-762-
Cognitive Criteria for Structuring
System Dynamics Models
Fahriye H. Sancar
Robert J. Cook
University of Wisconsin-Madison
ABSTRACT
This paper discusses the development of cognitive criteria for use in guiding
the problem definition phase of System Dynamics modeling. The System
Dynamics modeling process is presented as currently defined in the literature.
The problem definition phase of the process is then isolated because of its
overriding influence on model structure. Topics relating to information,
information processing and group decision making are discussed and the short-
comings of human judgment and inference are identified. These shortcomings
are related back to the tasks required for problem definition and criteria
are identified which can serve as guidelines for the development of cognitive
aids for structuring System Dynamics models. The paper closes with a brief
discussion on operationalizing the concepts of cognition, creativity and
social interaction as tests of the relative value of these criteria.
INTRODUCTION
Literature in System Dynamics presents a multi-staged process for use in
selecting model elements and specifying model structures (Roberts, et al.,
1983; and Randers, 1980). It is widely accepted, however, that this process
will not provide mechanistic rules for selecting and structuring model compo-
nents, for this is an activity commonly regarded as the "art" of modeling.
While we agree that problem formulation and model structuring can be consid-
ered artistic and therefore not amenable to mechanization, we believe that
the entire process can be better facilitated by taking into account the infor-
mation processing capabilities of individuals and groups.
It is important to recognize that model structures, once complete, become
decision environments. As such, a model may be conceived of as a “cognitive
aid" which reflects an enhanced but unique understanding of the problem situ-
ation, both guiding and constraining the alternative courses of action.
Since model structure has such a significant impact on the subsequent phases
of problem solving, appropriate criteria to guide the structuring process
need to be developed. This need becomes even more apparent when the initial
phase of modeling involves the direct participation of a number of people
including the client, and where communication and issues of accountability
become important. The goal of this paper is to present the tasks required
for guiding and organizing these tasks so that the cognitive capabilities and
limitations of individuals are taken into account.
To determine which specific tasks to consider the guidelines provided by
Randers (1980) and by Roberts and her colleagues (1983) will first be com-
bined to give a full account of the decision process employed by SD modelers.
That portion of the process which relates to problem definition will be iso-
lated for further study. Then, to develop criteria for guiding problem defi-
nition behaviors, topics related to information and information processing
-763-
will be discussed with respect to 1) the types of information used by
modelers, 2) the manner in which the information is processed for decision
purposes, and 3) the cognitive pitfalls common to such decision situations.
Since modeling under socially complex conditions will often require that
several individuals participate, group process will also be discussed as it
relates to decision making. The discussion will close with a set of proposed
criteria which are thought helpful in enhancing both the cognitive and group
processes previously identified. Such criteria may therefore serve to guide
the task completion behaviors which collectively lead to a problem definition.
THE MODEL STRUCTURING PROCESS
Randers (1980) identifies four stages of model construction; they are con-
ceptualization, formulation, testing, and implementing. The conceptualization
stage is facilitated through application of the SD paradigm and the formula-
tion stage through use of the DYNAMO computer language. Testing is performed
both logically and empirically and may require the iteration of stages one
and two. The implementation stage involves model application for policy
analysis, for sensitivity analysis, and for generating potentially useful
information.
Regarding the initial aspects of model development, Randers assumes a rele-
vant problem to exist as a prerequisite for model conceptualization. The
conceptualization stage is therefore presented as the most critical, for it
is here that the basic assumptions concerning the behaviors to be modeled and
the components used in doing so are developed. Procedurally speaking,
Randers calls first for a definition of the reference mode, i.e., a descrip-
tion of the time development of interest; and second, for the identification
of basic mechanisms, i.e.,.the most fundamental set of interrelated compo-
nents which produce the reference mode. A time horizon is necessarily iden-
tified and made explicit in the reference mode, and both the system boundary
and the required level of aggregation are defined through identification of
the basic mechanisms. From this, the modeler formulates a flow diagram if
desired and writes the specific code required by the DYNAMO compiler.
