They
enable the outcome of a decision process to be modified by unmodeled aspects of
the decision. Additional research, however, must be conducted to determine the
kinds of map displays that are most effective in communicating spatial
information to groups, and whether different kinds of maps are most effective
during different stages of group decision-making.
Decision-makers must be provided with the means to interact with problems in
near-real-time so that they may visualize the effects of making adjustments to
the parameters that define the solution space of a problem. Currently,
locational models are so computationally intensive (Armstrong and Densham,
1992) that near-real-time interaction is precluded except for trivial or small
problems. Instead, the state-of-the-practice is to create scenarios in what
amounts to batch mode because realistic problems require several minutes of
execution time, even on the current generation of high-end workstations.
Research must be performed to determine how computer architectures can be
exploited to improve performance to a level that is required to support true
interactive modeling and design of service systems in a group SDSS environment.
Such capabilities can be supported by multi-processor systems, in which
different processors are assigned specific tasks (e.g. modeling with
alternative criteria) that ultimately lead to the creation of a completed
scenario. Alternatively, a single locational problem can be decomposed into a
set of independent constituent parts to improve solution times through parallel
processing (e.g. Armstrong and Densham, 1992). The use of coordinated
ensembles of processors (Carriero and Gelernter, 1990; Karp et al.,
1993) appears to be especially promising for parallel processing of locational
problems. Note, however, that such ensembles will require greatly improved
network bandwidth to be effective. Myers (1993) describes one ensemble, as a
"metacomputer" that requires a one gigabit-per-second network to link
heterogeneous computational resources.
When response times improve to permit real-time interaction, users will be able
to manipulate directly two (or perhaps more) parameters or constraints much
like a driver simultaneously releases the clutch and presses the accelerator in
an automobile with a manual transmission (Armstrong, Densham and Lolonis,
1991). For example, in many analyses the imposition of a maximum travel
distance constraint will lead to infeasibility of solutions when there are too
few facilities to serve demand. By being able to visualize and evaluate the
interplay between these two parameters and how it affects solutions (in this
example the size and location of unserved areas), decision-makers will gain new
insight into the nature of trade-offs in multi-objective decisions.
Although decision-makers are knowledgeable, they typically are not experts in
methods of spatial analysis and decision analysis. Consequently, a CSDM
environment must actively help decision-makers employ its often complex
analytical and evaluative capabilities (Armstrong et al., 1990). One
way to provide this assistance is to incorporate knowledge in a CSDM system.
Environmental knowledge describes the problem domain. While some environmental knowledge can be captured and represented within a system - including spatial relationships and patterns of spatial interaction - other forms of environmental knowledge are brought into the decision-making process by individuals - an understanding of the local political milieu, for example.
Procedural knowledge is domain-dependent knowledge which can be used to restrict the solution space that will be searched. Computer systems that successfully exploit procedural knowledge often are called "intelligent" and increasingly are used in diagnostic tasks ranging from medicine to automobile maintenance. In a spatial context, procedural knowledge can be used to select a general problem-solving strategy or a specific analytical approach.
Structural knowledge is used to reduce the amount of computation required when an algorithm is applied to a particular problem. Structural knowledge consists of representations of the spatial relationships which are being analysed; exploiting this structure can reap large savings in computation.
To build knowledge-based CSDM environments, we must answer four questions:
Supporting visual interactive modelling for groups requires a fresh approach to the design, representation and implementation of spatial models. One approach is to develop a model base management system (MBMS) that enables users to access spatial modelling capabilities and to combine them in flexible sequences (Densham, 1991). In a MBMS, models typically are reduced to some set of atomic components which can be stored, manipulated, and recombined to yield the original algorithms. If they can be made independent of each other, model atoms can be combined in new ways. The ability to combine atoms in a flexible manner greatly extends the capabilities of the model base. Moreover, defining procedural knowledge for atoms is much simpler than for whole algorithms. The design and implementation of MBMS for CSDM environments raises several questions:
Armstrong, M.P., De, S., Densham, P.J., Lolonis, P., Rushton, G. and Tewari, V. (1990) A knowledge based approach for supporting locational decision-making. Environment and Planning B: Planning and Design, 17: 341-364.
Armstrong, M.P., Rushton, G., Honey, R., Dalziel, B.T., Lolonis, P., De, S., and Densham, P.J. (1991) A decision support system for regionalizing service delivery systems. Computers, Environment and Urban Systems, 15 (1), 37-53.
Armstrong, M.P. and Densham, P.J. (1992) Domain decomposition for parallel processing of spatial problems. Computers, Environment and Urban Systems, 16 (6), 497-513.
Armstrong, M.P., Densham, P.J., Lolonis, P., and Rushton, G. (1992) Cartographic displays to support locational decision-making. Cartography and Geographic Information Systems, 19 (2), 154-164.
Carriero, N. and Gelernter, D. (1990) How to Write Parallel Programs: A First Course. MIT Press, Cambridge, MA.
Carver, S.J. (1991) Integrating multi-criteria evaluation with geographical information systems. International Journal of Geographical Information Systems, 5(3), 321-339.
