Modeling as organizational learning:
an empirical perspective
Jac A.M. Vennix
Willem J. Scheper
Utrecht University, Fac. of Social Sciences,
Dept. of Gamma-Informatics, P.O. Box 80140,
3508 TC Utrecht, The Netherlands
ABSTRACT
Many system dynamics modelers consider the process of model-building more important
than the model itself. Model-building is supposed to generate considerable learning about a
policy problem. Not only at the individual level but also at the organizational level. From the
point of view of empirical evaluation research the question is how the occurrence of organiza-
tional learning as a consequence of a model-building process might be established. In this pa-
per we will explore some of the key issues and difficulties involved in establishing organiza-
tional learning from model-building empirically.
THE PROBLEM
Organizational learning is a common feature of organizations. It is part of the process in
which organizations adapt to their changing environment. Adaptation of organizations, how-
ever, is not merely a passive process. On the one hand organizations adapt to changing cir-
cumstances, on the other hand they make interpretations (Daft and Weick, 1984) and actively
manipulate their environment (Beer, 1979; Luhmann, 1976). Observation and interpretation
of the results of this process of adaptation is called learning. As Hedberg puts it: “Learning
takes place when organizations interact with their environment: organizations increase their
understanding of reality by observing the results of their acts." (Hedberg, 1981, 3).
Learning about the environment and adapting to changed circumstances is an important pre-
requisite for an organization's survival, and survival might be seen as the ultimate criterium
of organizational performance. For many organizations, however, environments have be-
come more complex and dynamic and hence more turbulent over the last few decades (Emery
and Trist, 1965; Hart, 1983). Consequently the process of learning and adaptation has been
severely impeded, since the more turbulent the environment the more difficult the learning
and adaptation process (Hedberg, 1981, 13). Particularly in turbulent environments it is not so
much the learning process per se as the speed with which the organization learns, which is the
key factor for the survival of an organization. As for instance De Geus points out:
“In fact, the normal decision process in corporations is a learning process, because
people change their own mental models and build up a joint model as they talk. The
problem is that the speed of that process is slow-too slow for a world in which the ability
to learn faster than competitors may be the only sustainable competitive advantage
(....). The issue is not whether a company will learn, therefore, but whether it will
learn fast and early." (De Geus, 1988, 71)
Various system dynamics modelers have recognized that model-building entails consider-
able learning about a complex policy problem (Meadows and Robinson, 1985; Morecroft, 1988;
Lane, 1989). It is claimed that modeling promotes and possibly accelerates the organizational
(or institutional) learning process (cf. Senge, 1989; Kim, 1989; De Geus, 1988; Stata, 1989).
1199
System Dynamics ‘90
However, the literature about organizational learning effects of model-building with organi-
zational client groups is largely speculative. Empirical evidence on organizational learning
effects as a consequence of model-building is scarce. This is partly understandable, since
empirical research in this area is frought with problems. However, this type of evaluation re-
search is important for at least two reasons. First, to test and possibly verify theories on orga-
nizational learning as a consequence of model-building. This will lead to refinement of these
theories and provide more detailed insight into the process of organizational learning.
Second, maybe even more important, it might improve the process of model-building with
client groups itself. There is no doubt that evaluation research will generate insights, which
can be used by modelers to adapt their model-building procedure, for instance with regard to
the process of eliciting and mapping knowledge from client groups.
This paper will discuss the phenomenon of model-building as organizational learning from
an empirical perspective. We focus on problems related to establishing organizational learn-
ing empirically, for instance by evaluating the learning effects of actual model-building pro-
jects. We first discuss the concept of learning and the differences between individual and or-
ganizational learning. Next, we introduce the cognitive mapping approach as a way to estab-
lish individual learning from model-building. In the subsequent section we will explore how
cognitive mapping can be used to establish organizational learning. Finally, we will formu-
late some conclusions and discuss implications for the process of building system dynamics
models with client groups.
INDIVIDUAL AND ORGANIZATIONAL LEARNING: THE CONCEPT OF ORGANIZATION
Research on individual learning has a much longer history than research on organizational
learning. Hence, most of the literature on organizational learning somehow refers to the lit-
erature on individual learning. As for instance Hedberg phrases it: "Although the concept of
organizational learning is widely used, the empirical observations are almost always taken
from studies of how individuals and animals learn in laboratories.” (Hedberg, 1981,3).