Roberts and colleagues (1983) identify a model building process having six
phases: they are problem definition, system conceptualization, model repre-
sentation, model behavior, model evaluation, and policy analysis and model
use. This is nearly identical to that proposed by Randers but is more
explicit with respect to identifying both the definition of the problem and
the analysis of model behavior as separate stages or phases in the process.
Important. and distinguishing aspects of the Roberts presentation are first,
the stated need for modelers to concern themselves with an initial problem
definition phase, and second, the ensuing discussion on analyzing less-struc-
tured problems.
Four critical components of the problem definition phase are introduced
which essentially add detail to Randers' conceptualization stage and intro-
duce elements of real-world constraints into the process. These components
are the "perspective", the "time horizon", the "reference mode", and the
“policy choices". The reference mode is described by taking a particular
modeling perspective, i.e., an explicit assumption of a point of view, and
deciding on the modeling time horizon or time period of interest. Only by
combining these two concepts can one clearly identify the reference mode,
-764-
i.e., the time development of interest. The fourth component, the policy
choices, requires that initial proposals for change be identified. They must
appear realistic in the face of any social, political, legal, or other con-
straints which could significantly influence future action. These constraints
become part of the modeler's conceptualization and influence the form and
content of the resulting model.
These two descriptions of the modeling process can be combined into what
Passini (1984) refers to as a “decision-plan". Figure 1 illustrates this
System Dynamics decision-plan and represents the process in a way useful for
distinguishing decisions, tasks, and behaviors and for relating each to one
another and to the process as a whole. Briefly stated, decisions give birth
to tasks which in turn foster behaviors designed to complete each task. For
example, the SD modeler's decision to define the reference mode generates
two tasks: first, identifying the perspective, and second, identifying the
time horizon. Each of these tasks implies certain behaviors designed to
complete them.
To test or
Figure 1. System Dynamics Decision-Plan.
The actual development of a decision-plan may therefore be thought of as both
a task-generating activity and a solution-generating activity (Passini, 1984)
The decisions in figure 1 read hierarchically from left to right and collec-
tively generate the set of tasks one then attempts to accomplish. The tasks,
which read in reverse order from right to left, illustrate the sequential
nature of the solution-generating activities or behaviors which eventually
lead to the completion of the most basic of tasks, i.e., test for policy
consequences.
-765-
Once a decision-plan has been specified, the behaviors invoked enroute to a
solution require modelers to use various types of information for each task
completion exercise. The more formalized a process and its many tasks
become, the more guided are the behaviors. Hence, appropriate information is
easily identified and made readily available, and the manner in which it is
to be used is clearly understood. In SD modeling, those tasks corresponding
to Randers' formulation, testing and implementing stages have become quite
formalized, with solutions readily available both from the formalisms defined
by the DYNAMO computer language and from prior experiences in model testing
and implementing. However, for the majority of tasks corresponding to the
conceptualization stage, or problem definition phase, similar formal solution
procedures have not been developed. These subtasks remain rather vague and
therefore problematic. The types of information which should be brought to
bear on these tasks and the manner in which these need to be integrated or
interpreted are not at all clear. The result is that the behaviors invoked
for purposes of problem definition are not given adequate guidance.
In order to develop criteria useful for guiding modelers through the various
tasks on the decision-plan, the types of information and corresponding
strategies for information processing will be presented first. These will
be based on the cognitive limitations and inferential biases inherent in
human judgment and decision making.
INFORMATION AND INFORMATION PROCESSING
There are three types of information inputs into the problem definition phase
of modeling; sensory information, memory information and inferred information
(Passini, 1984). Sensory information is that obtained through direct sensory
contact while memory information is evoked from past experience. Memory
information once went through the process of being directly perceived or
sensed. The distinguishing characteristic is that at a given moment memory
information can be obtained without or independent of sensory inputs, while
sensory information cannot. In this way, sensory information is directly
related to the environment or the setting within which one must operate
whereas memory information is at least one step removed from this setting.
Inferred information is that obtained from any combination of sensory and
memory information where the decision-maker manipulates one or the other
based on the influences each brings to the situation.
Corresponding to each of these information types are three task situations.
