Full text of DOI:10.1111/0004-5608.00233

GIS-Supported Collaborative Decision Making:
Results of an Experiment
Piotr Jankowski* and Timothy Nyerges**
*
Department of Geography, University of Idaho
*
* Department of Geography, University of Washington
This article addresses research questions about the sociobehavioral dynamics of geographic information system (GIS)
use during collaborative decision making in small interorganizational groups. Using an experimental design of a con-
ference room setting, a study of human-computer-human interaction was conducted with 109 volunteer participants
formed into 22 groups, each group representing multiple organizational stakeholder perspectives. The experiment in-
volved the use of GIS integrated with multiple criteria decision models to support group-based decision making con-
cerned with the selection of habitat restoration sites in the Duwamish Waterway of Seattle, Washington. Findings re-
presentative of four categories of investigation are presented. In the first category, the experiment demonstrated that
groups used maps predominantly to visualize the evaluation results and much less to structure/design the decision
problem. Maps played only a limited support role in various decision stages of the experiment. In the second category,
while the use of multiple criteria decision models by groups remained steady throughout different phases of the deci-
sion process, the use of maps was much lower during the initial exploratory-structuring phase than during the later
analytic-integrating phase. In category three, the amount of prior and acquired group member experience with com-
puter tools had no influence on the appropriation of decision aids. In category four, different phases of the decision
process had two different levels of conflict: the exploratory-structuring phase was characterized by a lower level of
conflict, and the analytic-integrating phase was characterized by high conflict level. The higher level of conflict dur-
ing the analytic-integrating phase tells us that analytical decision aids aimed at conflict management are likely to
help work through conflict, such conflict now being recognized as a necessary part of making progress in public de-
cision problems. Key Words: collaboration, GIS, group, site selection, spatial decision support, stakeholder.
n increasing number of problems in public policy
formation and implementation are being recog-
arises over public-private problems due to differences in
values, motives, and/or locational perspectives with re-
spect to what is to be accomplished (Popper 1981;
McGrath 1984; Susskind and Cruikshank 1987; Gregory
1999). In this context, the spatial character of problems is
special due to externalities that arise from adjacent loca-
tions being “valued” and perhaps used differently. Loca-
tional conflict arises from those spatial externalities. In
such situations, conflict—and therefore negotiation man-
agement in shared decision making—is a fundamental
concern in coming to consensus about choices to be made
(Duffy, Roseland, and Gunton 1996; Susskind and Field
1996; Simosi and Allen 1998).
Dealing with locational conflict in an open manner is
becoming more important as citizen (stakeholder) partici-
pation increases in land use, natural resource, and envi-
ronmental decision making (Parenteau 1988; Crowfoot
and Wondolleck 1990; Gregory 1999). The primary ra-
tionale for enhanced stakeholder participation in public
land planning is based on the democratic maxim that
those affected by a decision should participate directly in
the decision making process (Smith 1982; Parenteau
1988; Gregory 1999). It has been said that decision
A
nized as candidates for public-private collabora-
tions. Examples of such problems in a geographic domain
include locally unwanted land uses (LULUs) that instigate
“
not in my back yard” (NIMBY) controversies, such as
landfill and hazardous waste facility siting (Popper 1981;
Lake 1987; Susskind and Cruikshank 1987; Couclelis and
Monmonier 1995; Schneider, Oppermann, and Renn
1
998), polluted urban land use (so-called brownfield) re-
development projects (Bartsch and Collaton 1997; Davis
and Margolis 1997), and salmon habitat restoration plans
(
NOAA 1993; Brunell 1999). Most of those public-
private problems are called “wicked” and “ill-structured”
Rittel and Webber 1973) because they contain intangi-
bles not easily quantified and modeled, structures only par-
tially known or burdened by uncertainties, and potential
solutions mired by competing values, interests, and per-
spectives (Susskind and Cruikshank 1987). Public-private
problems involve a mix of concerns, some public (like
those about legitimacy and credibility) and some private
(
(
like those about financial efficiency due to the expendi-
ture of large amounts of funds). Locational conflict often
Annals of the Association of American Geographers, 91(1), 2001, p. 48–70
©
2001 by Association of American Geographers
Published by Blackwell Publishers, 350 Main Street, Malden, MA 02148, and 108 Cowley Road, Oxford, OX4 1JF, UK.

GIS-Supported Collaborative Decision Making
49
making groups are both fundamental building blocks
and agents of change within organizations, communities,
and society (Poole 1985). Further, Zey (1992, 22) states
of a situation (Roschelle and Teasley 1995). Working in
a collaborative manner, participants create synergy, and
each comes away with a sense of the way in which to un-
dertake decision making.
“
. . . that decisions [in society] are most frequently made
by groups within the context of larger social collectives.”
In Zey’s 1992 book on decision making, a group of ex-
perts in decision making research conclude that “. . . the
resolution of conflicts of values (individual and group)
and of inconsistencies in belief will continue to be highly
productive directions of inquiry, addressed to issues of
great importance to society” (Simon et al. 1992, 53).
The above perspectives indicate a broad-based need
for research on both group decision making in general
and collaborative spatial decision making (CSDM) more
specifically. Research about collaborative spatial decision
making is being encouraged by the research agenda of the
National Center for Geographic Information and Analy-
sis (Densham, Armstrong, and Kemp 1995). That agenda
is linked to an international research effort concerning
the development and use of group-based GIS (geographic
information systems) technology that can be integrated
with other computer technologies to facilitate group col-
laboration, problem-solving, and decision making on
projects with an inherently geographical character. Thus,
the general goal of our research is to develop geographic
information technology that can address the needs of a
diverse set of participants involved in CSDM.
In undertaking such research it is important to recog-
nize at least four cumulative levels of “social interaction”
during group decision making that can be described un-
der the umbrella term of “participation”: communication,
cooperation, coordination, and collaboration. At a basic
level of participation, people communicate with each
other to exchange ideas in a fundamental process of social
interaction. In public decision contexts, the traditional fo-
rum of a public meeting provides for communicative inter-
action—but only at this basic level, a drawback in meet-
ings at which truly constructive comments are desired
In this article we emphasize collaborative decision
making, including a negotiated understanding of issues,
because this context demands the most of participants’
effort and time. Furthermore, we emphasize collabora-
tive spatial decision making because of a growing interest
in understanding societal impacts of the development
and use of geographic information technology that can
represent “value differences” related to spatial externali-
ties and locational conflict. Representing value differ-
ences is important to “idea differentiation,” which in
turn is necessary as a step toward “idea integration.”
Stakeholder behavior involving conflict requires both
idea differentiation and integration before conflict reso-
lution can occur (Susskind and Cruikshank 1987; De-
Sanctis and Poole 1994).
The background literature in our area of research relies
heavily on material and developments in the management
and decision sciences regarding concepts and findings re-
lated to group work and decision support. Development of
group support systems (GSS) technology (Coleman and
Khanna 1995), including group decision support systems
(GDSS; DeSanctis and Gallupe 1987; Hwang and Lin
1987), as well as theoretical and empirical studies of the
use of both (Gray, Vogel, and Beauclair 1990; Gray et al.
1992; Jessup and Valacich 1993; Chun and Park 1998),
have been carried out in the management and decision
sciences since the early 1980s. Most of the empirical stud-
ies have been laboratory experiments conducted in con-
ference room settings with subject-groups using GDSS
software. A major purpose of the experiments was to un-
derstand the implications for group decision processes
and decision outcomes of using decision support software.
Although these experiments have yielded mixed and in-
conclusive results, they provide valuable insights into the
effects on group performance, group member attitudes,
level of participation, and group conflict of using such
software (Chun and Park 1998).
(
Schneider, Oppermann, and Renn 1998). At the next
level of social interaction, building on a set of ideas de-
veloped through basic communication can be considered
to be cooperative interaction. Participants in a coopera-
tive activity each agree to make a contribution that can
be exchanged, but each can also take the results of the
interaction away with them and act on the results as they
see fit, with no further interaction required. A coordinated
interaction is one whereby participants agree to cooperate
and they agree to sequence their cooperative activity for
mutual, synergistic gain. A collaborative interaction is
one whereby the participants in a group agree to work on
the same task (or subtask) either simultaneously or in a
near-simultaneous manner with a shared understanding
As expected, the results suggested that group perfor-
mance—represented by decision time—depends on the
familiarity of users with GDSS tools, but that a skillful
group facilitator can compensate for the lack of user ex-
perience with these tools. According to the results, there
is no advantage to using GDSS for improving decision
quality in simple problems. However, its use becomes ad-
vantageous in complex decision problems. The quality of
decisions in such problems was higher in the experimen-
tal groups that used GDSS software than in groups with-
out GDSS support. User attitudes toward the computer-

5
0
Jankowski and Nyerges
supported decision process, measured by user satisfaction
with GDSS, depended strongly on the presence or ab-
sence of a group facilitator; the facilitator’s presence en-
hanced users’ satisfaction with the process.
(Godschalk et al. 1992; Faber et al. 1994; Faber, Wallace,
and Cuthbertson 1995; Faber, Wallace, and Miller
1996), their offspring, spatial decision support systems
(SDSS) (Carver 1991; Densham 1991; Armstrong 1993;
Eastman et al. 1995; Heywood, Oliver, and Tomlinson
1995; Jankowski 1995; Reitsma 1996; Jankowski et al.
1997; Nyerges et. al. 1998b; Jankowski, Lotov, and
Gusev 1999; Malczewski 1999; Thill 1999), and spatial
understanding (and decision) support systems (SUSS/
SUDSS) (Couclelis and Monmonier 1995; Jankowski
and Stasik 1997) were suggested as information technol-
ogy aids to facilitate geographical problem understand-
ing and decision making for groups, including groups em-
broiled in locational conflict.
Another interesting finding concerned heightened
conflict level among groups supported by GDSS, in con-
trast to earlier supposition that the anonymity of elec-
tronic communications would increase the number of in-
terpersonal exchanges and reduce the chance of one or a
few “strong” individuals dominating a meeting. The an-
onymity feature, enabled by computer network-driven
GDSS, did often embolden group members to communi-
cate more forcefully, thus increasing the perception of
conflict in the group (Chun and Park, 1998).
In contrast to laboratory experiments, the results of the
few field studies conducted in the 1980s are much more
consistent and positive. They demonstrate both increased
decision quality and shortened meeting time when using
GDSS as compared to conventional meetings (Chun and
Park, 1998). They also demonstrate high user satisfaction
and enhanced decision confidence independent of prior
user experience with GDSS. Field studies dealing with
multiworkstation-based GDSS show enhanced participa-
tion and—like laboratory experiments—increased con-
flict from the use of the system. In contrast, field studies of
single workstation-based GDSS led by a facilitator re-
ported both enhanced participation and consensus in ar-
riving at decision recommendations.
Clearly, research concerning collaborative decision
making for geographically oriented public policy problems
continues to gain momentum (Godschalk et al. 1992;
Shiffer 1992; Faber et al. 1994; Couclelis and Monmonier
1995; Densham, Armstrong and Kemp 1995; Faber, Wal-
lace, and Cuthbertson 1995; Golay and Nyerges 1995;
Faber, Wallace, and Miller 1996; Reitsma 1996; Reitsma
et al. 1996; Jankowski et al. 1997; Nyerges and Jankowski
1997; Nyerges, Moore, Montejano, and Compton 1988;
Nyerges, Montejano, Oshiro, and Dadswell 1998). Reduc-
ing the complexity of a decision process by reducing the
cognitive workload of decision makers is one goal of devel-
oping collaborative decision support systems by integrat-
ing capabilities from group decision support systems, GIS,
and multiple criteria decision models (MCDM). Hope-
fully, reducing this workload will lead to a more thorough
treatment of information, exposing initial assumptions
more clearly and resulting in more participatory decisions
(Obermeyer and Pinto 1994). Unfortunately, most of the
research on collaborative spatial decision making is about
GIS development rather than about GIS use, without a
strong theoretical link between the two. Until recently,
little had been done to study the use of GIS technology at
a decision group level. Even though the case can be made
for transferability of research results from group decision
support system experiments to collaborative spatial deci-
sion making, unlike the case of a business decision prob-
lem such as the selection of product marketing plan, spa-
tial decision problems are unique in making location and
associated spatial relationships an explicit part of a spatial
decision situation.
Because the development of commercial group support
systems and group decision support system software in the
1990s (e.g., Lotus Notes from IBM, GroupSystems from
Ventana Corp., and MeetingWorks from Enterprise Solu-
tions) was preceded by studies in the 1980s on group use
of computer technology, one would presume that similar
studies on group use of GIS systems are needed before
commercial GIS software for groups can be developed. A
noted GIS researcher wrote more than a decade ago, “I
conclude that a GIS is best defined as a decision support
system involving the integration of spatially referenced
data in a problem solving environment” (Cowen 1988,
1554). Others who argued that GIS technology fell short
of providing decision analysis capabilities quickly dis-
puted this definition (Densham and Goodchild 1989). A
few years later Lake (1993) offered another perspective,
arguing that GIS might be a step backwards due to its pos-
itivistic approach, encouraging rational planning in the
decision process rather than opening the decision process
to participatory behavior.
Has any progress been made in the 1990s with regard
to GIS technology’s role in decision support? If we mea-
sure the progress in terms of tool development, the
answer is a resounding “yes.” During the 1990s, GIS
Our study was partially motivated by this gap between
the understanding of the implications of using decision
support software in nonspatial group decision processes
and the understanding of the implications of using such
software in spatial situations. The study was further mo-
tivated by the need to develop an understanding of how
GIS software, combined with multiple criteria evalua-