Various authors have shown, however, that the concept of organizational learning is not a
straightforward extrapolation of individual learning. An organization is not the mere sum of
its individual members (Hedberg, 1981; Fiol and Lyles, 1985). On the other hand authors like
Argyris and Schén (1978) and Shrivastava (1983) point out that individuals are the actual
agents through which learning in an organization takes place. This clearly is a paradoxical
situation. Or as Argyris and Schén put it: "...organizational learning is not merely individ-
ual learning, yet organizations learn only through the experience and actions of individu-
als." (Argyris and Schén, 1978). This paradox between the individual and the organizational
level is also reflected in the various approaches to conceptualize organizations.
In the literature on organizations one can roughly distinguish between two major ways of ob-
serving the phenomenon of organization (cf. Miller and Rice 1967, Etzioni 1961, Dalton et al.
1970). The first one conceptualizes "organizations as such" (Etzioni 1961, 2) thereby largely
disregarding individual organization members, while on the other hand, the second approach
to organizations has as its basic assumption that in fact "organizations are people” (Perrow
1970, 2).
With respect to the first type of conceptualizing organizations (‘organizations as such’) the
term:‘organization' has several different meanings. First, organization is used to denote a
societal entity (for instance a hospital, a corporation) which itself may have have an organiza-
tion. Organization in this second sense refers to the organization's structure, which basically
consists of a set of formal positions. The most general distinction that can be made regarding
formal positions in organizations is between superior and subordinate. The relationships
between these two positions can be organized in several ways. In this third sense organizing
refers to the process of organization (cf. Peters and Scheper 1990). These processes of organiza-
tion in turn may materialize in all kinds of regulations, formal procedures, rules ete.
Observing organizations from this point of view (organizations as such) means that the em-
Pirical researcher of organizational learning would employ indicators for organizational
System Dynamics ‘90
1201
learning which have somehow materialized in written form: organization's regulations, pol-
icy documents, annual reports, information systems etc.
In the second approach to conceptualize organizations (i.e. ‘organizations are people’) it is not
formal structures, rules and procedures which are believed to be of primary importance when
studying organizations, but individuals. These individuals actually assume the various posi-
tions in an organization. As can be learned from the literature on organizations, organiza-
tion members’ actions are based on their perceptions or constructions of reality (Schutz 1973,
Berger and Luckmann 1979), which in its turn is determined by the organization member's
background, knowledge, values etc, in short -using Schutz’ terminology- their biographical
situation. From this perspective, psychological and social psychological phenomena become
relevant. Taking into account the values, ideas, interactions, expectations etc. of organiza-
tional members becomes important because these determine the way organizational members
perceive the organization (cf. Weber 1976, Schutz 1973, Berger and Luckmann 1979, Weick
1979), which in turn determines the way they will carry out their tasks (cf Weber 1976, Peters
1989). In this sense organizations are largely intangible, they are primarily conceptual enti-
ties residing in the heads of the members of the organization. Or as Weick and Bougon phrase
it: "Organizations exist largely in the mind and their existence takes the form of cognitive
maps. Thus what ties an organization together is what ties thoughts together." (Weick and
Bougon, 1986, 102).
Taking this perspective as the point of departure, the empirical researcher of organizational
learning would employ indicators which take into account these values, ideas, expectations
and interpretations of individual organization members. In short the empirical researcher
would observe changes in mental models as the way to establish organizational learning.
In this paper we will take this latter perspective as our point of departure for studying organiza-
tional learning. There are three reasons for this choice. The first has to do with sources of
knowledge in organizations. As for instance Forrester (1961) has pointed out, only a very
small portion of the institutional knowledge is laid down in written documents. Much of what
is known in any organization resides in the mental models of its individual members.
Consequently, taking these mental models into account in establishing organizational learn-
ing seems of major importance.
The second reason has to do with the way organizations function and the actual determinants
of organizational behavior. Above we have pointed out that in the second perspective on concep-
tualizing organizations (‘organizations are individuals’) it is the way organization members
perceive the organization which determines the way they will carry out their tasks.