The first is when sensory information is itself sufficient to complete a
task, the second is when memory information is also required, and the third
is when information must be manipulated to arrive at some inference. A
fourth situation frequently encountered is when little or no relevant infor-
mation seems available. For purposes of SD modeling and particularly for
defining less-structured problems, individuals are most likely to be in situ-
ations where both memory and sensory data are combined and an inference made,
or where little or no information is available either from one's memory or
from the setting. And when in the latter situation, individuals will ini-
tiate a search process for information acquisition, Thus, modelers typically
find themselves in two of the four task situations and subsequently use either
of two broad strategies for guiding behaviors and decisions. They will employ
an inference strategy when information is. both available and considered ade-
quate, or they will employ a search strategy to acquire the needed infor-
-766-
mation. The search strategy is performed until an inference becomes possible
and is considered acceptable. Indeed, as individuals search they also infer
and when satisfied with their inference the search is often ended. There-
fore, the inference strategy is ultimately employed in all modeling behaviors.
When engaged in inference, people use two broad types of intuitive implements,
knowledge structures and judgmental heuristics. Knowledge structures relate
to memory information and include sets of beliefs, theories, propositions and
schemas, all of which people acquire through experience. Structures such as
these are used to label, categorize and form expectations about objects or
events and to suggest appropriate responses to them. Judgmental heuristics
are cognitive strategies used to reduce complex inferential tasks to simple
judgmental operations. Nisbett and Ross (1980) suggest that judgmental .
heuristics are the chief determinants of the arousal and use of the various
knowledge structures.
Three judgmental heuristics are identified by both Nisbett and Ross (1980) and
by Einhorn and Hogarth (1982); they are the "representativeness" heuristic, the
"availability" heuristic, and the "relevance" heuristic. The representativeness
heuristic is used to reduce inferential tasks to simple similarity judgments.
The availability heuristic is used to judge frequency, probability and even
causation, with such judgments based entirely on the extent to which appli-
cable information is readily available in one's memory. The relevance
heuristic is used to assign inferential weights to data in proportion to
their salience and/or vividness in one's mind. These heuristics generally
result in successful evaluations of everyday life complexities. However,
the same mechanisms are also used inappropriately and without adjustment.
Unfortunately, individual decision-makers are not typically aware of the
natural limits on their cognitive abilities and, hence, of the possibility
for biased inferences from the use of these heuristics.
Indeed, researchers have found rather severe and systematic shortcomings in
people's ability to make judgments and inferences when confronted with complex
problems and data structures (Tversky and Kahneman, 1978). Collectively
speaking, judgmental errors can be grouped into two broad categories; first,
those involving the use of available information such as base rate data, i.e.,
information about the probability that the variables - objects, individuals,
or events - will take a specific course of action; and second, those involv-
ing the use of information previously acquired as reflected in one's know-
ledge structures. This second category of inference uses cause and effect
information, i.e., the probability that a specified course of action will
produce a specified outcome.
When making inferences using base rate information, people will tend to
underutilize it to the degree that they consider it of relatively lesser
quality or simply inapplicable to the decision. Quality and applicability
is often inappropriately assessed due to inattentiveness to normative prin-
ciples.
When inferring about cause and effect, individuals’ knowledge structures will
greatly influence their predictions. The literature clearly documents what
is referred to as the perseverance tendency, where one's beliefs tend to per-
severe for various reasons beyond the point at which behavioral norms call
for change (Nisbett and Ross, 1980). People have been found to adhere to
-767-
preconceived beliefs in the face of evidence that ought, rationally, to
weaken or even reverse their belief (Hovland, Janis and Kelly, 1953). People
also tend to assess new information asymetrically, meaning that an individ-
ual's apriori expectations and theories tend to bias the detection of co-
variation or relationships among base rate information (Nisbett and Ross,
1980), and therefore bias any inferences made. Thus, opinions once formed
are slow to change in response to new information. It is interesting to
note that scientists have also been observed adhering to theories well past
the point justified by evidence (Mahoney, 1976).
