-744~
Aids for Customizing Generic
System Dynamics Models:
A Community Development Example
Fahriye H. Sancar
Robert J. Cook
University of Wisconsin-Madison
ABSTRACT
The general considerations for the development and customization of a generic
system dynamics model for rural community development are presented. A
preliminary version of a generic model is presented. Alternative
customization strategies based on problem complexity as perceived by the
community are discussed. The customization involves generating a network
representation of the problem using a modified version of interpretive
structural modeling and a pattern transfer procedure to enhance the generic
model. The entire procedure is based on cognitive criteria to overcome human
information processing limitations; to provide a rational and systematic
approach to the simplification of problem complexity, and to promote a shared
understanding of the problem situation among the participants.
BACKGROUND
The integration of the scientific-rational and the social-consensual
dimensions into planning is a major challenge in community development today.
It is now recognized that social and economic issues should be addressed
together with the issues on environmental and visual-aesthetic quality. The
complex, time-dependent interrelations and interactions among all these issues
should be analyzed by incorporating the knowledge and insights made available
by more than one discipline. Since there are no widely accepted general
theories of socio-economic change and spatial development, community planning
has been largely based on partial analytic models with restrictive assumptions
regarding system behavior. The model outcomes are then interpreted using
expert judgement in making recommendations on the problem issues.
The social-consensual dimension requires that the local capabilities within a
problem context be mobilized to initiate and sustain planning efforts through
genuine participation of the inhabitants and of the public and private
agencies. This requirement provides the community members with means to
exercise control over their environment and facilitate adaptation to growth
and change. It is expected that such a strategy will maintain continuity in
the environment so as to establish a “planning culture" (Skinner 1953),
allowing adaptive responses to occur at the local level (Dunn 1971, Smith
1973).
The underlying theme of a planning approach to respond to both of the above
challenges is the notion of an “appreciative system," consisting of a group,
an institution or an organization charged with policy making. In this context
appreciation involves making judgements about the external and internal state
of the system, and about the significance of these judgements. The
appreciative judgement reflects the views currently held by those who make
them, describes which events and relations they will regard as relevant under
-745-
various conditions, and how they value these. Such judgements disclose what
can best be described as a “readiness” to distinguish certain aspects of a
situation from others and to classify and evaluate them in a particular way.
This state of readiness is described as an “appreciative system". It needs to
be learned much like other perceptual schemata (Vickers 1965).
The perception of a problem situation in the community development context may
take the form of internalized or externalized models. Appreciation, however,
refers to the achievement of a special kind of perception which is 1) shared
by the planning group - including the experts and lay people, 2) reflects
those characteristics of the problem situation perceived to be important by
them, and 3) represents the collective knowledge about the situation rather
than only those deemed “necessary and sufficient" by the experts. Therefore,
the formulation of appreciative judgement starts with each individual's mental
model, and ends with an external model containing the facts and values which
bear on the decisions. The processes leading to the appreciation are
cognitive and deal with perception and integration of information at the
individual and group levels.
In this paper we describe the main components of a decision support system for
community development which enhances the appreciative capabilities of the
policy making groups by aiding them in the effective use of available expert
knowledge and integration of relevant information to analyze their own
problems, to formulate policy options and to evaluate available alternatives
for the resolution of these problems. These components consist of 1) a
generic community development model, 2) a problem structuring algorithm, and
3) a situational system dynamics model. In the following sections, the larger
procedural framework will be discussed first. Then the details of the three
components will be described.
THE METHODOLOGY
The three components of the decision support system are based on a larger
procedural framework developed by Checkland (1981) and referred to as a
generalized “soft" systems methodology for resolving unstructured problems.
This methodology is a result of the reflective approach taken by Checkland in
his industrial engineering problem-solving practice, and basically represents
the refinements of departure from the “hard" systems thinking which was widely
employed in the late 1960s and in the 1970s. Figure 1 is taken from Checkland
(1981; p. 163) and is modified to serve as the basis to develop the specifics
of an approach to community problem solving. The methodology contains two
Kinds of activity. The stages above the dotted line in Figure 1 are the
“real-world" activities. Those below the dotted line are “systems thinking"
activities. It is by these two activities that the real-world complexity is
structured. In this study the structuring is achieved via the generation, use
and modification of externalized models.
In stages 1 and 2 an attempt is make to build the richest possible picture of
th problem situation. In these stages divergent cognitive processes are
encouraged via the application of group idea generation procedures such as
brainstorming (Rickards 1982) or the nominal group technique (Gill and Delbecq
1982). Stages 3 and 4a represent the ideas generated in the previous stages.
The outcome is the model of the situation as perceived by the participants.