Conklin, J., and Begeman, M.L. (1988) gIBIS: A hypertext tool for exploratory policy discussion. ACM Transactions on Office Information Systems, 6 (4), 303-331.
Densham, P.J. (1991) Spatial decision support systems. In D.J. Maguire, M.F. Goodchild and D.W. Rhind (eds) Geographical Information Systems: Principles and Applications, Longman, London, pp. 403-412.
Densham, P.J., and Armstrong, M.P. (1993) A Heterogeneous Processing Approach to Spatial Decision Support Systems. Manuscript available from the authors.
DePinto, J.V., H.W. Calkins, P.J. Densham, J. Atkinson, W. Guan and H. Lin (1994) Development of GEOWAMS: an approach to the integration of GIS and watershed analysis models. GIS and Civil Engineering, forthcoming.
DeSanctis, G., and Gallupe, B. (1985) Group decision support systems: a new frontier. Data Base, 16 (2), 3-10.
Eastman, J.R., Kyem, P.A.K., and Toledano, J. (1993) A procedure for multi-objective decision making in GIS under conditions of competing objectives. Proceedings, EGIS'93, 438-447.
Greenwell, M. (1988) Knowledge Engineering for Expert Systems. Ellis Horwood, Chichester.
Greif, I., and Sarin, S., (1987) Data sharing in group work. ACM Transactions on Office Information Systems, 5 (2), 187-211.
Greif, I. (1988) Computer-Supported Cooperative Work: A Book of Readings. Morgan Kaufman Publishers, San Mateo, CA.
Grudin, J. (1990) Groupware and cooperative work: Problems and prospects. In Laurel, B. (ed.) The Art of Human-Computer Interface Design. Addison-Wesley, Reading, MA, pp. 171-185.
Guariso, G., and Werthner, H. (1989) Environmental Decision Support Systems. John Wiley, New York, NY.
Harris, B. (1988) Prospects for computing in environmental and urban affairs. Computers, Environment and Urban Systems, 12 (1), 3-12.
Honey, R., Rushton, G., Armstrong, M.P., Lolonis, P., Dalziel, B.T., De, S. and Densham, P.J. (1991) Stages in the adoption of a spatial decision support system for reorganizing service delivery regions. Environment and Planning C: Government and Policy, 9 (1), 51-63.
Karp, A.H., Miura, K., and Simon, H. (1993) The 1992 Gordon Bell prize winners: Judges summary. IEEE Computer, 26 (1), 77-82.
Kaplan, S.M., Carroll, A.M. and MacGregor, K.J. (1991) Supporting collaboration processes with ConversationBuilder. ACM SIGOIS Bulletin, 12 (2&3), 69-79.
Kuhn, W. (1991) Are displays maps or views? Proceedings of the Tenth International Symposium on Computer-Assisted Cartography. American Congress on Surveying and Mapping, Bethesda, MD, pp. 261-274.
Kuhn, W. (1992) Paradigms of GIS use. Proceedings of the Fifth International Symposium on Spatial Data Handling. IGU Commission on GIS, Columbia, SC, pp. 91-103.
Lai, K., Malone, T., and Yu, K. (1988) Object Lens: A "spreadsheet" for cooperative work. ACM Transactions on Office Information Systems, 6 (4), 332-353.
Lanter D.P. (1991) Design of a lineage-based meta-data base for GIS. Cartography and Geographic Information Systems, 18 (4), 255-261.
Laurel, B. , ed. (1990) The Art of Human-Computer Interface Design. Addison-Wesley, Reading, MA.
Malanson, G.P., Armstrong, M.P. and Bennett, D.A. (1993) Fragmented forest response to climatic warming and disturbance. In press, Proceedings of the Second International Conference on Integrating Geographic Information Systems and Environmental Modeling, Breckenridge, CO, September.
Mark, D.M. (1992) Spatial metaphors for human-computer interaction. Proceedings of the Fifth International Symposium on Spatial Data Handling. IGU Commission on GIS, Columbia, SC, pp. 104-112.
Myers, W. (1993) Supercomputing 92 reaches down to the workstation. IEEE Computer, 26 (1), 113-117.
Newton, P.W., Zwart, P.R. and Cavill, M.E. (1992) Networking Spatial Information Systems. Bellhaven Press, New York, NY.
Nunamaker, J.F., Dennis, A.R., Valacich, J.S., Vogel, D.R., and George, J.F. (1991) Electronic meeting systems to support group work. Communications of the ACM, 34 (7), 40-61.
Press, L. (1992) Dynabook revisited- portable computers past, present and future. Communications of the ACM, 35 (3), 25-32.
Rasmussen, J. (1986) Information Processing and Human Machine Interaction: An Approach to Cognitive Engineering. North-Holland, New York, NY.
Stefik, M., Foster, G., Bobrow, D.G., Kahn, K., Lanning, S., and Suchman, L. (1987) Beyond the chalkboard: Computer support for collaboration and problem solving in meetings. Communications of the ACM, 30 (1), 32-47.
Tomlin, C.D. (1990) Geographic Information Systems and Cartographic Modeling. Prentice Hall, Englewood Cliffs, NJ.
Comments to Karen Kemp