In a similar manner, individuals recognize the relevance of data which con-
firm prior held hypotheses more readily than those which disconfirm them and,
in this way, tend to search for such data in evaluating their hypotheses
(Wason and Johnson-Laird, 1965). Furthermore, involvement in the process of
causal explanation or hypothesis formation has been shown to influence the
plausibility or subjective likelihood individuals place on the explanation
of events (Ross, et. al., 1977). And last, causal schemas appropriate to
some settings ate often inappropriately applied to others (Einhorn and
Hogarth, 1982). People are therefore prone to making inappropriate pre-
dictions.
Such shortcomings in judgment and inference pose real threats to modeling in
which judgmental data are used. With particular regard to the problem defi-
nition phase of SD modeling, two main issues become apparent. The first is
that individuals are likely to come to the modeling exercise with predeter-
mined notions of the problem and its causes. When the problem conditions are
complex and difficult to define, such premature frames of mind may so dominate
the definition phase that new knowledge and new understandings cannot be
brought to bear on model form and policy choice. The second issue involves
the use of information and the tendency for biased inferences through the
inappropriate use of judgmental heuristics. Information which is both rele-
vant and important may be ignored entirely or may be given inappropriate
weights.
Fortunately, however, research does support the notion that the tendency to
persevere in one's beliefs can be countered by providing individuals with an
opportunity to alter their hypothesized cause and effect relationships through
the timely intervention and proper presentation of relevant information. An
individual's theories and beliefs are subject to change, but such seems to
require information which is concrete, sensory, and personally relevant
(Nisbett and Ross, 1980). If relevant base-rate information is to have any
impact, it should be presented in a clear and vivid manner. In addition,
causal explanations for the base rate information presented will increase the
likelihood of it being used (Einhorn and Hogarth, 1982). And when individuals
go beyond making one-time-only judgments and instead are allowed to manipulate
and act on both the data and the model, i.e., if the "action" component is
present, they may learn how to better use information and to make more
informed judgments (Einhorn and Hogarth, 1982).
GROUPS AND GROUP PROCESS
This last discussion centers on group processes and is thought equally impor-
tant in terms of the problem definition phase of SD modeling. This is
because one of the advantages of SD is its suitability for collective decision
-768-
making where experts and laypeople are able to participate in various phases
of the modeling process. While group decision-making will generally influ-
ence the quality of decisions in a positive direction, both advantages and
disadvantages exist and need to be discussed as they relate to cognitive pro-
cesses,
Groups can first be characterized by the various dispositions of their mem-
vers. Each will possess what Checkland (1981) refers to as a worldview which
is composed of three elements: cognitive representations, evaluations, and
ideals. In the group decision making context there is no pre-given worldview
but one which is produced by the active participation of individuals and
through negotiations concerning differing interpretations of reality
(Checkland, 1981; p. 276-277). These worldviews may manifest themselves in
terms of knowledge structures acquired through professional affiliation
and/or in terms of one's motivations.
In addition to adhering to a unique worldview, each group member may utilize
a particular cognitive style in their approach to problem-solving (Mitroff
and Turoff, 1973), with each based on an internally consistent inquiry
system, and with no single style being the correct one. When dealing with
unstructured problems the adoption of only one of these inquiry models is not
justified, hence, the use of groups and group process to guarantee alter-
native cognitive styles.
Groups possess information processing capabilities which improve decision
quality beyond that normally achieved by individuals. Gibb (1951) found that
the number of different ideas generated by a group was greater than that of
any one individual. In addition, groups have been found superior to the
individual when the task involves a search process. There is a greater
probability that one of several people will produce needed information than
that a single individual will do so (Collins and Guetzkow, 1964). These
same authors found groups to be superior in producing accurate interpreta-
tions because a wider variety of views are voiced and because when opposing
views are considered, group process can be structured in order to guarantee
some amount of anonymity which in turn allows for the freer flow of ideas and
the joint assessment of all opposing views without individual hostilities
forming.
Groups also exhibit certain overt behaviors. A group will always be respond-
ing on either of two levels; 1) responding to the problem agenda or the busi-
ness at hand, or 2) responding to the hidden agenda, i.e., the motives,
desires, aspirations and emotional reactions of individual participants.
Regarding the hidden agenda, group perceptions of such factors as prestige,
power and education often serve to unduly influence interaction, whereas
each individual's self-awareness, their ability to listen and the breadth of
social acceptance each displays are factors which can positively influence
group success (Patton and Giffin, 1973). Groups therefore exhibit both
defensive behaviors which diminish effectiveness and supportive behaviors
which act to improve effectiveness.