GIS-Supported Collaborative Decision Making
51
tion techniques, is used in group decision processes, and
which components of computer technology fulfill deci-
sion support tasks and which do not. We believe that this
knowledge will enable a better understanding of how
GIS can successfully be used to support collaborative
work involving spatial decision making and problem
solving. We also believe that participatory and collabo-
rative decision making and problem solving will play an
ever-increasing role in public-private decision situations
of the future.
From a social and behavioral science perspective, a
major goal of our research agenda is to broaden and
deepen the conceptual underpinnings of GIS-supported
collaborative decision making by considering social-
behavioral aspects of geographic information use. This is
an important part of geographic information science that
concerns itself with the implications of geographic infor-
mation (system) use within organizations, community,
and society. The study reported in this article fuels the dis-
course about the role of GIS as “tool or science” (Pickles
The “tool versus science” debate parallels a “methodol-
ogy versus theory” debate. The difficulty lies with the
concept of “versus.” Even a discourse about this concept
is a misplaced pursuit for encouraging developments in
science! Recognition of a balance between substance,
theory, and methodology is what is needed to make the
theoretical turn (Pickles 1996; Wright, Goodchild, and
Proctor 1996b). Achieving such a balance involves im-
plementing a research design that includes all three re-
search domains, with one domain taking the lead to set
the stage while the others follow but with all present as
part of the overall contribution (Brinberg and McGrath
1985).
Our study about the dynamics of GIS use in small
groups intentionally balanced substance, theory, and
methodology in a meaningful way, recognizing the con-
straints on interpreting research outcomes. We chose a
substantive public/private problem regarding the selec-
tion of habitat redevelopment sites, with concerns about
legitimacy, credibility, and efficiency (NOAA 1993). The
task of decision making was addressed using various infor-
mation structures provided by a collaborative spatial deci-
1996; Wright, Goodchild, and Proctor 1996a), but along
different emphases from those articulated in the discourse
to date. Wright, Goodchild, and Proctor (1996a) provide
a service to GIS researchers by writing about the multiple
roles of GIS, which is sometimes viewed as a tool and
sometimes viewed as a science. Like Pickles (1996), how-
ever, we doubt the usefulness of an analysis of e-mail dia-
logue about GIS as “tool or science,” as reported by
Wright, Goodchild, and Proctor (1996a) because of a
lack of reflective thought on the topic in terms of theoret-
ical and philosophical grounding. We believe that a more
productive discourse involves unpacking the links be-
tween substance, theory, and method.
In this article, we demonstrate that tool development,
information (tool) use, and scientific examination in-
volving tool development and/or information use all
have their place. However, when that place is grounded
explicitly in theory, research on it may be more recogniz-
able as a contribution to science. In other words, we sup-
port research that is ready to “. . . engage directly and
more substantially the complexities and abstractions of
philosophies of science and the theories of knowledge
and society associated with them” (Pickles 1996, 396).
To ground our contribution to geographic information
science, we point out that research concerned with GIS—
including its social implications, as in any social and be-
havioral science research—should explicitly recognize the
balance and intimate interplay among three research do-
mains: substance, theory (concept), and methodology
1
sion support tool. As the topic of habitat restoration, the
concern about collaborative group decision making, and
the adoption of GIS technology continue to grow in sig-
nificance, we need to better understand the dynamics of
each individually and all three in mutual dependence.
From a theoretical perspective, we used Enhanced
Adaptive Structuration Theory (Nyerges and Jankowski
1997) to motivate certain research questions and subse-
quent hypotheses about social dynamics of collaborative
decision making in small groups. We were confident that
this theory would provide a good starting place from
which to understand the relationships among several sig-
nificant aspects of group interaction during use of GIS
motivated by a complex, practical, public-private prob-
lem. From a methodological perspective we chose to use
a somewhat unconventional approach to data capture,
videotape coding. For analysis, however, we chose a con-
ventional statistical approach. Given the combination
of substance, theory, and method, our empirical results
can tell only a partial story about the potential impacts of
GIS use in collaborative decision making for a site selec-
tion problem. Other research designs providing a differ-
ent balance of substance, theory, and method will of
course be needed to provide other valuable insights.
This article is directed toward making a balanced con-
tribution to the theoretical understanding of GIS as geo-
graphic information science. In the next section, we
present the theoretical framework that guided our artic-
ulation of research questions for examining the use of
maps and multicriteria decision aids in a group setting. In
(Brinberg and McGrath 1985; Silvey and Lawson 1999).
All three domains are present in any study; however, the
strength of each one’s presence is a matter of emphasis.

5
2
Jankowski and Nyerges
the context of that framework, we introduce a substan-
tive problem setting that deals with group-based site se-
lection for salmon habitat redevelopment. Next, we out-
line the methodology we used in undertaking this study,
providing insight into the way we organized the experi-
ment and collected data about group participation. That
leads us to a presentation of the findings for nine hypoth-
eses clustered into four categories, with a follow-up dis-
cussion of each hypothesis to interpret those findings.
We conclude this article by situating its concepts and
findings in the broader research literature about geo-
graphic information and society. We offer prospects for
future areas in which our own and other research en-
deavors might be well worth the effort.
use of certain types of information and not others, thus
creating a certain decision dynamic, is an important one
in bringing about change related to complex problems.
To help orient our research, we enumerated twenty-
one aspects of groups, information technology use, and/
or decision making for complex problem solving in a
review and synthesis of fifteen theoretical frameworks
(Nyerges and Jankowski 1997). No aspect of any theoret-
ical framework was intentionally left out of the enumer-
ation; that is, the list is inclusive. Of the theoretical
frameworks, adaptive structuration theory (AST) treated
the most number of aspects (fifteen), including core con-
cerns regarding use of advanced information technology
in a group setting. AST was originally developed to ex-
plain the dynamics of advanced information technology
use to help structure information development for deci-
sion making, directing change at the intraorganizational
and organizational levels concerned with private organi-
zation problems (DeSanctis and Poole 1994). Further-
more, AST was created from a reconstructionist perspec-
tive, which is why the term “adaptive” is used in the
name (Orlikowski 1992). Enhanced AST (EAST)
broadens AST by adding the remaining six aspects from
the inclusive list (Nyerges and Jankowski 1997); hence,
all aspects of all fifteen theoretical frameworks in the re-
view are represented in EAST.
Theoretical Framework for the Study
Our collaborative spatial decision making research
agenda presumes that decision making groups are both
fundamental building blocks and agents of change
within a variety of organizational, community, and soci-
etal settings. We recognize that groups as decision actors
can be studied at intraorganizational, organizational and
interorganizational levels of aggregation within those
settings. As we move into the twenty-first century, it is
often said that we are experiencing information over-
load—i.e., the cognitive workload is increasing for indi-
viduals and groups alike when it comes to complex prob-
lem solving. The issue of what motivates groups to make
EAST’s twenty-one aspects are significant because
they comprise the details of a conceptual domain that
correspond to potential variables in a methodological
domain, forming the basis of theoretically informed em-
Figure 1. The EAST framework, consisting of convening, process, and outcome constructs (including example aspects) plus the respective
premises (see Table 1), provides a conceptual map for understanding a group decision support situation.

GIS-Supported Collaborative Decision Making
53
Table 1. Research Questions Motivated by Aspects and Premises in EAST
Premises
Research Questions
Convening Concerns
Premise 1. Various information
structures can be appropriated.
• What is the relationship between the usage of maps and decision models? (Hypothesis 1)
What kinds of decision models are appropriated in relation to maps? (Hypothesis 2)
•
Premise 2. Social rules guide the type of • What social norms or rules from various stakeholder organization(s) influence
information structure appropriation.
appropriation?
•
How does task complexity (as established by number of sites and criteria used) influence
what maps and decision models are appropriated? (Hypothesis 9)
Premise 3. A group’s internal social
system influences appropriation.
• How do the different stakeholder perspectives influence the types of decision
aids appropriated?
•
Does prior knowledge/experience with decision aids promote more use of maps and decision
models? (Hypothesis 3)
•
Does knowledge acquired through group decisionmaking participation promote more
effective use of decision models? (Hypothesis 4)
Process Concerns
Premise 4. Appropriation during decision • What is the relationship between maps usage and decision phases? (Hypotheses 5 and 6)
phases influences dynamics of decision • What is the relationship between decision model usage and decision phases? (Hypothesis 7)
(social) interaction
• Are there differences in the level of group conflict associated with different decision phases?
Hypothesis 8)
Does task complexity influence appropriation during a given phase?
Premise 5. Emergent structures influence • What new kinds of information structures are called for as a result of group interaction at the
(
•
the dynamics of the decision process.
inter-organizational level?
What new information emerges as a result of changes in task management?
•
Outcome Concerns
Premise 6. Appropriation and phasing
during decision process influences
decision outcomes
• Do differences in stakeholder (organizational) perspectives and lower or higher levels of
group conflict result in continued collaboration?
• What is the general consensus about the use of information technology in a meeting
environment?
Premise 7. Initial conditions and
decision processes influence
reproduction of social structures.
• Does the relative percentage of time committed to map use during a decisionmaking process
have any influence on decision outcomes?
• Does the relative percentage of time committed to multiple criteria decision model use
during a decisionmaking process have any influence on decision outcomes?
Note: Research questions addressed in the article are denoted by their respective hypotheses. Premises 5, 6, and 7 are not addressed in this article.
pirical research studies (Brinberg and McGrath 1985).
Among the most significant of the six EAST aspects not
present in AST are those that deal with stakeholder
groups at the interorganizational level whose main con-
cern is public problems, where a spatial component is
quite often part of the decision problem. Thus, the con-
ceptual domain for EAST includes at least the three
kinds of settings and three levels of aggregation men-
tioned above, as well as an orientation toward public-
private problems, some of the biggest challenges facing
applications of GIS in society as we enter the twenty-first
century.
The eight constructs of EAST are grouped according
to convening, process, and outcome categories to commu-
nicate their role in decision situations (Figure 1). The
convening constructs articulate what is important in set-
ting up a decision task: the organizations to be repre-
sented, the people from those organizations who are to
participate, and the information technology that can be
made available. The process constructs include the dy-
namics of invoking decision aids, managing decision
tasks from phase to phase, and the emergence of informa-
tion structures such as maps, models, and databases. The
outcome constructs and associated aspects include direct
outcomes related to the specific decision task, and the so-
cial relations created, evolved, and/or destroyed when
the task is complete. Each of the eight constructs is pre-
sented with example aspects of a decision situation, fos-
tering a better understanding of the particular concerns
associated with decision support.
The design, implementation, and analyses of our ex-
perimental study about human-computer-human inter-
action made use of the EAST constructs and aspects in a
direct manner. The constructs and respective aspects
guided development of research questions (articulated in
Table 1, to be discussed later), whereas the variables cor-
responding to the aspects formed the basis of the hypoth-
eses to be tested. As such, EAST helped us link concep-