Hence, organizational performance is eventually the product of individual actions of organi-
zational members, which in turn is based on their view of the organization and its environ-
ment (i.e. their construction of reality). In trying to establish organizational learning it is
thus important to take these individual ‘constructions of reality’ into account.
The third reason for taking the ‘individual point of view’ on organizational learning is re-
lated to the model-building process itself. Model-building with client groups aims at sharing
individual mental models and integrating these into one overall view on the organization. In
this sense model-building with client groups can be considered a participative learning sys-
tem which is basically more individually than organizationally oriented (Shrivastava, 1983;
Shrivastava and Grant, 1985). Hence, organizational learning as organizational adaptation
to the environment will in this paper be studied from the viewpoint of the individual
organization members. However, this point of view produces a problem.
Although learning about perceived changes in the environment by a single organization
member is a necessary condition for organizational learning (individuals are the organiza-
tion's learning agents), it is not sufficient. Individual perceptions must have consequences
for the organization as a whole, otherwise only individual learning would occur. Stated dif-
ferently, individual learning must have consequences for actions of other organization mem-
bers. According to the above mentioned distinction between two ways of conceptualizing a
organizations, this can be achieved in either of two ways. In the first alternative the insights
are simply laid down in formal procedures and regulations that specify patterns of behavior.
1202 ‘ System Dynamics '90
No real exchange of thoughts between various organization members, no sharing of mental
models, is required. In the second alternative the exchange of thoughts, whether directly or
indirectly, is explicitly aimed at. As we are interested in organizations as composed of indi-
viduals, we consider the exchange of thoughts between different organization members to be
essential for organizational learning. Organizational learning thus refers to the process in
which the perception of environmental change made by an organization member results in an
exchange of ideas (a sharing of mental models) in order for the organization to respond to the
perceived changes. As such our understanding of organizational learning corresponds with
the ‘assumption sharing approach’ to organizational learning as distinguished by
Shrivastava (1983), since from the exchange of ideas eventually a shared understanding or a
joint mental model will result.
Having discussed the problems resulting from the paradox of establishing organizational
learning, we will now turn to the question how to elicit and employ mental models for the study
of organizational learning.
USING COGNITIVE MAPS TO ESTABLISH LEARNING
Authors on organizational learning employ terms like 'mental maps’ (Hedberg, 1981, 9),
‘cognitive maps’ (Weick and Bougon, 1986), the ‘organization's policy maps' (Hall et al.,
1989), and ‘theories of action’ (Argyris and Schin, 1978) in order to deal with the phenomenon
of the learning organization. For instance Hedberg points out that: " To identify stimuli
properly and to select adequate responses, organizations map their environments and infer
what causal relationships operate in their environments. These maps constitute theories of
actions, which organizations elaborate and refine as new situations are encountered.”
(Hedberg, 1981, 7) Hence, in this paper we will focus on the use of ‘cognitive maps’ (Axelrod,
1976) to establish organizational learning.
A cognitive map can be seen as an interrelated set of assumptions with regard to some
phenomenon. In this paper we are concerned with cognitive policy maps, i.e. maps "...that are
used to direct strategy and policy making.” (Hall et al., 1989, 3) A cognitive map is usually
represented by means of a directed graph containing concepts and relationships between
concepts. These cognitive maps strongly resemble causal loop diagrams used in system
dynamics modeling. Figure 1 presents an example of a cognitive map, derived from the study
of Axelrod (1976).
Figure1 Example of part of a cognitive map
bili a
policy of | ___y- amount of security ___ » eae pe British
withdrawal =~ in Persia ew + — utility
strength of
removal of ————— > Persian
better governors “government
ee Zt
present policy of allowing Persians ability of Britain
intervention in ——+ tohavecontinued ——P to iacwure on
Persia 7 small subsidy + i
Source: Axelrod (1976, p. 61)
System Dynamics '90
1203
Cognitive maps can be derived from interview protocols or written documents. This is done
through content analysis of these documents (cf. Axelrod, 1976; Vennix, 1990). Basically this
boils down to translating statements like "increasing strength of the Persian Government
will raise the security of this country" into concepts connected by arrows as can be seen in fig-
ure 1.