Several problems relating to behaviors and processes in group situations have
been identified by Patton and Giffin (1973). The first is that all too often
either the group facilitator or the individual participants assume that their
concerns are shared by others. But since the concern brought to the group
-769-
session strongly shapes one's responses on many levels of interaction, it
is imperative that mutual concern not be assumed. Rather, it should be veri-
fied by attempts to determine if it exists and to clarify what these mutual
concerns are. If mutuality of concern is not present other means of pro-
moting cooperative behaviors should be utilized.
A second problem which often plagues groups is a lack of specific information.
The ability to recognize this is key, for individuals are often unwilling to
admit that they are relatively ignorant on certain subjects. In a similar
vein, groups are often observed behaving with such conformity that informa-
tion which is relevant but yet unmentioned will not be brought to the discus-
sion. Group ideation techniques may be appropriately applied to increase the
amount of information made available and thereby unleash the superior ability
of the group to identify and interpret wide ranging data and opinion.
A third problem is that groups may sometimes prematurely focus on a narrowly
defined perspective or emphasize possible solutions well before it is appro-
priate to do so.. Members may take stands based on their own knowledge
structures well before the problem has been adequately defined. And while
the goal or solution being emphasized may indeed be that which is eventually
adopted by the group, it is again important that each member feel that their
inputs have been decisive in the selection process.
A fourth problem is that people assume that truth, and therefore consensus,
will result from adequate discussion even if entirely informal in character.
Particularly when the information content of the modeling exercise is complex,
more formal aids to cognition need to be applied to integrate the information
made available, to reduce it to a simpler form for decision purposes, and to
help the group achieve a high degree of consistency in its logic and decision
process.
CRITERIA FOR MODEL STRUCTURING
Having identified the limitations in individual judgment and inference as
well as the problems inherent to groups and group process, criteria can now
be established to guide model structuring. The specific tasks to be com-
pleted are 1) the identification of the perspective, 2) the identification
of the time horizon, 3) the establishment of the reference mode, 4) the
identification of the policy choices, and 5) the definition of the basic
mechanisms. The following discussion summarizes each and presents the recom-
mended criteria.
PERSPECTIVE 1
When identifying the perspective, or the group frame of mind, it is important
not to assume that participants have shared perceptions. Instead, the group
should attempt to clarify and determine the different perspectives that may
exist. In establishing a shared perspective it is also important to avoid a
premature focus on a single item or on a small set of items as a basis for
mutual agreement. Instead, the group should accomplish the task of identi-
fying the perspective through a process which generates an exhaustive list
of concerns, issues or problems brought to the meeting by the various parti-
cipants. The procedure used should not allow any single individual to
-770-
dominate the discussion and if possible should guarantee anonymity. Each
participant must be made to feel that their inputs are critical.
From the standpoint of the individual within a group, avoiding an early focus
on a narrowly defined perspective will promote flexibility in the application
of knowledge structures. Participants can therefore be expected to remain
open to new information and to different ways of interpreting the knowledge
they possess and that which they may acquire. Providing such guidance to
individual thinking has been shown to feed back to the group and to promote
more effective behavior through the open, communicative and generally sup-
portive actions of its participants (Patton and Giffin, 1973).
The criteria which should govern behaviors when identifying the perspective
are as follows:
1. Aid the individuals to externalize their dispositions regarding
the problem situation by probing them to describe the situation in
their own terms.
2. Ensure equal participation by the group members.
3. Avoid premature closure on a single ‘perspective:
i. Promote broad thinking, idea generation.
ii. Withold judgment, do not eliminate or limit the scope
of viewpoints.
iii. Entertain opposing viewpoints.
When these criteria are applied the outcome of this task will be a number of
issues and viewpoints which may or may not converge on a coherent perspective.
Before such convergence is even encouraged, the individuals in the group need
to explore the underlying assumptions and implications of their statements.
This may be accomplished by discussing the time horizon, the reference mode
and the policy alternatives corresponding to the perspectives that are
generated in the subsequent tasks.