-746-
Among a large number of structural modeling techniques (Martino 1972),
Interpretive Structural Modeling (ISM) addresses precisely the same
scientific-rational and social-consensual issues and objectives as described
earlier. The objectives and advantages of the method have been presented in
Reviow of existing SD
regions\ /covmnusty model,
Figure 1
detail elsewhere (Warfield 1976, Watson 1978). It has also been applied in
problem contexts such as priority setting in urban systems management (Waller
1975), analysis of public safety in a neighborhood (Fitz 1975), and analyzing
urban development (Malone 1975). It has also been suggested that the
technique be integrated with simulation modeling to achieve rigor in designing
the model structure (Burns 1975). The method still has some unresolved issues
in application such as the difficulty of incorporating minority views and
eliminating effects of group dynamics or pressure on the final
conceptualization (Watson 1978).
In the framework presented here, the main function of the ISM technique is to
generate the richest picture of the situation reflecting the variety of
perceptions and interests, beliefs and interpretations of available
information regarding variables and intérrelationships without imposing any
preconceived structure. The procedural guidelines and the algorithms used in
stages 2 through 4a comprise the first component of the decision support
system. Stage 4b is the second component of the proposed approach consisting
of a generic system dynamics community model, providing the widely accepted
expert knowledge about the structure and behavior of community subsystems.
The advantages of the system dynamics type simulation modeling in this context
(Sancar 1977), and the value of designing and using generic models have been
discussed elsewhere (Forrester, 1961). Here the main purpose of the generic
model is to provide the problem solving group with a menu of
conceptualizations which reflect an understanding of systems potentially
relevant to community development issues.
-747-
The situational system dynamics model in stage 4d, represents the completion
of the model externalization process. This model is constructed via a
customization procedure involving the integration of the initial structural
model of stage 4a with the generic model of stage 4b through comparison and
specification. The evaluation of the situational model takes place in stage
5. If the model is found to reflect a collective appreciation of the
situation, the process moves forward. If not, any combination of stages 2
through 5 may be repeated. Stage 6 involves the analysis of those scenarios
deemed important, desirable, or feasible by the participant group. Action to
improve the problem situation may then be taken with the aid of the
information and understanding obtained through the application of the entire
methodology (Stage 7). In turn, the methodology may be continuously
reiterated if problems or issues continue to emerge.
Within this methodological framework, the three components of a decision
support system mentioned earlier in the introduction are perceived as aids for
information processing and social interaction while reducing the biases
inherent in human cognition processes. In the long run, using these
procedures will result in accumulation of knowledge on the community systems
since the problem/decision context and its activities will be documented in
detail as part of the process. An enhanced understanding of the problem will
then be reflected from commonalities observed in a number of situational
models. In the following, the development of the generic community model and
its customization will be presented.
THE GENERIC COMMUNITY DEVELOPMENT MODEL
A generic system dynamics model contains structures and behaviors common to a
general class of phenomena or to a system. The identification of such general
relationships may be based on theory or experiential evidence. The general
type and form of information readily available to communities are also
important factors in developing a generic community model. Based on the
sociological/community development literature it was difficult to choose among
the various theoretical constructs regarding communities. Therefore an
inductive approach to model specification was taken. Accordingly, a survey of
system dynamics models that appeared to be similar in purpose and/or content
was made. At the same time, community studies undertaken by the University
Extension were surveyed to identify the recurring types of problems and the
data bases made available to the communities on an ongoing basis.
Eight system dynamics models which were documented in some detail were
studied. More specifically, the following questions were asked:
Which issues, clients and problems do the models intend to address?
How is the model limited in its scope (system boundary)?
What major subsystems are included and in what detail?
. How are the variables aggregated?
. What constitutes the major causal hypotheses?
UPONE
The answers to the first four of these questions were summarized by preparing
a table. The answer to the last question required a finer analysis.
Therefore, a commonly modeled variable, migration, was singled out and studied
within the context of three models which included this variable. This review
revealed that the structures given to the community systems and subsystems are
-748-
very different from.one another. This is true even when the researchers’
stated purposes are similar. In other words, this survey did not indicate a
common set of variables or relationships which could form the basis of a
generic model. However, two models given by Hamilton (1969) and Sancar (1977)
appeared to be the most general purpose ones.
Next, the types of community development problems and the data bases available
to communities in Wisconsin were also investigated. It was found that the
University Extension in cooperation with various departments followed an
established approach which has been applied to a large number of communities
and small towns. This approach consists of six separate studies: a visual
analysis, a community consensus survey, a trade area survey, a survey of local
business people, a threshold level analysis, and a business district design.