Cognitive maps provide a strong means to capture a person's mental model. Research in the
field of cognitive mapping has indicated that persons act largely in accordance with their cog-
nitive maps (cf. Axelrod, 1976; Hall, 1984; Weick and Bougon, 1986). Hence, establishing a
person's cognitive map allows to arrive at conclusions about the behavior of organizational
members and consequently about organizational performance.
In the next subsection we will provide an example of the use of cognitive maps to establish in-
dividual learning from model-building.
Using cognitive maps to measure individual learning from model-building
Various authors have studied the behavior of research subjects interacting with complex com-
puter models. For instance Dérner has made an extensive study of the decisions made by re-
search subjects in simulated complex decision making situations (cf. Dérner, 1980; Dérner et
al., 1983). In a related vein Brehmer (1989) and Sterman (1989a, 1989b) have studied dynamic
decision making behavior in simulated situations in which feedback processes occur. These
reseachers primarily focus on decision-making behavior within a simulated environment.
Stated differently: the simulation is used as a laboratory in which decision making processes
of research subjects can be studied. Although these studies do not primarily focus on the effects
of learning through model-building, they shed some light on learning during the dynamic
decision-making experiments. Sterman (1989a) for instance points out that in the investment-
accelerator model subjects often start to avoid instability after three to five trials. Hence, it
seems that learning takes place, although, as the author points out, the learning seems to be
highly situational, which might be attributed to the dynamic deficiency of mental models.
Hence, "Study of the processes by which people form and revise mental models of the feedback
structure of their environment would appear to be a fruitful extension of the present research.”
(Sterman, 1989, 330).
In an exploratory study on the effects of computer models on mental models Vennix (1990) has
used cognitive maps to establish changes in mental models. In his experimental study one
condition is a computer-based learning environment, using an econometric model of the
Dutch social security system, and the other consists of traditional classroom lectures and dis-
cussions. In order to establish differences in effects between both conditions Vennix presented
both groups of students with a policy problem in the Dutch social security system and had both
groups draft a policy document on this problem both before and after the experiment.
The research design is depicted in figure 2. The policy documents written by the research
subjects were used as a basis to extract a person's cognitive map about the social security
system. This was accomplished by following and slightly adapting Axelrod’s method of
coding policy documents to extract cognitive maps (Axelrod, 1976). Next a number of
indicators for the quality of the cognitive maps were developed. These were primarily based on
the Dutch literature with regard to the quality of policy theories (Hoogerwerf, 1984) or policy
maps, i.e. the set of assumptions people have with
1204 System Dynamics ‘90
Figure 2 Research design of evaluation of individual learning from model-building
7 experimental condition:
policy document computer-based learning ee
pretest environment
licy document control condition: icy document
Fett report and classroom lectures Pstiost
Source: Vennix, 1990
regard to a certain policy problem. Three important overall criteria to establish the quality of
cognitive maps are identified: epistemological, implementary and strategic criteria. Among
the epistemological are: validity of the theory, precision of concepts and relationships, consis-
tency and quantifications. The implementary criterium includes elements like differentia-
tion of the theory, the occurrence of (manipulable) policy variables and the integration of the
theory. Finally, strategic criteria are the occurrence of time delays in the map and its accor-
dance with societal conditions (cf. Leeuw, 1983, 1986; Kraan Jetten, 1986; Hoogerwerf, 1984).
Each of these subcriteria was translated into one or more indicators with which scores, based
on the cognitive maps, could be established. For instance the length of paths and feedback loops
in the cognitive maps was employed as an indicator for the differentiation of the cognitive
map. The integration of the map was established by calculating its density: the number of re-
lationships in the map divided by the number of concepts. In sum about 20 different indicators
were employed and scores on each of the indicators were determined for all-77 research sub-
jects in the pretest and posttest. Table 1 gives a summary of the indicators used to assess the
quality of the cause maps of the research subjects.