TIME HORIZON
When identifying the time horizon the group will attempt to decide how far
into the future to continue their analysis. This will be dependent on the
perspectives being discussed and on the quality of the information which can
be brought to bear on the situation. Better information allows for better
predictive capabilities which may in turn make participants more willing to
commit themselves to inferences further into the future. At this stage, the
types and the relative quality (availability and certainty) of information
which is supportive of the various perspectives need to be defined. For each
perspective the behaviors of interest and their associated assumptions are
stated. Next, the information types needed to justify/operationalize the
perspectives are defined. Then a tentative, judgmental evaluation of their
quality and availability is made. Finally, based on these investigations, a
time horizon is to be assigned to each perspective. The main criteria for
carrying out these subtasks are:
1. Further clarification of assumptions and implications of each
perspective.
-771-
2. Establishment of a group language based on a common understanding
of terms.
3. Identification of the data requirements for future stages.
4, Avoiding the elimination‘of and a premature closure on any of
the perspectives at an early stage until the subsequent tasks are
accomplished.
REFERENCE MODE
The "reference mode" is the time dependent behavior of interest and may refer
to either an existing or a desired state. In establishing the reference mode
the group will generate a number of scenarios and graphs corresponding to
each perspective based on the outcomes of the previous task. In order to
record the initial perceptions as a base-line, the group should complete this
task once, using only the information available to the individuals without
initiating a search. Then appropriate base-rate information should be
acquired and the group should modify the reference mode. This can only be
accomplished if the reference mode involves the past behavior of the system.
The main criteria for carrying out this task is to provide visual aids to
the group, and to apply vividness criteria in presenting data. An implied
eriteria is to maintain equal participation in generating the reference modes.
POLICY CHOICES
Policy choices are the alternative courses of action which will either result
in the desired reference mode or will alter an undesirable reference mode.
All of the previous tasks will imply alternatives. Groups and individuals
have a tendency to limit their thinking when they are discussing action
alternatives to those that are thought "controllable" or feasible. Asa
result of this tendency, the model structure is unnecessarily constrained and
it resembles the existing system which is responsible for the problem behav-
ior in the first place. In order to encourage the generation of creative
options, the main criteria is that the policy alternatives should not be
evaluated as to feasibility of implementation. 4
BASIC MECHANISMS
The basic mechanisms refer to the causal relationships or interactions among
variables which generate the system behavior that is of interest. Their
definition requires a thorough review of all perspective components and their
interrelationships so as to establish a more detailed causal framework in
achieving the reference mode. Since the previous tasks have generated a num-
ber of perspectives with corresponding time frames, reference modes, assump-
tions and policy choices, the group will have to concern itself with a com-
plex set of variables and the interrelationships which are in force. At this
point, the challenge is to structure these components by combination and elim-
ination in a non-arbitrary way while enhancing the group capability to inte-
grate information.
Given the limitations of human judgment and information integration, the
following criteria are proposed:
-772-
1. If unaided, individuals as well as groups suffer severe limitations
when dealing with complex situations, a concept commonly referred
to as "bounded rationality” (Simon, 1969). These limitations
reflect themselves in arbitrary selections of perspectives and
information which have been generated in the previous steps. To
surpass these limitations, cognitive aids to structure the com-
plexity should be employed.
2. When exploring the interactions between components, the group
should search for base-rate information relevant to the causal
relationships and present it visually along with concrete case
examples.
3. In cases of disagreement concerning the validity or interpreta-
tion of the base-rate information, a structured debate should be
organized to present the various assumptions and counter-
assumptions,
The application of these criteria to SD modeling is presented in another
paper presented at this conference where a combination of structural model-
ing procedures and a generic SD model are proposed as the cognitive aids.
EXPECTED OUTCOMES
The operationalization of these criteria for SD modeling applications will
vary, but various measurements may be taken to generate information which is
useful for understanding the cognitive strategies that are used during the
problem definition and model structuring process. Three concepts which can
be operationalized and implemented to evaluate the procedure are cognition,
creativity and social interaction.