Any or all of the studies may be undertaken at the request of the
communities. In addition to the above studies two other data sources were
identified consisting of a database for community economic analysis (Hustedde
et al. 1984) and a detailed survey of industrialists’ concerns and intentions
to expand or relocate, conducted by the Bell Telephone Company. These studies
provided the general concerns, the established common language regarding the
variables and concepts, and the secondary sources of data which may be used in
customizing a generic model.
Based on the above information a tentative community model was constructed by
combining the main features of the Hamilton (1969) and Sancar (1977) models so
as to reflect various documented concerns and take advantage of the available
data sources. At this stage the generic model consists of a global
causal-loop diagram showing the main subsystems and their interaction, and a
list of state variables (Figures 2 and 3). It was assumed that the
situational system dynamics model would reflect the global causal-loop diagram
and all of the subsystems; however, it may contain more or less number of
sectors and any number of additional parameters.
PROBLEM STRUCTURING
Structural models portray the features of the situation and employ graphic
techniques to show how these are interrelated. Therefore they may be used to
build the necessary conceptual bridge between the mental/prose models and the
dynamic models and to enhance and document the thought process. The general
criterion for such techniques is that they should reflect human perceptions
and help to integrate them into systems modeling by providing the necessary
cognitive aids for individuals and groups to cope with complexity. In this
section, the general characteristics of structural modeling, more specifically
of Interpretive Structural Modeling, will be explained. Then, an enhanced,
more detailed version will be proposed which further specifies how each
decision within the structuring process will be carried out in order to
encourage creativity and improve human inferences and judgements.
The Interpretive Structural Modeling (ISM) technique which is an advanced
representative of the general class of structural models, has been developed
to compensate for bounded rationality (Simon 1969), which imposes severe
limitations to mental activity in dealing with complex situations.. This
limitation results in premature structuring and avoidance of complexity
-749-
leading to an arbitrary or biased selection of entities or variables, objects,
events, and relationships to be considered in problem solving. ISM is a
content-free technique. Its main advantage lies in the fact that once an
entity list and a relational proposition is specified, the situational
structure can be generated without being adversely influenced by the above
mentioned mental limitations. This is achieved by evaluating the entities two
at a time via an interaction matrix. There are three major steps in the
implementation of ISM (Wilson 1982). These are: 1. the determination of the
variables and the relational proposition; 2. the generation of the interaction
matrix; and 3. the derivation of the structural network.
In the first step, the variables are listed and the corresponding relational
proposition is determined. These variables may be activities, elements,
attributes, issues, problems, objectives, or opportunities. The relational
Market Employment
row
ecgencus inns!) eiccer
Relative Cost ees i
Relative Wages Haseena Saving Occupancy {
Emplennort
Relative Accaue
to Martactes Level of pe L i sox"
Unen ployment industry] [wage
woe Environment
afo——y) Precnace On |
FS Popalationm Migaion ee ay
Destine ———T ]
| Concumer Government
High income Revewes
Med income Expenses
Lew income Cottvels
Figure 2. Generic Community Development Model
proposition, is in turn, a statement framed as “facilitates the solution of,"
“influences,” “is subordinate to," or the like. Each of these expressions
represents a different way of "framing" the situation, or an alternative
perspective. Once a perspective is chosen, all variables have to be
consistent with it. In general, this consistency check is completed by the
facilitator. The relational variable is determined based on the nature of
variables. The generation of the interaction matrix is conceptually
straightforward. It may be done by paper-and-pencil, or interactively on a
computer; by the individuals or as a group. Finally, the matrix entries may
be evaluated as to importance by ranking and rating. The derivation of the
structural graph is accomplished by the use of an appropriate software
designed to generate “di-graphs" from the given set of matrix interactions.
-750-
In this study the main function of the ISM prior to simulation modeling is to
reflect a collective appreciation of the situational context. It is this
representation which subsequently becomes the decision environment, and is
expected to influence future decisions. There are alternative ways of
POPULATION EMPLOYMENT
Children (0-13) Export Industry Employment
Teenagers (14-19) Business-Serving Employment
Young Adults (20-24) Household-Serving Employment
Prime Age (25-44) Skill Level
Middle Age (45-64) Wage Income
Old Age (65+) Unemployment.