5
4
Jankowski and Nyerges
tual aspects of human-computer-human interaction to
experimental variables that focused on the selection of
habitat redevelopment sites in the Duwamish Waterway
of Seattle Washington—an interorganizational decision
task. Since our interest during the experiment concen-
trated on the interaction between decision support tools
and decision situation, the EAST framework helped to
illuminate the potential influences of a decision situation
on the use of decision aids, and the reverse influence of
decision aids on the decision situation. Researching this
mutual influence is the crux of a reconstructionist ap-
proach, i.e., the adaptive nature of information technol-
ogy use (Orlikowski 1992). Each of the eight constructs,
together with the more detailed aspects thereof, is de-
scribed below.
ground and perspective issue, we asked participants to
adopt one of five stakeholder perspectives, which in-
cluded a regulatory agency representative, elected offi-
cial, technical expert, environmentalist, and business
leader. Each perspective was described in terms of stake-
holder values in a concept document compiled by a hab-
itat restoration panel organized by the National Oceanic
and Atmospheric Administration (NOAA 1993).
Social and organizational structuring (construct 2) is
likely to have a major influence on decision making. As-
pects of that structuring include the task goal and struc-
ture as well as the organizational rules adopted by a
group. Research on interorganizational collaboration
suggests that one of the most significant reasons people
(organizations) collaborate is because they share a con-
cern (goal) about a topic and want to see something
done, even if for different reasons (Wood and Gray
Convening Constructs and Aspects in
Collaborative Spatial Decision Making
1
991). Using that finding to frame our approach, we
stated in our fliers that the purpose of the experiment was
to better understand “habitat redevelopment site selec-
tion,” a topic that is somewhat popular among citizens of
the Pacific Northwest. We had a reason to believe that
manipulating the complexity of the task might have an
effect on the types of decision support tools used and the
nature of the process (Renn, Webler, and Wiedemann
1995). From a theoretical perspective, each new task def-
inition brings about a different decision situation; thus,
we encourage readers of this article to imagine reapply-
ing the EAST framework (Figure 1) to our attempts to
characterize decision processes relevant to those tasks.
The information technology (construct 1) available
to a group can foster or hinder the way information is
created and used. It has been shown that group-based in-
formation technology can help larger groups come to a
quicker understanding of an array of ideas than is possi-
ble without such technology (Vogel 1993). GIS technol-
ogy has been growing in popularity as a data integration
tool. Decision modeling technology has been growing in
popularity since its introduction on a PC platform. We
presumed that group support technology that integrates
these two kinds of technology would be an interesting
combination of tools that could provide significant syn-
ergy for efficient, effective, and more equitable participa-
tion in decision making processes. In this project we de-
veloped a software extension for ArcView2 (1995) to
generate specialized maps, linked that to a set of multiple
criteria decision models, and linked a group voting tool
to both of those to create a software system called Spatial
Group Choice (Jankowski et al. 1997). Comparing a va-
riety of information structures generated by that tool
formed the basis of the decision support aspects treated
in this construct.
Among the convening constructs of a decision situa-
tion are the group of people who come together to resolve
a decision situation(s) (construct 3), the social structures
and institutional arrangements that provide an organizing
context for the situation (construct 2), and the informa-
tion structures that people will use to address the issues
that are part of the situation (construct 1). There are sev-
eral aspects to each of these constructs.
In this study, the participants responded to a flier that
called for people to participate in an “environmental de-
cision making experiment focusing on habitat redevel-
opment site selection in the Duwamish Waterway of
Seattle, while making use of GIS.” The research team
counted on participants’ familiarity with phrases in the
flier to encourage them to investigate the activity. Not
all people who investigated the activity became a part of
the participant pool, usually due to either lack of further
interest or time constraints. Those that did participate in
the experiment were probably attracted by the one or
more of the following phrases: “environmental decision
making,” “habitat site selection,” and/or “GIS use.” As-
pects (variables) of a group include the type of back-
ground they have in the subject of a problem, their ex-
pertise with software tools, and the personalities they
bring. Together, these aspects form the character of a
group. Getting the appropriate people together to ad-
dress a situation is as challenging as the situation itself,
because it is the people who will turn the collaboration
into a success or failure (Wood and Gray 1991). We sus-
pected that participants with interest and/or background
in environmental decision support would be able to at
least try to work together, although they had no mandate
to do so other than interest. To help with this back-

GIS-Supported Collaborative Decision Making
55
Process Constructs and Aspects in Collaborative
Spatial Decision Making
rules. Sampling these requests of the group was part of
the data gathered for “emerging information.”
Several researchers (e.g., Rohrbaugh 1989; Todd
1995) recommend that decision research should empha-
size “process” rather than “outcomes,” which were em-
phasized in the 1980s, particularly in experimental set-
tings. In terms of outcomes that can be compared to
other studies, because experimental settings are con-
trived, researchers are less likely to identify meaningful
results than to identify spurious ones. For this reason, our
experimental study also focused on process dynamics, as
opposed to outcomes. Nonetheless, the EAST frame-
work is still useful for casting the character of outcomes
because the framework as presented here can be used for
framing studies using any research strategy, including
field experiments, case studies, sample surveys, usability
studies, and laboratory experiments.
Among the process constructs of a decision situation
are the way sociotechnical structures are invoked at any
given time (construct 4), the way a decision process (task)
is managed (construct 5), and the type of information that
emerges during the process (construct 6). Together, these
constructs form the basis of human-computer-human in-
teraction during information use.
Invoking the structures for technology and rules is
called appropriation (construct 4). The way technology is
invoked includes exactly what software and databases are
put to use. Hence, we intended to sample the group uses
of all maps and tables invoked through the Spatial Group
Choice software. Technology is often appropriated and
put to use in ironic, i.e., unexpected ways, but these are
not understood as right and/or wrong (DeSanctis and
Poole 1994). We expected to be able to sample such
uses.
Outcome Constructs and Aspects in Collaborative
Spatial Decision Making
As information technology is put to use, in whatever
form, planning/decision tasks are managed in various
ways to enhance/hinder the chance of successful out-
comes (construct 5). Decision processes are commonly
managed as a set of task activities that require social in-
teraction. Thus, sampling these “social moves” was im-
portant to sampling the decision process. Several re-
searchers have described decision processes in group
settings that helped frame our understanding. Simon
The outcome constructs of a decision situation include
the decision outcomes (construct 7) as well as the social
relationships among participants (construct 8). Sampling
the character of outcomes can provide researchers with
valuable information to use in understanding the directed
nature of processes. Given the multiple criteria by which a
group decision outcome (construct 7) can be assessed, it is
likely that no single “correct” decision exists. Conse-
quently, there are several aspects of decision outcomes
that are important. Among them are the substantive na-
ture of results discovered, the satisfaction with and con-
sensus regarding those results, and the efficiency, effective-
ness, and equity associated with the results.
Changed participant relationships (construct 8) are
often seen as outcomes of decision processes due to the
intense social interaction involved. Some relationships
are created, others reinforced, others tarnished, and still
others broken. Change in social relations is an inherent
part of decision situation interaction. As with decision
outcomes, social outcomes are less meaningfully ob-
served in an experimental setting than in a field setting.
Consequently, this study did not directly pursue research
on this aspect of the collaborative decision situation.
(
1977, 1979) describes managerial decision making in
organizations as a rationally bounded process involving
four steps: intelligence gathering, design of a problem
structure, choice of a set of alternatives, and review of
the process. Renn et al. (1993) describe a public partici-
patory decision making process, used in both the U.S.
and Germany, consisting of three steps: values-criteria
development, alternatives assessment, and alternatives
evaluation (see also Schneider, Oppermann, and Renn
1
998). There is also evidence that many people perform
decision making in something akin to cycles, i.e., as an
iterative process (Poole and Roth 1989; Bhargarva,
Krishnan, and Whinston 1994). Part of the study in-
volved investigation of the dynamics of decision process
in the context of geographic information technology use.
As technology is put to use, and as project tasks are
managed, certain information emerges from the process
Research Questions Motivated by
Premises in EAST
(
construct 6). The emergence of information during the
process reinforces whether technology should be contin-
ually put to use. New databases and/or maps may emerge
from the social interaction. When information emerges
that is useful, this intermediate result encourages the
continued appropriation of technical capabilities and
The constructs and aspects of EAST provide a guide-
line for sampling data. However, the principal reason the
constructs and the aspects of them exist in the way they
do is that each pair of constructs is related through a

5
6
Jankowski and Nyerges
“
premise” (sometimes called a proposition in various
Orthophoto Image: shows area using a photo image
Previous Rank Map: display of current and previous
site ranks using bars
Situation/Context Map: situational/contextual charac-
teristics for the sites
pieces of literature). DeSanctis and Poole (1994) articu-
lated seven premises in AST related to the use of ad-
vanced information technology in decision settings. Al-
though we retain those same premises in EAST, the
interpretation is somewhat broadened due to the en-
hancements of the constructs as described above. Each
premise motivates a series of research questions (see
Table 1) and subsequent hypotheses. It is these premises
and related research questions that comprise the explan-
atory character of EAST. Without the premises, the
framework is simply a descriptive task model.
The decision modeling component, Group Choice, uses
multiple criteria decision models for prioritizing alterna-
tives using various weighting and aggregation schemes, to-
gether with a group voting capability. The following mul-
tiple criteria modeling aids are available in this module:
Select Criteria: dialog box used to select the criteria,
which the group uses for decision modeling
Criteria Valuation: dialog box used to value the criteria
Pairwise Comparison: weighting method where each cri-
terion is compared against every other for preference
Ranking: weighting method that assigns ranks to each
criterion on a scale from 1 to 9
Rating: weighting method that allocates 100 points
across all criteria
Select Alternatives: dialog box that reduces the number
of alternatives considered in the decision modeling
MCDM Window: use of the weights or the decision
model evaluation scores and rank list
Research Design
Our research design involved a laboratory experiment
setting in which we were able to videotape participants
in groups working with computer-oriented geographic
information. The following description of the research
design includes the sociobehavioral setting, subjects, de-
cision tasks as treatments, and research instruments for
data capture.
The sociobehavioral setting involved groups of five par-
ticipants assisted by a facilitator/chauffeur using specially
developed collaborative spatial decision making software
in a decision laboratory. Our choice of five-person groups
stems from Vogel’s (1993) review of several experiments in
GSS research that showed mixed results with groups of
three or four, but beneficial results starting with a group
size of five. The facilitator/chauffeur provided less media-
tion than would a facilitator in a large group, and more
software (technical) support with the overall problem
than would a chauffeur. (In the remainder of this article,
we will refer to this role simply as “facilitator.”)
Sensitivity Analysis Window: use of the sensitivity
analysis window
The maps in the Spatial component are meant to pro-
vide a synoptic view of information about habitat sites,
taken to be the options in the decision task, that are de-
scribed by attributes of those options. The MCDM tables
in the Group Choice component depict options and pri-
orities based on criteria (selected attributes from the
Spatial component) as the fundamental basis of decision
making from an analytical perspective. In addition, the
Spatial component links to the Group Choice compo-
nent to depict graduated symbols that show rank priority.
Decision research is about sorting through criteria associ-
ated with “values” and options described in terms of
those criteria and setting priorities for those options.
The decision lab used for the experiment, located in
the Department of Geography at the University of Wash-
ington, included six 486 PC stations with seventeen-
inch graphics monitors connected to a local area net-
work (LAN). The six PCs were configured in a U-shaped
layout so that the five participants and the facilitator/
chauffeur (the latter played by the same research assis-
tant throughout the study) could see each other and the
public display screen with relative ease. One station was
specially configured for use by the facilitator/chauffeur to
drive the public display screen. The Spatial Group
Choice software installed on the facilitator/chauffeur’s
The CSDM software, called Spatial Group Choice, is
a research prototype described elsewhere in greater detail
(
Jankowski et al. 1997). The software has two modules.
The Spatial component is a GIS/mapping module com-
prised of ArcView 2.1 software from Environmental
Systems Research Institute (ESRI), with a specially con-
figured, simplified user interface and a set of functionality
developed with the aid of ESRI’s Avenue scripting lan-
TM
TM
guage. The following maps are available in the GIS/
mapping module:
Bar Map: display of site attribute values using bars, as
in a bar chart
Consensus Rank Map: display of consensus rankings of
the sites using circles
Graduated Circle Rank Map: display of site ranks using
graduated circles
Site Location Map: site locations and names only