Cognition is defined as the mental operation involving active manipulation
of information, including perception, learning, memory and thinking for pur-
poses of problem solving (Mayer, 1983). Measures of the concept seek to
determine if the procedure as applied promotes accuracy, efficiency and con-
sistency in the acquisition and use of information. The benefit of using
such measures is that judgments and inferences are expected to be more
accurate due to the attention given to the quality of information and its
interpretation. The specific uses of information such as base-rate data can
be monitored and compared with a normative model of the judgment and infer-
ence process to determine the level of accuracy achieved, Furthermore, each
participant's: general understanding and knowledge of the systems and issues
being modeled is expected to improve. One could perform a simple before and
after test of each participant's knowledge of relevant information. And
finally, individuals are expected to find the procedure relatively easy or
painless when compared to the unstructured approach to complex problems such
as those typically encountered in public and political meetings. Partici-
pants may simply be asked to compare these group sessions with others they
have attended.
Creativity refers to the process and/or the product, The essential feature
of the creative process is the grasping of previously unrelated and essential
parts of a problem in a new pattern, As a result, a novel and appropriate
product is created and the heuristic process that the group utilizes to
-773-
achieve this elicits an aesthetic response from both the participants and the
observers, such as surprise, satisfaction and stimulation (Amabile, 1983).
Operationalizations of the concept should therefore attempt to measure both
the number of solutions generated by the group and the originality of those
solutions. In addition, participants can be queried to determine the level
of satisfaction they experience from their involvement with the development
of such original solutions. It is expected that use of the outlined criteria
will result in greater numbers of solutions being generated. A simple count
can be made at appropriate points in the process. It is also expected that
the solutions will exhibit greater originality due to requirements of the
procedural criteria. Either the participants themselves or a separate group
of problem-solving experts can be asked for their opinions regarding the
originality of solutions. It is also expected that the individuals will
benefit from greater satisfaction or pleasure based on their involvement in
producing such original work. The aesthetics of the procedure itself are
thought: significant and will most likely increase the satisfaction of those
participating. The participants can then be queried at the end of the model
structuring exercise to measure the levels of satisfaction achieved.
Social interaction is defined in terms of the level of participation achieved,
the numbers of participants or size of the group, the diversity of that
group in terms of knowledge, interests, and perhaps cognitive styles, and the
clarification of values which occurs because of factors associated with groups
and group process. The level of participation is expected to be high if care
is taken to guarantee anonymity and promote individual contributions. In
addition, by establishing a shared perspective, each individual is expected
to feel at ease knowing that others also have the same concerns and are per-
haps more receptive to ideas pertinent to overall group opinion. Individuals
can be scored for the number of contributions made, whether they are ideas,
opposing views, or conforming behaviors.
The number of participants is also expected to be greater when the above
eriteria are applied because of the gradual enlargement of the group when
either expert knowledge or interest group responses are required. As the
group searches for information, experts will be added for their specific
knowledge. Others will be included because of the interests they may have
in proposed solutions. Numbers of participants can easily be counted and
differentiated into original members, new expert members and new
interest members. ‘This will also allow for a measure of participant diver-
sity. The individuals can be scored for the specific information and prob-
lem-solving styles they bring to the group. And finally, participants are
expected to experience a general clarification of values, whether consensus
is reached or divergence and eventual polarization of opinion occurs. Again,
a before and after test can be used to determine each individual's values and
to measure whether clarification occurs.
In general, making these outcome measurements in the field will not provide a
rigorous evaluation of a procedure used to satisfy the above criteria. How-
ever, a modeling procedure which keeps a running record of the evolving task
structure, the information base and the relevant aspects of social inter-
action, and which evaluates these using the suggested outcome measures, will
seperate a valuable data base for gaining insight concerning human problem-
solving.
-774-
CONCLUSIONS
When model structuring in SD is reviewed it is apparent that the major diffi-
culty in modeling less structured problems lies in the problem identification
stage. This stage consists of determining the perspective, the time horizon,
the policy choices, the reference mode and the basic mechanisms. Unfortu-
nately, the completion of each is left to the discretion of the model builder.
The objective should be to successively narrow the scope of the problem to
converge on a final representation in terms of a reference mode and the main
causal mechanisms. The major shortcoming of this procedure is an arbitrary
and premature closure in problem definition with possible judgmental biases
in information processing and insufficient participation of the various
stakeholders.