Births Industry Mix
Deaths Competitive Pressure on Wages
Migration Other Economic Variables
Market Growth
CONSUMER Relative Costs
High Income Relative Wages
Middle Income Relative Access to Markets
Low Income Relative Access to Materials
ENVIRONMENTAL OCCUPANCY
Open Space Industrial Establishments
Vacant Land Business Services
Building Quality Convenience Stores
Parking Shopping Stores
Attractiveness to Industry Specialty Stores
Entertainment Establishments
Housing
Vacancy
Figure 3. List of Generic Model Variables.
carrying out the three steps conceivably having different impacts on the final
outcome in terms of achieving the following additional objectives:
1. Generating creative options, or “systems design",
2. Learning through search and interpretation of information concerning
facts and values relevant to the situation and overcoming judgemental
and inferential biases, \
3. Generating and documenting information concerning the interactions
among participants; their negotiations concerning different
interpretations of reality, and the different perspectives and how
they are changed and modified.
To incorporate the above considerations into the ISM, four explicit decision
points were introduced. This modified version of the ISM is given in Figure
4, where the decision points are represented by the diamonds in the flowchart.
Procedural guidelines to improve these decisions are derived from applicable
literature in various fields dealing with human judgement and. decision making
and social-psychology of creativity. The application of these guidelines also
-751-
depend on which of the prototypical planning problems" (Kristensen 1985), best
represents the situation. The activities at each decision point have been
called 1) framing, 2) categorization, 3) interaction specification, 4)
derivation and testing of the network. These activities will now be described
in detail since they constitute the major novel elements of the proposed
decision support system using ISM.
L
List of Varioules
Retationsl Statement
st d
fit: e Rees
Enter Matrix
Interactions
Interoction : .
|-—X_ Derive >
Matrix.
Gop
4®) in
Figure |.
Figure 4. Flowchart for Problem Structuring
FRAMING
The outcome of the first decision point is a list of variables and a
relational proposition. During framing each question or probe used to
generate the list of variables which are thought to be relevant/important by
the participants represents a different way of “framing” the situation, or a
different perspective. The choice of a perspective (for example, listing
variables as problems rather than opportunities) and the name given to the
-752-
situation in terms of scale and specificity, may enhance or limit the group's
problem solving capabilities.
Working within a “scenario/schema" or perspective increases fluency in
generating ideas and in noticing relevant elements, while limiting the
possibility of discovering original ideas which will contribute the most to
the process. Once a perspective is established, it influences causal
inferences as well as predictions. These judgements, then, may prove to be
very resistant to modification even in view of new information (Nisbet and
Ross 1981). Each member of the problem solving group is likely to have a
dominant perspective of the situation determined by cognitive style (Mittroff
and Turoff 1973), motivational factors, and their past involvement with the
issue at hand (i.e. situational factors). These should not be externalized or
imposed on the group at this point. Therefore, the situation should be
referred to in less specific terms than any individual member would avoiding
value-laden statements. For example, statements such as “We have been
assigned to this citizens task force to discuss and recommend solutions
regarding the various problems our community is facing; such as increasing
property taxes, crime and vandalism in various neighborhoods, and closing down
of the Happy Brewery" should be avoided, because 1) the situation has been
framed as a “problem” situation, and 2) a certain: set of problems have been
prematurely emphasized.
Whatever the scale or specificity of the perceived situation may be, there are
two opposing ways of framing. One is the problem analysis perspective and the
other is the system design perspective (Sancar 1983). In most cases,
especially when the situation is "messy", the former perspective is likely to
be dominant. Since the advantages of adopting the latter perspective are
numerous (Nadler 1970, Ackoff 1978, Checkland 1981, Sancar 1983), if the
statements turn out to be problems, the opposing perspective may be introduced
by “re-framing" (Bandler and Grinder 1982).
The size of the variable list should not be Limited at this stage, since such
an effort imposes arbitrary screening of ideas and limits free thinking. To
achieve this objective and to ensure that ideas are generated by all the group
members alternatively (rather than each person presenting a more or less
self-contained perspective), a combination of triggering and round robin
brainstorming session (Rickards 1982), brainwriting or "method 635" (Warfield
1975) may be used in this step. From the initial listing, participants may
also be asked to identify which of the elements are controllable and which are
not. This type of inquiry may foster additional branched thinking and may
serve to broaden the base of elements from which the entire analysis begins.
During editing, the mediator takes the preliminary list of elements and edits
them for eventual inclusion into an intersection matrix. This is done by
checking redundancy of statement substance, uniqueness and ambiguity of each
statement. The resulting list is then presented to the group for approval.
Selecting the relational statement is an important step following the editing
phase. This statement provides the conceptual building block for the
selection of matrix interactions and the relationships which will eventually
appear in the situational model. This statement should be consistent with the
appropriate options listed in Figure 5 below. For instance, if they are
-753-
comparing opportunities or alternatives, they will be asked to consider
whether they are operating within a positive (desired state) mode or in a
negative (removing the constraints) mode. They are then presented with the
accompanying verbal operator to test whether it fits their perceptions of a
proper comparison.