GIS-Supported Collaborative Decision Making
station contained several special features to facilitate
57
group interaction.
The study used 109 participants formed into twenty-
two groups (one group had only four members). They were
recruited from across the University of Washington cam-
pus, and in a few cases from off campus, through an-
nouncements in classes and fliers posted on bulletin
2
boards around campus. No special competence was
sought, only an interest in the environmental decision
task to be undertaken.
Of the 109 participants, 104 finished the study. The av-
erage age of the participants was 28 years. The average ed-
ucation attainment was close to completion of an under-
graduate degree, although there were several graduate
students and participants from off campus with an interest
in GIS and habitat restoration. Rating “attitude toward
working in groups” on a 5-point Likert scale (ꢀ2 ꢁ strong
dislike, 0 ꢁ indifferent, and 2 ꢁ strong like), the partic-
ipants rated an average of .72. Rating “previous work-
group experience” on a 4-point Likert scale (0 ꢁ none,
1
ꢁ experiment group only, 2 ꢁ some work experience,
and 3 ꢁ management), the participants rated an average
of 2.12. Rating “working with computers on a per week
basis” on a 4-point Likert scale (0 ꢁ none, 1 ꢁ 1–5
hours/week, 2 ꢁ 6–20 hours/week, 3 ꢁ greater than 20
hours/week), the participants rated an average of 2.02.
Rating “experience with using GIS maps” on a 4-point
Likert scale (0 ꢁ do not know, 1 ꢁ heard about it and
tried, 2 ꢁ use some, 3 ꢁ use frequently), the participants
rated an average of 1.21, which indicates that overall the
participants were GIS map novices. Rating “experience
with using MCDM models” on a 4-point Likert scale
Figure 2. Map representing the decision situation area used in the
experiment, the Duwamish river corridor with twenty restoration
sites. The gray-shaded land area represents the city of Seattle.
(
0 ꢁ do not know, 1 ꢁ heard about it and tried, 2 ꢁ use
some, 3 ꢁ use frequently) the participants rated an aver-
age of 0.90, which indicates that overall the participants
were also MCDM model novices. Rating “experience
with habitat restoration” on a 4-point Likert scale (0 ꢁ
none, 1 ꢁ education only, 2 ꢁ work only, 3 ꢁ education
and work experience) the participants rated an average
of 0.95, which indicates the vast majority of participants
were novices in experience with habitat restoration, as
well as in use of both GIS tools.
nants into Puget Sound (Elliott Bay), degrading fish and
wildlife habitat. A GIS database for this site selection prob-
lem was compiled from Seattle and King County sources.
The site selection decision process was expected to involve
conflict management during social interaction due to the
different perspectives inherent in the views of participating
members. Thus, these site selection activities were particu-
larly interesting from the standpoint of software tool use
and its interplay with group interaction.
We adopted a realistic decision task to structure our
treatments about site selection for habitat restoration (de-
velopment) in the Duwamish Waterway of Seattle, WA
Each decision group met for five sessions, one in each
of five consecutive weeks (or as close as possible to that
schedule), and worked on a different version of the hab-
3
(
see Figure 2). The decision task was being performed by
the National Oceanic and Atmospheric Administration
NOAA) Habitat Restoration Panel (NOAA 1993) as a
itat site-selection task. In each of the five sessions we
asked each group to work toward consensus on the selec-
tion of three preferred sites (or as many as the $12 mil-
lion budget would allow) out of the total number of sites
presented to them. The total number of sites varied from
eight to twenty. At the end of each session, we asked
each group to fill out a questionnaire, which provided a
(
result of a lawsuit settled against the city of Seattle and King
County for inappropriate storm sewer drain management.
For years, storm sewer drains had been releasing unfiltered
storm water containing highway gasoline and oil contami-

5
8
Jankowski and Nyerges
means for the individuals to assess group use of the tools,
group interaction, and the level of satisfaction with the
overall group selection.
We used a counterbalanced repeated measure design
for the treatments (Girden 1992). The treatments in-
volved varying: task complexity as the number of sites
five-person decision sessions. Our data sources for our
eventual data analyses were this videotaped record and
the session questionnaires. As we were interested in both
software tool use and the overall group interaction, we
focused one video camcorder on the public display screen
and another camcorder (with a remote microphone that
hung overhead in the middle of the participants) on the
group interaction.
(eight versus twenty), cognitive conflict as the number of
criteria (three versus eleven per site), and access to tech-
nology (group and individual access versus group-only ac-
cess). With the group-only access to technology, the
number of sites was set at twenty and the criteria at
eleven. A total of five tasks were constructed and num-
bered from one to five. Groups were randomly assigned a
task number for their first session. They were given a new
task sheet in task sequence at the beginning of each of the
five sessions, and thus cycled through all five tasks by the
fifth session. Although we set up the experiment to exam-
ine cognitive conflict by varying criteria, we later de-
duced that this was a change in task complexity as well.
Consequently, we did not bother to examine cognitive
conflict, but instead looked at task complexity as a varia-
tion in both number of sites and number of criteria, with
the simplest task involving eight sites and three evalua-
tion criteria per site the most complex task involving
twenty sites and eleven evaluation criteria per site. In
general, we expected that the information technologies
would have more positive influences as the task complex-
ity increased.
Data were collected by session (and hence by task)
using questionnaires and coding interaction of video-
tapes. Each participant filled out a background question-
naire (education, sex, age, etc.) and attended a two-hour
collaborative spatial decision making software training
session. At that time, we passed out materials introduc-
ing the overall wildlife habitat site-selection task, as-
signed the participants to groups based on schedule avail-
ability, and handed out stakeholder roles that they could
adopt by the time their first decision session convened.
Based on interviews completed by the NOAA Restora-
tion Panel (NOAA 1993), these roles included those of
business/community leader (twenty participants adopted
it), elected official (ten adopted it), regulatory/resource
agency staff member (twenty-two adopted it), technical/
academic advisor (twenty-three adopted it), or environ-
mental group representative (twenty-nine adopted it).
Roles were self-selected to encourage subjects to partici-
pate based on their inherent interests. We made sure that
no less than three different stakeholder roles were repre-
sented in each of the groups.
We developed and used a set of interaction coding sys-
tems to perform data capture from videotapes on which
we recorded the use of CSDM software as a process of
group interaction. An interaction coding system is a set
of keywords that reliably summarizes the character of a
process from a thematic perspective. We make use of
three coding systems developed for this research: a deci-
sion aid coding system to describe what decision aids are
being used, a decision phases coding systems to describe
what phases exist, and a group working relations coding
system to describe group conflict (Nyerges, Moore, Mon-
tejano, and Compton 1998). Each of the coding systems
is described in more detail in the section of this article
where we report our findings.
Coding was implemented at the level of “group atten-
tion.” Group attention is a level of attention whereby at
least four out of five participants show head-directed
awareness of a group conversation/discussion underway
as the predominant activity in a given time cell. A time
cell is a one-minute interval (called a count) using which
the predominant event/activity occurring is observed and
coded to a database. Thus, for each coding system we cre-
ated an event/activity sequenced track, observing the
videotape three times to code the three coding system
tracks needed to create entries in the database for each
videotape session. Although we used a code category
called “individual work” in the group working relations
coding system, we did not code any details about such
work, since our focus was on “group attention.” Group at-
tention leads to “social cognition,” which is different
than “individual cognition” (Golay and Nyerges 1995).
Therefore, the research associated with “individual atten-
tion”—individual map use coding as reported throughout
the cartography literature—was not applicable to our
work on “group attention” coding. We devised a “map
use coding” system in our initial work on coding sys-
tems, but did not implement it because of our focus on
decision aid “appropriation.” We leave work on map use
and its relation to appropriation to a near future re-
search effort.
To create the database, a SVHS video recorder was
connected to a Macintosh computer running the Mac-
SHAPA software (Sanderson et al. 1994). This software
is specialized to perform exploratory sequential data
Group interaction was videotaped in each session. A
total of seventy-four sessions were used for coding, due to
group participant attrition within the requirement for

GIS-Supported Collaborative Decision Making
59
analysis, which is defined by Sanderson and Fisher
1994, 225) as
tools. In general, we expected that maps and decision
models would have more positive influences on group de-
cision processes as the task complexity increased. In this
research, we were interested in a group response to a spa-
tial decision task considering varying groups’ familiarity
with decision support tools. We did not collect data on
gender, because we randomized this variable, trying not
to bias male and female participation in a group. How-
ever, some groups had more males and some groups had
more females due to scheduling preferences.
(
.
. . any empirical undertaking seeking to analyze systems,
environmental, and/or behavioral data (usually recorded)
in which the sequential integrity of events has been pre-
served. The analysis of such data (a) represents a quest for
their meaning in relation to some research or design ques-
tion, (b) is guided methodologically by one or more tradi-
tions of practice, and (c) is approached (at least at the out-
set) in an exploratory mode.
The selection of hypotheses for this study does not in-
corporate research questions about new information struc-
tures likely to emerge from group interaction during the
course of solving a spatial decision task (premise 5 in Table
However, in addition to data analysis, MacSHAPA
also provides an outstanding set of capabilities for se-
quential data collection from videotape. An interpreted
data observation created by MacSHAPA, a concept used
frequently in the analyses below, is termed a “move.”
MacSHAPA can aggregate several one-minute counts of
the same contiguous code into an interval-event. A
move is a change in an event within a coding system,
e.g., a move from one type of map to another type of map,
or one decision phase to another decision phase. Thus,
moves are “social event-coding changes” that are coded
at the level of “group attention.” We recognize that
1) because groups were limited to the technology provided
to them. A hypothesis around these questions could be
formulated more readily for a field experiment or a case
study where participants are free to identify new informa-
tion structures at will. Our hypotheses concentrate on the
group use of technology during a decision process. They do
not address research questions about process outcomes
(premises 6 and 7 in Table 1) because researchers seem to
agree that testing such concerns in a laboratory experi-
ment is a meaningless endeavor due to the absence of con-
sequences (Rohrbaugh 1989; Todd 1995). Again, such hy-
potheses could be included in a future field experiment or
a case study involving an organizational setting with real
stakeholders and decision makers.
“
map moves” in particular and “decision aid moves” in
general are not strictly speaking the same as “use,” if we
interpret use to be a “set of operations on a decision aid.”
However, as mentioned earlier, the map use coding sys-
tem not yet implemented has been left for a future re-
search effort. Consequently, we use “moves” and “cell
counts” as the basis of our analysis to mirror what has
been used in other group decision support research (De-
Sanctis and Poole 1994).
Appropriating Maps and Decision Models
The hypotheses in the first group, derived from
premise 1 focus on appropriations of maps and decision
models during collaborative spatial decision making.
The frequency and type of interaction between maps and
multiple criteria decision models used during group deci-
sion process have not been reported in the literature.
Findings
The research questions outlined in Table 1 guided the
development of hypotheses, which cluster into four cate-
gories. The categories are associated with EAST premises
1
to 4 respectively. Under premise 1, we report on the re-
Hypothesis 1: No relationship exists between the
number of map moves and the number of MCDM
moves. We tested hypothesis 1 using a Pearson correla-
tion statistic. The value of the Pearson correlation coef-
ficient (r ꢁ ꢀ0.124, p ꢁ 0.001) is statistically significant,
but the correlation is very low. Although we reject the
null hypothesis, the result indicates a very weak inverse
relationship, meaning that the map moves and MCDM
moves are not likely to occur in a systematic manner
across tasks. To determine whether task complexity had
any effect on the interaction between the number of map
moves and the number of MCDM moves, we used a Gen-
eral Linear Model (GLM) statistical procedure (SPSS
Base 8.0 1998). A GLM procedure provides regression
lationship between appropriation of maps and appropria-
tion of decision models during human-computer-human
interaction. Under premise 3, we report about group
member experience and its influence on appropriation of
decision aids. Under premise 4, we report on appropria-
tions of decision aids and the dynamics of the decision
process, described in terms of phases. Finally, under prem-
ises 2 and 4, we report on the relationship between deci-
sion processes and group conflict. In all sections, we ex-
press a hypothesis in terms of a null hypothesis.
Our selection of hypotheses to be tested was driven by
the experimental treatments allowing us to control for
task complexity and group access to decision support