In order to avoid the above shortcoming it was necessary to examine the nature
of human judgment and information processing, and of group dynamics. Elements
of each which are capable of enhancing problem definition activities were
identified and used to develop applicable criteria.
These criteria correspond to each of the problem definition tasks in SD.
When identifying the perspective the main objective is to externalize and
include the variety of viewpoints regarding the situation. In defining the
time horizon, the sources of information required to operationalize these
viewpoints are identified and the data requirements for future modeling stages
become clear. Similarly, the reference modes corresponding to the perspec-
tives are established with the help of appropriately presented base-rate infor-
mation. The main theme in defining policy choices is to encourage the
generation of creative options. Application of the criteria to these tasks
implies the use of divergent processes rather than the convergent procedures
typical of the problem identification stages of SD modeling. As a result,
completion of the final task, i.e., the identification of the basic mecha-
nisms, involves structuring the complex sets of values and information com-
piled in the previous stages. The expected benefits of these criteria in SD
modeling are in enhancing the cognitive, creative and social interaction
aspects of problem solving.
REFERENCES
Ackoff, R.L. The Art of Problem Solving. New York: John Wiley and Sons,
1978.
Amabile, T.M. The Social Psychology of Creativity. New York: Springer-
Verlag, 1983.
Checkland, P. Systems Thinking, Systems Practice. New York: John Wiley
and Sons, 1981.
Collins, B.E. and H. Guetzkow. A Social Psychology of Group Processes for
Decision-Making. New York: John Wiley and Sons, 1964.
Einhorn, H.J. and R.M. Hogarth. "Behavioral Decision Theory: Processes of
Judgement and Choice", In G.R. Ungson and D.N. Braunstein, eds., Decision
Making: An Interdisciplinary Inquiry. Boston: Kent Publishing Company,
1982.
-775-
Gibb, J.R. "The Effects of Group Size and of Threat Reduction Upon Creati-
vity in a Problem Solving Situation," American Psychologist, Volume 5, 1951,
p. 324.
Hovland, C.I., I.L. Janis and H.H. Kelley. Communication And Persuasion.
New Haven: Yale University Press, 1953.
Mahoney, M.J. Scientist As Subject: The Psychological Imperative.
Cambridge: Ballinger, 1976.
Mayer, R.E. Thinking, Problem Solving, Cognition. New York: W.H. Freeman
and Company, 1983.
Mittroff, I.I. and M. Turoff. "The Whys Behind the Hows," IEBE Spectrum,
Vol. 10, March 1973, pp. 62-71.
Nisbett, R. and L. Ross. Human Inference: Strategies and Shortcomings of
Social Judgement. Englewood Cliffs: Prentice-Hall, Inc., 1980.
Pasini, R. Wayfinding In Architecture. New York: Van Nostrand Reinhold
Company, 1984.
Patton, B.R. and K. Giffin. Problem-Solving Group Interaction. Evanston:
Harper & Row, 1973.
Randers, J. "Guidelines for Model Conceptualization," In J. Randers, ed.,
Elements of the System Dynamics Method. Cambridge: The MIT Press, 1980.
Roberts, N., D.F. Andersen, R.M. Deal, M.S. Garet, and W.A. Shaffer. Intro-
duction to Computer Simulation: The System Dynamics Approach. Menlo Park:
Addison-Wesley, 1983.
Ross, L., T.M. Amabile, and J.L. Steinmetz. "Social Roles, Social Control,
and Biases in Social-Perception Processes," Journal of Personality and Social
Psychology, Volume 35, 1977, pp. 485-494,
Simon, H. The Sciences of the Artificial. Cambridge: The MIT Press, 1969,
1981. ‘
Tversky, A. and D. Kahneman. "Causal Schemata in Judgements Under Uncer-
tainty," In M. Fishbein, ed., Progress in Social Psychology. Hillsdale:
Lawrence Erlbaum, 1978.
Wason, P.C. and P.N. Johnson-Laird. Psychology of Reasoning: Structure
and Content. London: Batsford, 1965.
CC BY-NC-SA 4.0