Format Option Operator
1. problem positive/negative influences
2. opportunity/alternative positive/negative helps/hinders
3. phenomenon/issue neutral influences
4. attributes/functions neutral is related to
Figure 5. Options For a Relational Proposition
CATEGORIZATION
This decision point consists of two tasks which are categorization of
statements and creation of an initial causal loop diagram using the category
names. If the list generated in the previous stage is large (more than 7-10
statements), completion of the relational matrix becomes consistently harder.
In this case, categorization is advisable. Categorization will introduce the
simplest way of structuring or achieving some coherence in a situation. This
process will force the group to create a common understanding of terms and
definitions. The participants will begin to see initial interactions as those
within the categories and those between them prior to actually working on the
matrix. The resulting element categories will become an anchor for the group
during the discussion of matrix interactions. In this way, keeping the
responses consistent during the next step will be facilitated.
There are two alternative strategies to categorize the elements, in one case
the generic model is introduced to the group, whereas in the alternative
strategy it is not and the categorization is achieved discursively by making
similarity judgements. In the first case the group is asked to assign each
element to one of the sectors, or to a “new sector". In the second case, a
randomly chosen statement is used to start the process, and. it is compared to
the next statement in terms of similarity (of difference). If it is found
different, then a second category is established. The third statement is then
compared with one and two; and either assigned to an existing category, or is
used to create a new one. Prior to the assignments, discussion takes place
which may result in modifications of the original statements. The process
continues until all the statements are assigned to a category and the number
of categories is not larger than approximately ten.
The choice of one of these strategies over the other depends on the type of
problem the group appears to be dealing with in terms of uncertainty due to
goal conflicts and/or availability of alternative solutions (Kristensen
1985). In those communities where prior research on needs (Community
Concensus Study) have been done, both of these conditions can be readily
identified. Otherwise, the same measures can be applied to the initial
unedited list of statements. Goal conflicts are measured by the agreement
between public officials, community leaders, and residents regarding the
importance of problems. Availability of alternative solutions is reflected by
the number of problem statements which in effect contain solutions.
-754-
When uncertainty level is high, both goal conflicts and lack of alternative
solutions are observed. Also, instead of emphasizing one or two types of
problems (such as traffic or sewage treatment), a number of problems receive
relatively high ratings. When this is the case, it is expected that the group
will not be ready to accept the generic model categories, and the second
strategy will be more appropriate. On the other hand if the group appears to
have a more clear idea about available alternatives, and when they perceive
the challenge to be exploring the various interactions between these, then it
is more likely that the generic model will prove to be a valuable guide. Once
the categorization is finished, before taking the next step, studying a global
causal diagram showing relations between categories establishes the general
framework to be used to enter the larger matrix interactions.
INTERACTION SPECIFICATION
This decision point involves two tasks; the specification of the existence and
the nature of interactions among elements and the evaluation of the specified
relations by ranking and whenever possible, rating. The first task may be
performed interactively on the computer or manually; individually or in
groups. If the task is completed individually, aggregation of responses need
to be considered. If the relational proposition assumes transitivity, then
most computer programs automatically reduce the number of interactions that
need to be completed, thereby increasing the efficiency and the reliability of
the procedure.
Completing the above tasks requires that the participants make causal
inferences. From a cognitive standpoint, the framing of the situation, the
information or data available to them, and the procedures that are used to
interpret the data are important factors which influence their judgements.
People use various judgemental heuristics in making inferences, and are
successful in many cases. The observed shortcomings and biases in human
judgements are due to either inappropriate use of these heuristics or due to
failing to make adjustments on initial judgements in view of new information.
Causal inferences: are of two kinds which are contingency framing and outcome
framing or prediction. Contingency framing may lead to erroneous inferences
due to inappropriate use of the “representativeness heuristic, and the
“availability” heuristic. Representative heuristic is the tendency to
attribute overt behaviors or problems to systemic dispositions or anomalies
rather than environmental influences by ignoring situational factors. An
example would be to attribute a drug and vandalism problem to a general lack
of law and order instead of attributing it to the closing down of the local
YMCA and the discontinuation of various youth programs due to a lack of
funds. The availability heuristic refers to the effect of the degree of
availability of causal factors; their salience and ease of retrieval, on
causal inferences. An example would be the influence of a recent news item
about police negligence in a youth committed robbery in making the same kind
of attribution as above. This points out the effects of both perceptual and
verbal manipulations on inferences.