6
0
Jankowski and Nyerges
analysis and the analysis of variance for one dependent
variable by one or more factors and/or variables. The de-
pendent variable was map moves, the covariate was
MCDM moves, and the factor was task complexity. Task
complexity ranged from task 1 (the least complex, 8 sites
and 3 evaluation criteria) to task 4 (the most complex,
2
0 sites and 11 evaluation criteria). The model explains
only 8% of the variability between the use of maps and
2
multiple criteria decision model aids (adjusted R ꢁ .081,
F ꢁ 10.682, p ꢁ 0.000), and task complexity is not a sig-
nificant effect in explaining variability (F ꢁ 1.368, p ꢁ
.252). However, there is an interesting relationship be-
tween map moves, MCDM moves, and task complexity
that is not captured by the above model (see Figure 3).
The increasing task complexity is accompanied by an in-
creasing number of MCDM moves and decreasing num-
ber of map moves. This indicates a dichotomy between
multiple criteria decision model aids and maps: multiple
criteria decision model aids were used more frequently in
more complex collaborative decision tasks, whereas
maps were used more often in simpler tasks.
Discussion of findings for Hypothesis 1. We were espe-
cially interested in the extent to which maps were used
concurrently with MCDM aids. However, based on our
findings, we can only conclude that participants used
maps more often independently from MCDM aids than
in concert with them. During the entire experiment (74
sessions), maps and MCDM aids were used together
Figure 4. Frequency of map use expressed as the percentage of map
moves by map type.
photo images (see Figure 4). The bar map was used most
frequently in conjunction with the MCDM functions
that support criteria selection, criteria valuation, and pri-
oritization (67% of the total conjunctive use of this map
in the experiment). This finding validated our intended
use of the map: i.e., the bar map was included in Spatial
Group Software to help support decision criteria selec-
tion and the elicitation of criterion weights. The other
three maps (site location map, situation map, and ortho-
photo) were used most often in conjunction with the
tabular output of ranked sites (71% of the total conjunc-
tive use of these maps). This finding tells us that general-
purpose maps are used predominantly to visualize the lo-
cations of decision alternatives and can potentially be
used to evaluate tradeoffs among these alternatives.
The combined use of two maps intended for the pre-
sentation of site evaluation results—graduated circle
rank map and previous rank map—comprised less than
3
4.4% of the time (conjunctive map use). In the remain-
der of cases (65.6%), maps were used independently from
MCDM aids (disjunctive map use).
The four types of map most frequently used were the
bar map, site location map, situation map, and ortho-
2
% of the total map use. This was surprising, because
these maps were designed to combine rank information
from the multiple criteria tables with the site locations.
Furthermore, a third map, called the consensus rank
map, was intended to combine the results of consensus
voting, which we thought would provide significant in-
formation for the participants, but its use comprised only
about 3% of the map moves. Perhaps the groups did not
need to make use of rank and consensus maps very often
in their conversation because of the “rich information
display.” Alternatively, perhaps a basic map showing site
locations together with a list of ranked sites is easier to
interpret than a map that includes both the location and
the rank. The more complex maps, such as the graduated
circle rank map, previous rank map, and consensus rank
map, might have been harder to use for those not accus-
Figure 3. Mean of map moves and MCDM moves by experimental
task.

GIS-Supported Collaborative Decision Making
61
tomed to using specialized maps. Another alternative ex-
planation for this finding might be that individuals used
these maps at their own workstations, but the group as a
whole did not have the facilitator project the maps on
the public display. Since we coded “group attention”—
that is, what was on the public display—we do not have
data on individual map use at each workstation. Either
keystroke logging or a minimum of three more video
cameras would have been needed to capture and code in-
dividual map use.
sensitivity analysis, and affecting the position of the
group on the final ranking of decision alternatives. The
questions of how to improve the existing maps and which
direction should be taken in the design of new types of
maps and visualization aids are open ones. We offer some
suggestions in this matter in our conclusions below.
Group Member Experience and Its Influence
on Decision Aid Appropriation
The hypotheses of the second category, following from
premise 3, focus on the relationship between prior knowl-
edge and acquired experience (the latter during group de-
cision making sessions) in relation to the frequency of use
of maps and decision models. Previous experiments with
nonspatial group decision support systems found a “learn-
ing effect,” in that the frequency and effectiveness of de-
cision aid appropriations increased with prior experience
with computer-supported group decision making pro-
cesses (Chun and Park 1998). In addition, previous stud-
ies report a learning effect in the use of decision aids that
results from frequent participation in a single computer-
supported group decision making task (Chun and Park
1998). However, the transferability of those research re-
sults to collaborative spatial decision making is question-
able because of differences between business decision
problems used in the past research experiments and the
spatial decision problem used in this study.
Hypothesis 2: No relationship exists between spe-
cific MCDM aids used by participants and the number
of map moves accompanying the use of these aids.
The value of Pearson’s correlation coefficient between
the MCDM aids used by the participants and the corre-
sponding map moves (r ꢀ ꢁ0.048, p ꢀ 0.213) was not
statistically significant; hence, we accept the null hy-
pothesis of a lack of a relationship. To determine
whether the use of specific MCDM aids had any effect on
the interaction between the number of MCDM moves
and map moves, we used a GLM statistical procedure.
The dependent variable was MCDM moves, the covari-
ate was map moves, and the factor was MCDM aids.
MCDM aids were encoded by 12 categories. These cate-
gories included 8 aids as described above under the capa-
bilities of Geo Choice module, 3 combinations of these
aids, and the category “None” to represent the absence of
using MCDM aids at a given observation interval. The
model explains 58% of the variability of MCDM moves
Hypothesis 3: No relationship exists between the
frequency of decision aid moves—both MCDM aids
and maps—and the prior knowledge of/experience
with decision support aids of group members. To address
this hypothesis, we classified all participants into four
levels of prior knowledge/experience. The classification
was based on participants’ responses to a background sur-
vey questionnaire completed by each one prior to the ex-
periment. The questions included “What is your experi-
ence with using GIS-generated maps?” and “What is
your experience with multiple criteria decision models?”
The responses were summarized using the frequencies of
rank codes corresponding to “do not know,” “heard
about,” “used some,” and “used frequently.” We used an
analysis of variance (ANOVA) procedure to analyze the
differences in the means of the variable representing
user’s knowledge/experience. The dependent variable
was the frequency of decision aid use and the inde-
pendent variable was the level of user knowledge/
experience. The differences in the means were not sig-
nificant for all levels of knowledge/experience (F ꢀ 0.742,
p ꢀ 0.531); hence, we accept the null hypothesis.
2
(
adjusted R ꢀ .575, F ꢀ 26.714, p ꢀ 0.000); however,
almost all of it is explained by the category of MCDM aid
used (F ꢀ 26.792, p ꢀ 0.000). The covariate—i.e., map
moves—is insignificant (F ꢀ 0.011, p ꢀ 0.917) and does
not contribute to explaining the variability of MCDM
moves.
By far the most map moves were invoked indepen-
dently of any MCDM aid. There were three multiple cri-
teria decision model aids that attracted a higher number
of map moves than others: MCDM Window, Select
Alternatives, and Select Criteria. Combined with the
frequency of map use (see Figure 4), this suggests that
specialized thematic maps (graduated symbol maps), ref-
erence maps representing the distributions of auxiliary
data themes (other than the decision criteria), and high
resolution areal images (orthophotos) can be useful in se-
lecting evaluation criteria and evaluation tradeoffs
among the sites options.
Discussion of findings for Hypothesis 2. There is little
or no evidence to support an assertion that maps in-
cluded in Spatial Group Choice software were effective
in prioritizing evaluation criteria, displaying the results of
Discussion of findings for Hypothesis 3. The result

6
2
Jankowski and Nyerges
might be explained in several ways. Perhaps we did not
have a sufficient range or data to distinguish participant
expertise. Perhaps, unexpectedly, the use of decision sup-
port aids in the experiment (maps and simple analytical
evaluation techniques, at least) is independent of user
experience with decision support aids. That is, maybe a
certain minimum level of computer competency is all
that is needed to make use of such aids. Another issue
that may have influenced the result is educational back-
ground of an average computer user. Our participants
came from different academic fields, but at a minimum
they had completed one year of university-level studies.
All participants were computer-literate, and over one
half of them made use of a computer six or more hours
per week.
Hypothesis 4: Knowledge and competency in using
computerized decision aids, acquired through partici-
pation in the group decision sessions, do not promote
more effective use of decision aids. We used as a mea-
sure of effectiveness the frequency of decision aid moves
Figure 5. Mean number of decision aid moves per decision session.
(
MCDM aids and maps), in which higher effectiveness
of using decision aids is characterized by fewer moves to
access decision aid functions. To test this hypothesis,
we performed a one-way ANOVA. The frequency of
decision aid moves was the dependent variable and
session number (SESSION) was the factor. The differ-
ence of means was significant for one pair only, i.e., the
first session (1) and last session (5), respectively (see
Table 2). Therefore, we reject the null hypothesis for
that pair only. Remember that “session” is the sequence
order 1, . . . , 5 of the randomized tasks 1–5; hence, the
complexity of tasks 1–5 is not at issue. The bar graph of
the mean frequency of decision aid moves per session
Discussion of findings for Hypothesis 4. Although we
did find a learning effect, we are skeptical of the results
due to the limited number of sessions. More sessions in
sequence and better measurements of effectiveness can
provide data for a better test of the learning effect. None-
theless, a recognizable diminishing number of decision
aid moves was observed across the experimental sessions.
Decision Support Aids in Relation
to Decision Phases
The hypotheses of the third category, following from
premise 4, focus on relationships between the use of de-
cision support aids, both maps and MCDM, and the
phase of alternatives evaluation in the group decision
process. Alternatives evaluation is the last phase of the
three-step public participatory decision making process
described above. During values-criteria development,
stakeholders articulate what is important to them in a
decision situation. During alternatives assessment, ex-
perts assign data values to alternatives that best charac-
terize those alternatives in the context of the decision
situation. During alternatives evaluation, stakeholders,
the public, and/or experts identify what criteria are im-
portant and examine the tradeoffs among alternatives to
prioritize them for a recommended solution to a decision
situation. The Renn and Schneider studies (Renn et al.
1993; Schneider, Oppermann, and Renn 1998) reported
only the results of noncomputer-supported decision
processes. Consequently, little is known about how
(
see Figure 5) shows a trend hinting at a learning effect.
The effectiveness of using decision aids increases with
the number of decision sessions. This observation is sup-
ported by a weak but significant negative correlation be-
tween the frequency of decision aid moves and session
number (Pearson’s r ꢀ ꢁ0.3, p ꢀ 0.023).
Table 2. Results of ANOVA
for Session 1 and Sessions 2, 3, 4, and 5
Mean
Difference
Std.
p
(I) SESSION (J) SESSION
(I ꢁ J)
Error 95%
1
2
3
4
5
.7821
2.6154
2.6923
3.9154*
1.845 .673
1.939 .183
1.808 .142
1.939 .048
*
The mean difference is significant at the .05 level.

GIS-Supported Collaborative Decision Making
63
computer-supported groups might use aids at each phase
Table 3. ANOVA of Map Moves by Decision Phases
of the decision process.
Mean
Our experiment implemented the alternatives evalua-
tion phase of a participatory decision task as presented by
Renn et al. (1993). Within the alternatives evaluation
phase, we would expect to see four activities—perhaps
better thought of as microphases—undertaken, as articu-
lated by Simon (1977, 1979): intelligence (gather infor-
mation), design (organize the information), choice (se-
lect alternatives), and review (reassess what was done).
Consequently, we coded the decision phases based on
the following activities:
Difference
(I ꢁ J)
(I) Dec Phase (J) Dec Phase
Std. Error
p
EA/SA
CI/CV/CP
FUNCS
NONE
PE
1.6486
1.9730
1.6157
1.6757
.227
.227
.228
.227
0.000
0.000
0.000
0.000
Note: Only the significant differences are presented in the table.
variable was map moves and the factor was decision
phase. The decision phases included 1) function struc-
turing; 2) problem exploration; 3) criteria identification,
valuation, and prioritization; and 4) evaluation and se-
lection of alternatives. We also included category “none”
to represent the lack of activity. The differences in map
move means were statistically significant for evaluation/
selection of decision alternatives, i.e., EA/SA only in
comparison to other phases (see Table 3). Thus, only in the
case of evaluation/selection of decision alternatives can
we reject the null hypothesis and accept the alternative
hypothesis suggesting significant differences in the mean
of map moves with respect to decision phases.
Discussion of findings for Hypothesis 5. The finding
highlights an observation that groups used maps predom-
inantly to visualize the evaluation results and much less
to structure/design the decision problem. The question
then arises: were the maps provided in Spatial Group
Choice simply not adequate for problem exploration, cri-
teria identification, valuation, and prioritization? Based
on the analysis of variance, maps implemented in Spatial
Group Choice played only a limited support role in the
decision stages of the experiment. In an attempt to find
an answer we examined the frequencies of map use by de-
cision phase. We discovered that the two decision phases
in which group participants used maps most frequently
were evaluation and selection of alternatives (EA/SA)
and identification, valuation and prioritization of criteria
FUNCS (Function Structuring): Group and/or facili-
tator activity with a focus on “what’s next?” and
“how do we want to do this?” with respect to which
decision function to enter. This would include pe-
riods in which the group develops possible ap-
proaches to navigating through the decision func-
tions or to deciding which function should be
visited next.
PE (Problem Exploration): Periods in which the
group attempts to gain a better understanding of
the overall problem. The focus of the discussion is
on learning about or investigating, the problem.
CI (Criteria Identification): Group and/or facilitator
activity with a focus on identifying, discussing, and
selecting criteria that are potentially important for
the decision problem.
CV (Criteria Valuation): Group and/or facilitator ac-
tivity with a focus on selecting the preferred
method of valuing the criteria.
CP (Criteria Prioritization): Group and/or facilitator
activity with a focus on differentiating the criteria
to determine the relative importance of the criteria
for the decision problem, or to gain a better under-
standing of individual priorities.
EA (Evaluate Alternatives): Group and/or facilitator
activity with a focus on comparing and contrasting
the alternatives.
SA (Select Alternatives): Group and/or facilitator
activity involving statements about the group’s ac-
tions that explicitly refer to the selection of alter-
natives, voting, or the reduction of the set of
alternatives for the decision selection.
(
CI/CV/CP; see Figure 6). The three maps used most fre-
quently were bar map (BM) representing attribute val-
ues, site situation map (SM), and orthophoto image
(
OI). The high frequency of bar map use, especially in
EA/SA and CI/CV/CP decision phases, demonstrates
that participants found this specific type of multivariate
graduated symbol map useful in elucidating tradeoffs
among criterion priorities and in illuminating the rank-
ing of decision alternatives. A bar map allows the user to
visualize both quantitative relationships among the mul-
tiple decision criteria and their spatial distribution (Jan-
kowski et al. 1997). The participants found this type of
visualization helpful in making judgments about the rel-
Hypothesis 5: No difference exists in map moves
with respect to decision phases. We were interested to
find out if the frequency of map appropriations differed in
different phases of the decision process within the con-
straints of the predefined task. We tested this hypothesis
using a one-way ANOVA procedure. The dependent

6
4
Jankowski and Nyerges
Figure 6. Frequency of map moves by decision phase.
Figure 7. Mean of map moves in the two session halves of the
experiment.
ative importance of decision criteria. We posit that the
relative judgments about the importance of decision cri-
teria are influenced by spatial relationships among the
decision criterion values, and the visualizations of crite-
rion outcomes in geographic space help elucidate these
judgments. The high frequency of map moves for situa-
tion maps and orthophoto images, especially during eval-
uation/selection of decision alternatives, shows the use-
fulness of the general reference situation map and the
very realistic orthophoto image in these areas.
ally, the comparison of means of MCDM moves across
the tasks confirmed that MCDM aids were used with
similar frequency in both session halves.
Discussion of findings for Hypotheses 6 and 7. In contrast
to the mean of MCDM moves, regardless of task com-
plexity, the mean of map moves markedly differs between
the two session halves for each task. We were surprised to
find that the participants used MCDM aids without much
difference between the two halves of the experiment in
the frequency of moves. The Spatial Group Choice soft-
ware used during the experiment offered the participants
both analytical functions useful for alternative evaluation
and exploratory/visualization functions that could poten-
tially be used during the problem exploration phase. We
speculated that the first, more “exploratory” half of the
experiment would be marked by more frequent use of
maps than the second half—but only because of anec-
dotal evidence about maps as exploratory aids in various
literature. The much less frequent use of maps during the
first half of experimental sessions indicates the possible
need for new types of maps, such as those charting the lin-
eage of collaborative discourse and problem structuration,
as in a spatial understanding and decision support system
(Jankowski and Stasik 1997; Moore 1997). However, it
also points out the need for analytical visualization aids
that might help in exploratory decision situations.
Hypothesis 6: No difference exists between session
halves in the frequency of using map aids. We tested
this hypothesis with a paired-samples T-test statistical
procedure. The results of the T test, with the first half
mean ꢀ 0.62 and the second half mean ꢀ 1.84 (t ꢀ
ꢁ4.570, p ꢀ 0.000), allow us to reject the null hypothe-
sis and indicate a difference in the pattern of map use be-
tween the two session halves. In all tasks, maps were used
much more frequently during the second, more analyti-
cal half of the experiment than during the first explor-
atory half (see Figure 7).
Hypothesis 7: No difference exists between session
halves in the frequency of using MCDM aids. We ex-
pected to see less use of MCDM (analytical) aids during
the first half of the session than during the second half,
since the first half of each experimental session was ex-
pected to involve more exploratory discussion than anal-
ysis. We used a paired-samples T-test statistical proce-
dure to test the hypothesis. The results of the T test were
not significant (t ꢀ ꢁ0.427, p ꢀ 0.670), thus validating
the null hypothesis and indicating a lack of difference
between the session halves in MCDM moves. Addition-
Group Conflict by Decision Phase
The hypotheses of the fourth category, following from
premises 2 and 4, use as a point of departure the results of

GIS-Supported Collaborative Decision Making
65
Table 4. Group Work Conflict Code and
length of time during which the participants were en-
gaged in conflictual activity as coded using the group
work conflict (GWC) code within the group working re-
lations coding system (Nyerges, Moore, Montejano, and
Compton 1998). This code included several types of
group work conflict (see Table 4).
Its Interpretation Source
Group Work Conflict—periods when the group is organized and
working together, but members disagree with each other or express
a different perspective on the topic. This may include attempts by
group members to manage or diffuse the conflict (as listed below),
or periods when the group is in disagreement because of a
misunderstanding.
Hypothesis 8: No difference exists in the level of
conflict by decision phase. To test this hypothesis we
used a one-way ANOVA in which the dependent variable
was conflict level (measured by the length of time the
groups were engaged in conflict situations) and the factor
was decision phase. The results revealed an interesting di-
chotomy between the “intelligence gathering” phases
(FUNCS and PE) and the “criteria or alternative evalua-
tion” phases (CI/CV/CP and EA/SA; see Table 5). Both
types of phases are internally homogeneous in terms of the
conflict level, meaning that the differences within the in-
telligence gathering and evaluation phases are not signifi-
cant. However, statistically significant differences in con-
flict level appear between the decision phase groupings.
Discussion of findings for Hypothesis 8. The finding in-
dicates that different phases of computer-supported
group decision making have different levels of conflict,
an analytical-detail phase characterized by high conflict
level and an exploratory-structuring phase characterized
by low conflict level. Perhaps there was less conflict dur-
ing problem exploration because interests and values
were not at odds with each other. There was more con-
flict during criteria selection and alternative evaluation
because specific interests surfaced. The higher level of
conflict during evaluation phase tells us that analytical
decision aids aimed at conflict management are likely to
help move through conflict; such conflict now being rec-
•
•
•
•
•
•
Opposition: periods in which disagreements are expressed
through the formation of opposing sides
Accommodation: a mode of resolution of opposition in which
one side gives in
Tabling: a mode of resolution of opposition in which the
subject is tabled or dropped
Negotiation: a mode of resolution in which the group
negotiates to manage conflict
Compromise: a mode of conflict resolution in which the group
compromises
Justification: periods when a supporting rationale for a
particular position is posed to the group for consideration
Note: The codes were extracted from the coding guide developed within
the project and reported in Nyerges et al. (1998a).
many previous studies on group decision support systems,
studies which analyzed the effect of computerized group
decision support on conflict management (Chun and
Park 1998). The findings were based on comparisons be-
tween nonsupported and supported groups. Based on
these studies, there is strong evidence that electronic
mail and anonymous voting, prominent features of group
decision support systems, reduce conflict in collaborative
decision making. Since we did not have control groups
solving the decision problem without computerized deci-
sion support (i.e., nonsupported groups), we decided to
analyze the relationship between decision phase and
group conflict. We measured the level of conflict by the
Table 5. Results of ANOVA for Conflict Level by Decision Phase
9
5% Confidence Interval
Mean Difference
(I) Decision Phase (J) Decision Phase
(I ꢁ J)
Std. Error
p*
Lower Bound
Upper Bound
CI/CV/CP
EA/SA
FUNCS
PE
EA/SA
FUNCS
PE
ꢁ.36
2.31
2.35
.375
.375
.375
.766
.000*
.000*
ꢁ1.33
1.35
1.39
.60
3.28
3.32
CI/CV/CP
FUNCS
PE
.36
2.68
2.72
.375
.375
.375
.766
.000*
.000*
ꢁ.60
1.71
1.75
1.33
3.64
3.68
CI/CV/CP
EA/SA
PE
ꢁ2.31
ꢁ2.68
4.05E-02
.375
.375
.375
.000*
.000*
1.000
ꢁ3.28
ꢁ3.64
ꢁ.92
ꢁ1.35
ꢁ1.71
1.01
CI/CV/CP
EA/SA
FUNCS
ꢁ2.35
ꢁ2.72
ꢁ4.05E-02
.375
.375
.375
.000*
.000*
1.000
ꢁ3.32
ꢁ3.68
ꢁ1.01
ꢁ1.39
ꢁ1.75
.92
*
The mean difference is significant at the .05 level.

6
6
Jankowski and Nyerges
ognized as a necessary part of making progress in envi-
ronmental disputes (Simosi and Allen 1998). This is an
important finding for future designs of collaborative spa-
tial decision support software.
rather than use individual work to examine alternatives
before posing suggestions.
Conclusion
Hypothesis 9: No difference exists in the level of
conflict between task 1 (least complex) and task 4
Our research helps fill a gap in knowledge about group
use of geographic information systems applied to value-
laden, public-private problems. The question of testing
differences in information use versus software use often
arises. To address this concern, we chose a “classic” site
selection preference problem that dealt with emerging
concern about habitat restoration in the context of
nature-society debates. The geographic information used
in the experiment was adopted from the realistic Duwa-
mish Waterway decision task. The information was con-
veyed to participants through somewhat common map
and decision table information structures (Nyerges
1991), and those information structures were imple-
mented in GIS software. We purposely narrowed the in-
formation structures to those two to be able to address
differences in the classic “tables” and “maps” concern:
which is better for what activity? The information struc-
tures are general enough in nature that any GIS could
implement them. Thus, we believe our findings address
differences in information use as much as differences in
software use, the two being intimately connected.
The findings reported here broaden and deepen our
understanding of the conceptual underpinnings of GIS-
supported collaborative decision making by way of em-
pirical analysis linked to a theoretical base. Examining
the premises and their respective hypotheses in EAST
provides a way of systematically addressing research ques-
tions and subsequent hypotheses about the dynamics of
geographic information (system) use in groups. Address-
ing research questions using an experimental design pro-
vides a foundation for examining details about decision
process dynamics involving geographic decision aids.
The existing research literature encouraged us to focus
on decision process rather than decision outcomes. Re-
searchers working on decision support topics in the man-
agement sciences suggest that process-based research is
more meaningful than outcomes-based research in a lab-
oratory setting, since the latter fosters little ownership of
results (Rohrbaugh 1989; Todd 1995). From a case study
perspective, Yin (1994) recommends a process-based ap-
proach for research contexts, even for field studies, and
an outcomes-based approach for program evaluation
contexts. Because outcomes in public (group) problems
are often difficult to predict, Renn, Webler and Wiede-
mann (1995) and Wood and Gray (1991) also recom-
mend process-oriented research.
(
most complex). To test this hypothesis, we used a one-
way ANOVA in which the dependent variable was con-
flict level (measured by the length of time the groups were
engaged in conflict situations) and the factor was task
complexity (tasks 1 and 4). The results showed the lack of
statistically significant difference between the group con-
flict in tasks 1 and 4; hence, we accepted the null hypoth-
esis. We then modified our hypothesis to include only
tasks 4 and 5, assuming that the level of conflict between
these tasks is different because of the difference in the or-
ganization of the collaborative decision making process
(access to individual workstation and the presence of fa-
cilitator in task 4 versus the access to public screen only
and the presence of facilitator in task 5). The results of
the ANOVA indicate a statistically significant difference
in conflict level between tasks 4 and 5 (F ꢀ 5.504, p ꢀ
0.02); hence, we rejected the modified null hypothesis.
Discussion of findings for Hypothesis 9. Our findings
that task complexity was not associated with the level of
conflict between task 1 (simpler) and task 4 (more com-
plicated) are somewhat contrary to current literature.
Renn, Webler, and Wiedemann (1995) present a dia-
gram indicating that high levels of complexity are associ-
ated with higher levels of environmental conflict. How-
ever, their diagram also takes into consideration higher
conflict being associated with a difference among three
levels: knowledge level, experience and trust level, and
worldviews and values level. Since all participants vol-
unteered to take part in this experiment, and the experi-
ment was situated in an environmental context, it is
likely that most participants had similar environmental
perspectives, even though we asked each participant to
choose from four different stakeholder perspectives. Per-
haps this finding results from the influence of a con-
trived, experimental setting versus a more realistic set-
ting. A review of the participant database indicates that
2
8% of all participants chose to be aligned with an envi-
ronmental stakeholder perspective. The results of the
difference between task 4 and task 5 might be explained
by the fact that people are more likely to voice conflict
when they are talking to each other than when they are
not talking to each other. Since task 5 probably involved
more “public discussion” than task 4, given its use of only
a public screen, observation of increased conflict was to
be expected; participants had to talk through differences,