Biases in making predictions are mainly due to failures to use “base rate" —
‘data such as prior probabilities, population proportions, and central
tendencies. All too often people make judgements on the basis of single
-755-
events which seem to support their pet theories, or which are vivid, even when
they are presented with contradictory objective data. For example a
businessman may refuse to participate in a beautification program, because he
knows of a business in another town which went bankrupt a year after spending
all that money on a face-lift. All the base rate information which shows the
positive effect of beautification on sales, will not change his mind.
Both of these biases may be traced back to the “theories” people bring to
situations, and also to the types and forms of data/information available.
The impact of data on changing or modifying prior theories is very little
unless certain conditions are met. The most important condition is to
supplement base rate information with causal explanations, i.e. with an
alternative theory. The generic model, and a study of the global causal
relations between the sectors after categorization is expected to facilitate
the introduction of causal explanations for the data. The second condition
has to do with presentation of data. People weigh data in terms of its
vividness which is based on emotional interest, concreteness, and spatial or
temporal proximity. The research in this area suggest that unless the
situational base rate data is presented vividly, they will be ignored.
From the above discussion it becomes apparent that the awareness of various
biases, and efforts to minimize them via framing and data presentation, will
lead to more enlightened inferences. Nevertheless it is also acknowledged
that there is no one correct way of interpreting the situation in a social
context since judgements are also guided by concerns other than “correctness,”
such as fairness and justice. In a group context, in addition to introducing
the above statistical realities (e.g. the base rates), which clearly favors
quantitative information, a more flexible approach which accepts the presence
of competing explanations and embraces qualitative information will be more
appropriate. An example of such a procedure is “assumptional analysis" (Mason
and Mitroff 1981) developed to organize various parts of an argument, such as
“warrant", “backing”, “claim”, “rebuttal"; in interpreting information. The
importance ratings can then be used to simplify the matrix so as to have three
to five interactions per column (or row) and avoid complicated networks
lacking conceptual clarity and visual impact.
DERIVATION AND TESTING OF THE NETWORK
In the fourth decision point, the matrix is evaluated by using an ISM computer
program. The program output is a “directed graph" that shows a hierarchy of
variable clusters and relationships among them. This output is translated
into a graphic network representation using a heuristic algorithm which will
not be given here for lack of space. An overly complex network (too many
lines crossing) reflects either the presence of too many matrix interactions
or unclear thinking. In the first case the situation may be remedied by
vesimplifying the matrix interactions, while in the second case the logic of
interactions must be reconsidered.
After the network graphics is finalized, its general features are discussed to
see whether this is an acceptable representation of the situation. At this
point it may be possible to reduce the problem into separate routine tasks to
be assigned to appropriate existing public or private organizations.
Otherwise, the group is ready to start building one or more situational
simulation models to generate and test alternative solutions.
-756-
PROCEDURAL TESTING OF PROBLEM STRUCTURING
The problem structuring procedure described in the preceeding sections were
tested using two cases of community development previously studied and
extensively documented by the University of Wisconsin Extension for the towns
of Jefferson (Such, 1983) and Racine (Drewiske, 1984) in Wisconsin.
The Jefferson case was an example of a problem with high uncertainty. There
were a high number of problem statements with few solution implications
suggested by the community members at a downtwon retreat. The relational
statement used was "The solution of problem A will help/hinder the solution of
problem B." Categorization was done by clustering the elements in terms of
their differences.
The Racine case, on the other hand, represented a more structured situation.
All the issues were stated as opportunities related to a harbour and marina
development along with other recreational, residential, and commercial
development potential. Hence it was easier to use this generic model as a
guide to categorize the variables and to specify the interactions. In this
case, the relational proposition was “Achievement of A will support/hinder the
achievement of B."
The resulting networks for the problem are shown in Figures 6 and 7 for
Jefferson and Racine, respectively. Based on these two examples and their
networks, several observations can be made. The Racine network in Figure 7 is
conceptually more clear even though the number of variables are almost twice
as many as those for the Jefferson case. This is due to the use of the
generic model structure in defining the matrix interactions. However, in both
networks it is possible to clearly identify meaningful clusters of variables
and their relation to the generic model sectors. Furthermore, by examining
the input/output arrow configurations, the major dependent variables; the
throughput variables; and those variables which appear as either
uncontrollable, trivial or external policy/decision variables can be
determined. At this time, the major feedback loops involving the main
dependent and throughput variables and the paths at their interfaces also
become apparent.
Since the two cases described above have not involved the active participation
of stakeholders or the base rate information, no conclusions could be drawn
regarding the strategies for information presentation and group dynamics.
However these aspects are being investigated in a follow-up application and
will be reported subsequently.