GIS-Supported Collaborative Decision Making
67
Findings representative of four categories were devel-
oped in this study. One category of findings demon-
strated that groups used maps predominantly to visualize
the evaluation results and much less to structure/design
the decision problem. Maps played only a limited support
role in various decision stages of the experiment. The
simplicity of the task might have constrained a more cre-
ative role for maps. In the second category of findings,
while the use of multiple criteria decision models by
groups remained steady throughout different phases of
the decision process, the use of maps was much lower
during the initial exploratory-structuring phase than
during the later analytic-integrating phase. This was a bit
surprising, since maps are thought to be idea generators.
However, the simplicity of the task might again have ac-
counted for the difference. In the third category of find-
ings, neither the amount of prior group member experi-
ence with computer tools nor the amount of acquired
group member experience with such tools had any influ-
ence on the appropriation of decision aids. The group
backgrounds of participants were reasonably homoge-
neous in comparison to those of a cross-sample of the
general population. We had very few “spatial profession-
als” with multiple years of experience. In the fourth and
final category of findings, different phases of the decision
process had two different levels of conflict. The explor-
atory-structuring phase was characterized by a lower
level of conflict, whereas the analytic-integrating phase
was characterized by a higher level of conflict. The
higher level of conflict during the analytic-integrating
phase suggests to us that analytical decision aids aimed at
conflict management are likely to help move through
conflict, since such conflict is now recognized as a neces-
sary part of making progress in public decision problems.
Conflict early on indicates that people are likely aware of
each other’s valued interests.
Claims about the validity of our findings are guided
by the considerations of human-computer-human labo-
ratory experiments (Kidder and Judd 1986). With re-
gard to our claims about external validity, the partici-
pant group we used is admittedly a sample from a
student population, with a few community walk-ons.
We did not try to enlist professionals, but relied on
those who were interested to self-select, having interest
in environmental decision making and the time for a
battery of five decision sessions. However, we claim that
we can generalize to other groups that are characterized
by containing a majority of novices in tool use and hab-
itat restoration. We can argue that the participants we
used are likely to use such technology in the future be-
cause of their interests, as evidenced by their volunteer-
ing for the project.
As in most laboratory experiments, we opted to em-
phasize construct validity and internal validity. We were
interested in the influence of task complexity and tech-
nology exposure (construct validity) on the fine-grain
dynamics of human-computer-human interaction as ex-
pressed through relationships between aspects of EAST
(internal validity). The experimental design allowed us
to set up and control tasks; the videotape allowed us to
capture the fine-grain dynamics. Getting professional
analysts to do this is a very difficult endeavor. However,
we promoted external validity by adopting a realistic de-
cision task for participants to address, thereby attracting
student subjects with a sincere interest in environmental
decision tasks. As further evidence of the external valid-
ity of the professional abilities of students, for the past
three years, in the winter quarter offering of a wetlands
restoration class taught by one of the NOAA panelists at
the University of Washington, we have made use of our
database and software to conduct the Duwamish Water-
way habitat restoration decision experiment. Each year,
five groups of five students, each group playing a stake-
holder role consistent with the NOAA panel, have come
together in consensus on choosing two of the top three
sites that have been restored.
To our knowledge, only three major experiments have
been conducted that make use of spatial decision support
technology: one by Reitsma et al. (1996), one by Stasik
(1999), and this one. In terms of comparability across
all experiments, like the GDSS experiments and the
Reitsma et al. (1996) resource negotiation experiment, it
appears that groups would rather have facilitators and/or
chauffeurs help them work through problems. Individu-
als having access to technology in the Reitsma experi-
ment were frustrated with the complexity of the soft-
ware, but in our experiment we did not detect such
frustration. Admittedly, the water resource simulation
model in the Reitsma study was probably more difficult
to understand than our multicriteria models. Ranking
options is an activity people do quite often. In land use
planning experiments conducted by Stasik, the role of
facilitator was partially fulfilled by a computer server,
which managed the collaboration process. The partici-
pants expressed a strong preference for making complex
collaborative spatial decision support software, which
they used, as easy and as familiar to use as standard Web
browser. Many of them essentially wanted the software to
look and act like such a browser. However, regardless of
whether the support comes through software and/or hu-
manware, the advantage of facilitation and technical as-
sistance is undeniable at the current time.
The theoretical approach and findings reported in this
article are intended to fuel the debate about the intended

6
8
Jankowski and Nyerges
and unintended impacts of geographic information tech-
nology on society, particularly in the context of small-
group work. As geographic information technologies dif-
fuse through society, they are being criticized because
they might in fact support a one-sided story. Lake (1987,
ing to prepare the data for analysis for this article.
Thanks go to Victoria Lawson, William Beyers, Richard
Morrill, and Christina Drew for comments on an earlier
draft of the manuscript.
1
41) warns that “. . . ultimately at issue is whether the in-
Notes
tegrative capacity of GIS technology proves robust
enough to encompass not simply more data but funda-
mentally different categories that extend considerably
beyond the ethical, political, and epistemological limita-
tions of positivism.” In a special issue of Cartography and
Geographic Information Systems, Sheppard (1995) and
others echo the same concern, since use of GIS based on
efficiency and effectiveness might result in opposite out-
comes from an equity perspective. Others who are re-
searching public participation GIS, as reported in a more
recent special issue of Cartography and Geographic Infor-
mation Systems, recognize a number of those shortcom-
ings but are suggesting GIS-based solutions that can
broaden the voice of societal participants (Obermeyer
1
. Note that the tool in this case is part of the “substantive” do-
main and not the “methods” domain, since the tool itself is
an aspect of study—an important point recognized by Pickles
(1996).
2
. Not all participants were students; many were community
members who had seen the flyer. Participants received a
voucher for a four-hour training session in ArcView GIS
upon completion of all five sessions. Groups received a prize
of $25 per group member from a random drawing after com-
pleting all five sessions. If students wanted, they received two
credit hours of independent study. Only about 10 out of 109
opted for the credit, so this was not seen as a main motivating
factor for participation.
3. To motivate students to return to subsequent sessions, we
awarded monetary prizes and training only upon completion
of all five sessions. Another motivation factor for many par-
ticipants was the chance to work with GIS in an environ-
mental decision making setting.
1
998). It is important to recognize that “[d]ifferential ac-
cess to GIS data, hardware, software, and ‘humanware’ is
a significant component of the political economy of spa-
tial decision making” (Harris and Weiner 1998, 69).
Regardless of where one stands in this debate, a better
understanding about the impacts of geographic informa-
tion technology use in collaborative decision making set-
tings is needed, whether we are investigating use of GIS,
SDSS, SUSS, SUDSS, or the “umbrella category” of par-
ticipatory geographic information systems. We hope that
our framework and findings motivate others to pursue, or
at least encourage a critical interest in, sociobehavioral
studies about geographic information (system) use.
References
ArcView 2.1. 1995. Environmental Systems Research Institute,
Inc., Redlands, CA.
Armstrong, M. P. 1993. Perspectives on the development of
group decision support systems for locational problem solv-
ing. Geographical Systems 1 (1): 69–81.
Bartsch, C., and E. Collaton. 1997. Brownfields: Cleaning and re-
using contaminated property. Westport, CT: Praeger.
Bhargarva, H. K., R. Krishnan, and A. D. Whinston. 1994. On
integrating collaboration and decision analysis techniques.
Journal of Organizational Computing 4 (3): 297–316.
Brinberg, D., and J. McGrath. 1985. Validity and the research pro-
cess. Thousand Oaks, CA: Sage Publishers.
Acknowledgments
Brunell, D. 1999. Not just any salmon plan will do the job. East-
side Business Journal 5 (6): 23.
The co-authors shared equally in the research effort
reported herein. Their research has been supported in
part by National Science Foundation Grant No. SBR-
Carver, S. 1991. Integrating multicriteria evaluation with GIS.
International Journal of Geographical Information Systems
5
:321–39.
Chun, K. J., and H. K. Park 1998. Examining the conflicting
results of GDSS research. Information & Management 33:
313–25.
Coleman, D., and R. Khanna. 1995. Groupware: Technologies
and applications. Upper Saddle River, NJ: Prentice Hall.
Couclelis, H., and M. Monmonier. 1995. Using SUSS to resolve
NIMBY: How spatial understanding support systems can
help with the “not in my back yard” syndrome. Geographical
Systems 2 (2): 83–101.
9
411021, funded jointly by the Geography and Regional
Science Program and the Decision, Risk and Manage-
ment Science Program. The authors acknowledge the
helpful input of members of the Elliott Bay/Duwamish
Restoration Panel, especially Robert Clark, Jr., and Jen-
nifer Stegers of NOAA Restoration Northwest, Curtis
Tanner of the U.S. Fish and Wildlife Service, Glen St.
Amant of the Muckleshoot Tribe, and Bob Matsuda of
King County Department of Metropolitan Services, who
provided a critique of the Spatial Group Choice software
at an early stage of development. T. J. Moore helped tre-
mendously as a research assistant facilitator and in help-
Cowen, D. 1988. GIS versus CAD versus DBMS: What are the
differences? Photogrammetric Engineering & Remote Sensing
5
4 (11): 1551–55.
Crowfoot, J. E., and J. M. Wondolleck. 1990. Environmental dis-
putes: Community involvement in conflict resolution. Wash-
ington, DC: Island Press.

GIS-Supported Collaborative Decision Making
69
Davis, T., and K. Margolis, eds. 1997. Brownfields: A comprehen-
sive guide to redeveloping contaminated property. Chicago:
American Bar Association.
Densham, P. J. 1991. Spatial decision support systems. In Geo-
graphical information systems: Principles and applications, ed-
ited by D. J. Maguire, M. F. Goodchild, and D. W. Rhind.
New York: John Wiley and Sons.
Harris, T., and D. Weiner. 1998. Empowerment, marginaliza-
tion, and “community-integrated” GIS. Cartography and
Geographic Information Systems 25 (2): 67–76.
Heywood, D. I., J. Oliver, and S. Tomlinson. 1995. Building an
exploratory multicriteria modeling environment for spatial
decision support. In Innovations in GIS 2, edited by P.
Fisher. London: Taylor and Francis.
Densham, P. J., M. P. Armstrong, and K. Kemp. 1995. Report
from the specialist meeting on collaborative spatial deci-
sion making, Initiative 17, National Center for Geographic
Information and Analysis, University of California, Santa
Barbara, 17–21 September.
Densham, P. J., and M. F. Goodchild. 1989. Spatial decision sup-
port systems: A research agenda. In Proceedings, GIS/LIS
Hwang, C.-L., and M.-J. Lin. 1987. Group decision making under
multiple criteria. Berlin: Springer-Verlag.
Jankowski, P. 1995. Integrating geographic information systems and
multiple criteria decision making methods. International Jour-
nal of Geographical Information Systems 9 (3): 251–73.
Jankowski, P., A. Lotov, and D. Gusev. 1999. Application of
multiple criteria tradeoff approach to spatial decision mak-
ing. In Spatial multicriteria decision making and analysis: A
geographic information sciences approach, edited by Jean-
Claude Thill. Aldershot: Ashgate.
Jankowski, P., T. L. Nyerges, A. Smith, T. J. Moore, and E. Hor-
vath. 1997. Spatial Group Choice: a SDSS tool for collab-
orative spatial decision-making. International Journal of
Geographic Information Systems 11 (6): 577–602.
Jankowski, P., and M. Stasik. 1997. Design consideration for space
and time distributed collaborative spatial decision making.
Journal of Geographic Information and Decision Analysis 1
(1):1–8. http://publish.uwo.ca/~jmalczew/gida.htm (14 Feb-
ruary 2001).
Jessup, L., and J. Valacich, eds. 1993. Group support systems: New
perspectives. New York: Macmillan Publishing Company.
Kidder, L., and C. M. Judd. 1986. Research methods in social rela-
tions. 5th ed. New York: Holt, Rinehart, and Winston.
Lake, R. W., ed. 1987. Resolving locational conflict. New Bruns-
wick, NJ: Rutgers Center for Urban Policy Research.
Malczewski, J. 1999. GIS and multicriteria decision analysis. New
York: John Wiley and Sons.
’
89, Orlando, FL. Vol. 2. Bethesda, MD: American Con-
gress on Surveying and Mapping.
DeSanctis, G., and R. B. Gallupe. 1987. A foundation for the
study of group decision support systems. Management Sci-
ence 33:589–609.
DeSanctis, G., and M. S. Poole. 1994. Capturing the complexity
in advanced technology use: Adaptive structuration theory.
Organization Science 5 (2): 121–47.
Duffy, D. M., M. Roseland, and T. I. Gunton. 1996. A prelimi-
nary assessment of shared decision making in land use and
natural resource planning. Environments 23 (2): 1–16.
Eastman, J. R., W. Jin, P. A. K. Kyem, and J. Toledano. 1995. Raster
procedures for multicriteria/multiobjective decisions. Photo-
grammetric Engineering & Remote Sensing 61 (5): 539–47.
Faber, B., J. Knutson, R. Watts, W. Wallace, J. Hautalouma, and
L. Wallace. 1994. A groupware-enabled GIS. Proceedings,
GIS ’94 Symposium. Fort Collins, CO: GIS World, Inc.
Faber, B., W. Wallace, and J. Cuthbertson. 1995. Advances in
collaborative GIS for land resource negotiation. Proceed-
ings, GIS ’95. Vol. 1. Ninth Annual Symposium on Geo-
graphic Information Systems, Vancouver, B.C. March. Fort
Collins, CO: GIS World, Inc.
McGrath, J. E. 1984. Groups: Interaction and performance. Engle-
wood Cliffs, NJ: Prentice-Hall Inc.
Faber, B. G., W. Wallace, and R. M. P. Miller. 1996. Collabora-
tive modeling for environmental decision making. Pro-
ceedings, GIS ’96. Symposium, Vancouver, B.C., March.
CD-ROM.
Girden, E. R. 1992. ANOVA: Repeated measures. Newbury Park,
CA: Sage Publications.
Godschalk, D. R., G. McMahon, A. Kaplan, and W. Qin. 1992.
Using GIS for computer-assisted dispute resolution. Photo-
grammetric Engineering & Remote Sensing 58 (8): 1209–12.
Golay, F., and T. Nyerges. 1995. Understanding collaborative
use of GIS through social cognition. In Cognitive aspects of
human-computer interaction for geographic information sys-
tems, edited by T. Nyerges, D. M. Mark, R. Laurini, and M.
Egenhofer. Proceedings of the NATO ARW, Mallorca,
Spain, 21–25 March. Dordrecht, Netherlands: Kluwer
Academic Publishers.
Gray, P., S. L. Alter, G. DeSanctis, G. W. Dickson, R. Johansen,
K. L. Kraemer, L. Olfman, and D. R. Vogel. 1992. Group
decision support systems. In Information systems and decision
processes, edited by E. A. Stohr and B. R. Konsynski. Los
Alimitos, CA: IEEE Press.
Moore, T. J. 1997. GIS, society, and decisions: A new direction
with SUDSS in command. Proceedings, AutoCarto 13. Vol.
5. American Society for Photogrammetry and Remote
Sensing, Seattle, WA. 7–10 April. Bethesda, MD: Ameri-
can Society for Photogrammetry and Remote Sensing.
National Oceanic and Atmospheric Administration (NOAA).
1993. Technical notes from the Elliot Bay/Duwamish restora-
tion program, April 14, 1993. NOAA Restoration Program.
Seattle, WA: Sand Point Office.
Nyerges, T. 1991. Analytical map use. Cartography and Geo-
graphic Information Systems (formerly The American Cartog-
rapher) 18 (1): 11–22.
Nyerges, T, and P. Jankowski. 1997. Enhanced adaptive structu-
ration theory: A theory of GIS-supported collaborative de-
cision making. Geographical Systems 4 (3): 225–59.
Nyerges, T., R. Montejano, C. Oshiro, and M. Dadswell. 1998.
Group-based geographic information systems for transpor-
tation site selection. Transportation Research C: Emerging
Technologies 5 (6): 349–69.
Nyerges, T., T. J. Moore, R. Montejano, and M. Compton. 1998.
Interaction coding systems for studying the use of groupware.
Journal of Human-Computer Interaction 13 (2): 127–65.
Obermeyer, N. 1998. The evolution of public participation GIS.
Cartography and Geographic Information Systems 25 (2): 65–66.
Obermeyer, N., and J. Pinto. 1994. Managing geographic informa-
tion systems. New York: Guilford.
Gray, P., D. Vogel, and R. Beauclair 1990. Assessing GDSS em-
pirical research. European Journal of Operational Research
4
6:162–76.
Gregory, R. 1999. Identifying environmental values. In Tools to
aid environmental decision making, edited by V. H. Dale and
M. R. English. New York: Springer-Verlag.
Orlikowski, W. J. 1992. The duality of technology: Rethinking

7
0
Jankowski and Nyerges
the concept of technology in organizations. Organization
Science 3 (3): 398–427.
Parenteau, R. 1988. Public participation in environmental decision
making. Ottawa: Federal Environmental Assessment Re-
view Office.
Pickles, J. 1996. Tool or science? GIS, technoscience, and the
theoretical turn. Annals of the Association of American Geog-
raphers 87 (2): 363–72.
Shiffer, M. J. 1992. Towards a collaborative planning system. En-
vironment and Planning B 19 (6): 709–22.
Silvey, R., and V. Lawson. 1999. Placing the migrant. Annals of
the Association of American Geographers 89 (1): 121–32.
Simon, H. 1977. The new science of management decision. 3rd edi-
tion. Englewood Cliffs, NJ: Prentice-Hall.
———. 1979. Rational decision making in business organiza-
tions. American Economic Review 69:493–513.
Poole, M. S. 1985. Tasks and interaction sequences: A theory of
coherence in group decision-making interaction. In Se-
quence and pattern in communicative behaviour, edited by R. L.
Street, Jr., and J. N. Cappella. London: Edward Arnold.
Poole, M. S., and J. Roth. 1989. Decision development in small
groups IV: A typology of group decision paths. Human
Communication Research 15:3, 323–56.
Simon, H., G. Dantzig, R. Hogarth, Ch. Plott, H. Raiffa, T.
Schelling, K. Shepsle, R. Thaler, A. Tversky, and S. Winter.
1992. Decision making and problem solving. In Decision
making: Alternatives to rational choice models, edited by M. Zey.
Newbury Park, CA: Sage Publications.
Simosi, M., and P. T. Allen. 1998. Public perception of risk man-
agement in environmental controversies: A U.K. case
study. Risk: Health, Safety and Environment 9 (4): 309–27.
Smith, L. G. 1982. Alternative mechanisms for public participa-
tion in environmental policy making. Environments 14 (3):
21–34.
Popper, F. J. 1981. Siting LULUs. Planning 47:12.
Reitsma, R. 1996. Structure and support of water-resources man-
agement and decision making. Journal of Hydrology 177:
2
53–68.
Reitsma, R., I. Zigurs, C. Lewis, V. Wilson, and A. Sloane. 1996.
Experiment with simulation models in water-resources ne-
gotiation. Journal of Water Resources Planning and Manage-
ment 122 (1): 64–70.
Renn, O., T. Webler, H. Rakel, P. Dienel, and B. Johnson. 1993.
Public participation in decision making: A three-step pro-
cedure. Policy Sciences 26:189–214.
Renn, O., T. Webler, and P. Wiedemann. 1995. Fairness and com-
petence in citizen participation: Evaluating models for environ-
mental discourse. Dordrecht: Kluwer Academic Publishers.
Rittel, H., and M. M. Webber. 1973. Dilemmas in a general the-
ory of planning. Policy Sciences 4:155–69.
Rohrbaugh, J. 1989. Demonstration experiments in field set-
tings: Accessing the process, not the outcome, of group de-
cision support. In The information systems research challenge:
Experimental research methods, edited by I. Benbasat. Cam-
bridge: Harvard Business School.
SPSS Base 8.0. 1998. Applications Guide. SPSS Inc., Chicago, IL.
Stasik, M. I. 1999. Collaborative planning and decision making
under distributed space and time conditions. Ph.D. diss.,
University of Idaho.
Strauss, A., and J. Corbin. 1990. Basics of qualitative research.
Newbury Park, CA: Sage Publications.
Susskind, L., and J. Cruikshank. 1987. Breaking the impasse: Con-
sensual approaches to resolving public disputes. New York:
Basic Books.
Susskind, L., and P. Field. 1996. Dealing with an angry public. New
York: The Free Press.
Thill, J.-C., ed. 1999. Spatial multicriteria decision making and
analysis: A geographic information sciences approach. Alder-
shot: Ashgate.
Todd, P. 1995. Process tracing methods in the decision sciences.
In Cognitive aspects of human-computer interaction for geo-
graphic information systems, edited by T. Nyerges, D. Mark,
M. Egenhofer, and R. Laurini. Dordrecht: Kluwer.
Vogel, D. R. 1993. Electronic meeting support. The Environmen-
tal Professional 15 (2): 198–206.
Wood, D. J., and B. Gray. 1991. Toward a comprehensive theory
of collaboration. Journal of Applied Behavioral Sciences 27
(2): 139–62.
Wright, D., M. Goodchild, and J. Proctor. 1996a. GIS tools or
science?: Demystifying the persistent ambiguity of GIS as
“tool” versus “science.” Annals of the Association of Ameri-
can Geographers 87 (2): 346–62.
Roschelle, J., and S. D. Teasley. 1995. The construction of
shared knowledge in collaborative problem solving. In
Computer supported collaborative learning, edited by C.
O’Malley. Berlin: Springer-Verlag.
Sanderson, P. M., and C. Fisher. 1994. Exploratory sequential
data analysis: Foundations. Human-Computer Interaction
9
:251–317.
Sanderson, P., J. Scott, T. Johnston, J. Mainzer, L. Watanabe,
and J. James. 1994. MacSHAPA and the enterprise of ex-
ploratory sequential data analysis (ESDA). International
Journal of Human-Computer Studies 41:633–81.
———. 1996b. Reply: Still hoping to turn that theoretical corner.
Annals of the Association of American Geographers 87(2): 373.
Yin, R. K. 1994. Case study research. 2nd ed. Thousand Oaks,
CA: Sage Publications.
Zey, M., ed. 1992. Decision making: Alternatives to rational choice
models. Newbury Park, CA: Sage Publications.
Schneider, E., B. Oppermann, and O. Renn. 1998. Implement-
ing structured participation for regional level waste man-
agement planning. Risk: Health, Safety and Environment 9
(
4): 379–95.
Sheppard, E. 1995. GIS and society: Towards a research agenda.
Cartography and Geographic Information Systems 22 (1): 5–16.
Correspondence: Department of Geography, University of Idaho, Moscow, ID 83844-3021, email piotrj@uidaho.edu (Jankowski); Department
of Geography, University of Washington, Box 353550, Seattle, WA 98195-3550, email nyerges@u.washington.edu (Nyerges).