DEVELOPING THE SITUATIONAL SYSTEM DYNAMICS MODEL
The situational model is generated by transposing the elements of the
completed structural model into a system dynamics model. We note here that
customization actually begins with the framing of the situation. The
integration of the situational and the generic models can then be initiated at
Figure 6. Network Representation of Jefferson Case
one of th following steps corresponding to the various decision points:
1. In the “editing phase,” where convergence between the terms in the
generic model and the participants' statements may be achieved,
-757-
2. In categorization, where the generic model categoires may be used to
recognize the participants’ statements into clusters,
3. While generating matrix interactions, when the various interactions
can be guided by the causal linkages of the generic model.
4. After the completion of the structural model, at which point items (1)
through (3) above may or may not have been applied.
After the completion of the structural model further customization can be
continued via the following activities:
1. Comparing the sectors of the generic model with the variable
cateegories in the interaction matrix,
2. Comparing the variables in the network by the state and flow variables
in the generic model,
3. Deciding on the level of aggregation of each sector based on the above
comparisons,
4. Comparing the main feedback loops in the structural model with those
in the generic model,
5. Deciding on the relationships to be included in the situational model
with their incorporation into the generic model.
With the completion of these activities the situational model can now be used
to generate and test various policy options. Since the community members‘
participation in the construction of the model has been maximized, it is
expected that they will now be able to take part in the model testing and
experimentation phases. These steps have not been carried out for the two
cases reported above. Complete application of the procedure, including the
development of the situational model will be implemented for another case in
the next phase of this study.
CONSUMER
USTRY ond |
ett
20 Vecotional
. . ~ PHYSICAL. ENVIRONMENT
Figure 6. Network Representation of Jefferson Case
~758-
BS Ave. length
< t
stay (marina users) — 21, Papin, ity moorings
rT . Pent ecandany oe
Ekta et onto far
sein dt [82 thd mit comel ee ba € tats
ey
#
eR
=
I
x
> hE
i;
eal
nga
i
% thy to Lane
Figure 7. Network Representation of Racine Case
CONCLUSIONS
Recent discussion among System Dynamics researchers and practitioners refers
to generic models as tools for the analysis of problems common to classes of
systems. Such developmental work will contribute to the achievement of a
common understanding of significant problems on the part of model builders and
users, and an efficient approach to the study of systems differing only in
unique, context specific ways.
In this paper, a procedure for the development and customization of a generic
model for rural community development has been described. The components of
the procedure have been presented as decision aids within the context of a
larger methodological framework developed for resolving unstructured
problems. These consist of a generic communty development model, a problem
structuring algorithm, and a situational system dynamics model.
A survey of various system dynamics models of community, regional, and urban
systems indicated that there were no common variables and interactions which
could form the basis of a generic community development model. A tentative
generic model structure and a list of variables were chosen from two of these
-759-
models which appeared to be the most general purpose ones, keeping in mind the
data sources commonly available to small towns and rural communities.
The problem structuring algorithm was developed by introducing explicit
decision points - framing, categorization, interaction specification,
derivation and testing - into the Interpretive Structural Modeling process.
To impelment each of these decisions, guidelines based on cognitive criteria
to enhance the participants' information processing capabilities while
reducing inferential biases were provided. The procedural mechanics of the
problem structuring was tested using previously compiled community development
reports of two towns in Wisconsin. The resulting networks showed meaningful
clusters of variabls and it was possible to relate them to the generic model
sectors. Finallyk the steps required to develop the situational system
dynamics model by integrating the structural networks and the generic model
were described.
In the following phase of the study, the entire procedure will be applied to a
community development case involving the community members. The benefits
expected from the implementation of th procedure can be summarized as follows:
1. Cognitive facilitation: the participants will be able to use more and
better informaton regarding their community; their understanding of
the systemic interrelationships will be enhanced and more consistent
with the objectives data and available theories.
2. Social facilitation: the number and diversity of prticipants will
increase as the problem structuring activity proceeds. The
participants will acquire a common language as reflected in the use of
similar terminology and consistent definitions, i.e. will have
developed a shared perspective of the situation. Reaching a consensus
will be easier, or sources of disagreement will be clear.
3. Creativity: the number and originality of solution proposals will
increase, and th participants will enjoy taking part in the activity
and express satisfaction with their contributions.
Even though it will not be possible to put the proposed procedure through a
rigorous testing with respect to these expectations, appropriate measures will
be taken to provide insights concerning the successes and failures.
ACKNOWLEDGEMENT
This study has been supported by a Hatch Grant No. WIS 02721. Under the title
of “Multi-Level Planning Model for Rural Development."
REFERENCES
Ackoff, R.L. The Art of Problem Solving. A Wiley Interscience Publication.
New York: John WlIley and Sons, 1978.
Bandler, R. and Grinder, J. Reframing. Moab: Real People Press, 1982.
Burns, J.R. Toward a Mathematically Rigorous Methodology for Simulation of
Social Processes. Summer Computer Simulation Conference. San Francisco, 1975.
-760-
Checkland, P. Systems Thinking, Systems Practice. Chichester: John Wiley
and Sons, 1981.
Drewiske, D.P. Economic Impact Potential for the Racine Harbor Development
Project. Madison: Recreation Resources Center, University of
Wisconsin-Extension, 1984.
Dunn, E.S. Social and Economic Development: A Process of Social Learning.
Baltimore: Johns Hopkins Press, 1971.
Fitz, R.B. ISM as a Technology for Social Learning. In: M.M. Baldwin, Ed.,
Portraits of Complexity, Battelle Monograph No. 9, Columbus: Battelle
Memorial Institute, 1975.
Forrester, J. Industrial Dynamics. New York: MIT Press, 1961.
Gill, S.L. and Delbecq, A. Nominal Group Technique In: Olsen, S.A. Ed.
Group Planning and Problem Solving Methods in Engineering Management. New
York: John Wiley and Sons, 1982.
Hamilton, H.R., Goldstone, S.E., Milliman, J.W. Pugh III, A.L., Roberts, E.B.,
Zellner, A. Systems Simulation for Regional Analysis: An Application to
River Basin Planning. Cambridge: MIT Press, 1969.
Hustedde, R., Shaffer, R., Pulver, G. Community Economic Analysis: A How to
Manual. Department of Agricultural Economics, University of Wisconsin,
Madison.
Kristensen, K.S. “Coping With Uncertainty in Planning," JAPA, Vol. 51, No. 1,
Winter 1985, pp. 63-73.
Malone, D.W. Application of ISM: Relating Factors for Urban Success. In
Baldwin, M.M. Ed. Portraits of Complexity, Battelle Monogrph No. 9,
Colombus: Battelle Memorial Institute, 1975.
Martino, J.K. Technological Forecasting for Decision Making. New York:
American Elsevier, 1972.
Mason, R.O. and Mittroff, I.I. Challenging Strategic Planning Assumptions.
New York: John Wiley and Sons, 1981.
Mittroff, I.I. and Turoff, M. “The Whys Behind the Hows," IEEE Spectrum, Vol.
10, March 1973, pp. 62-71.
Nadler, G. Engineering Research and Design in Socioeconomic Systems, In:
Moore, G.T., Ed., Emerging Methods in Environmental Design and Planning,
Cambridge. MIT Press.
Nisbett, R. and Ross, L. Human Inference: Strategies and Shortcomings of
Social Judgement, Englewood Cliffs: Prentice Hall, Inc., 1980.
Rickards, T. Brainstorming, In: Olsen, S.A. Ed., Group Problem Solving
Methods In Engineering Mangement, New York: John WIley and Sons, 1982.
-761-
Sancar, F.H. A Model of Planning Methodology for Community Participation and
Urban Development. Unpublished Ph.D. Dissertation, Pennsylvania State
University, University Park, 1977.
Sancar, F.H. “System Dynamics as a Heuristic For Systems Design," In: The
International System Dynamics Conference, Pine Manor College, Chestnut Hill,
Mass., July 27-30, 1983.
Simon, H. The Sciences of the Artificial. Cambridge: MIT Press, 1969, 1981.
Skinner, B.F. . Science of Human Behavior. New York: McMillan, 1953.
Smith, R.W. “A Theoretical Basis for Participatory Planning," Policy
Sciences, Vol. 4, 1973, pp. 275-295.
Such, D.C. City of Jefferson Community Revitalizaton Retreat Report of
Proceedings. University of Wisconsin Extension Jefferson County, May 26, 1983.
Waller, R.J. Application of ISM to Priority Setting in Urban Systems
Management. In: M.M. Baldwin, Ed., Portraits of Complexity, Battelle
Monograph No. 9, Colombus: Battelle Memorial Institute, 1975.
Warfield, TOTOS: Improving Group Problem-Solving. In: Approaches to
Problem-Solving No. 3, Colombus: Battelle Institute and the Academy for
Contemporary Problems, 1975.
Warfield, J. Societal Systems: Planning, Policy and Complexity. New York:
John Wiley and Sons, 1976.
Watson, R.H. “ISM - A Useful Tool for Technology Assessment?" Technological
Forecasting and Social Change, No. 11, 1978.
Wilson, H.G. ISM-UNI User Manual, University of Northern Iowa, 1982.
Vickers, Sir, G. The Art of Judgement, New York: Basic Books, 1965.