The results of the study show that there are only minor differences between both groups in the
experiment, when it comes to changes in cognitive maps. This is due to the fact that both condi-
tions in the experiment are more similar with regard to the way information about the simula-
tion model was exchanged than was originally conceived (cf. Vennix, 1990). However, for the
two groups together clear changes in a number of indicators based on these cognitive maps
could be established due to confrontation with a computer model. For instance the number of
delays and quantifications in cognitive maps increased significantly. This is also true for
the density of the cognitive maps of research subjects. The interested reader is referred to
Vennix (1990) for a more detailed description of the research procedure and the results of the
experiment. 7
Recently new research projects have been initiated aiming at providing more insight in the
learning effects emenating from model-building. For instance Gould (1989) has a research
design in which she tries to establish learning effects from modeling using a variety of mea-
surements: subjective as well as objective, both before and after as well as during the model-
building process. Bakken (1989) is conducting a study which sheds some more light on the
phenomenon of transfer of learning from models from one situation to another. Verburgh
conducts an experiment to establish learning effects as a consequence of participation in a
computer-beased learning environment for policy making in Dutch Health Care (cf. Vennix
and Verburgh, 1990). The latter study also employs objective as well as subjective measure-
ments to explore learning effects from model-building. To our knowledge, the three latter
studies are, however, still in the stages of research designing and data gathering.
System Dynamics ‘90
1208
Table 1 Operationalization of the quality of a policy map
main crit. | subcriteria | aspects indicators
| validity - empirical - correct relationships
validity
- consistency - consistency of
relationships
precision - precision of + identical concepts
concepts - converted concepts
epistemo- - diffuse concepts
logical = specific concepts
- precision of - signed relationships
relationships = quantified relationships
- precision of ~ quantification of policy
policy options interventions
and effects - quantification of policy
effects
policy - instr. variables | - number of instrument
variables : variables
- goal variables | - number of goal variables
differenti- |- scope - number of concepts
imple- ation - number of fields
mentary - degree of detail | - length of paths
+ length of feedback loops
- extension + number of paths
- number of feedback loops
- complexity - balancedness of paths
integration - density of map
accordance - number of exogenous
with societal concepts
conditions
strategic
time factor -_number of delays
Source: Vennix, 1990, 133
Using cognitive maps to establish organizational learning from model-building: deriving
the collective map
Establishing learning effects from model-building implies that measurements are performed
at least at two different points in time. This does not only go for individual learning as we
have seen in the previous section, it also holds for organizational learning. Measuring indi-
vidual learning from model-building can be accomplished by establishing changes in indi-
vidual cognitive maps before and after the model-building process. In order to establish orga-
nizational learning empirically in a similar way as is done at the individual level (i.e.
establishing changes in cognitive maps) one would somehow have to derive an
‘organization's policy map’ (cf. Hall et al., 1989) at two different points in time: before and
after the model-building process. The question is how an organization's policy map can be
constructed. Several methods have been used to arrive at collective maps. :
One method that has been used in arriving at a collective map is ‘averaging’ individual cog-
nitive maps (cf. Bougon et al. 1977; Ford and Hegarty, 1984). This is done by taking the indi-
1206 System Dynamics '90
vidual cognitive maps, adding these and inserting concepts and relationships between con-
cepts in the ‘organization's map’ if they exceed a certain prespecified value. For instance if
more than 80% of the individuals holds a particular relationship in their mental map, this re-
lationship is inserted in the ‘averaged’ organization's map.
As we have seen, however, this way of using cognitive maps to establish organizational learn-
ing must be considered largely incorrect, since "...organizational learning is not simply the
sum of each member's learning.” (Fiol & Lyles, 1985) Or as Hedberg phrases it: "Although
organizational learning occurs through individuals, it would be a mistake to conclude that or-
ganizational learning is nothing but the cumulative result of their members’ learning."
(Hedberg, 1981, 6)
In other words simply averaging the individual cognitive maps before and after a model-
building process, does not tell much about the fact whether the organization has learned some-
thing. Averaging individual cognitive maps creates an artificial organization's cognitive
map, which might be an innaccurate representation of the organization's knowledge base.
This is also clear from the fact that the average organization's map from the pretest and the
posttest in an empirical evaluation study might be the same while individual maps might have
changed drastically in the model-building process.
Another method to arrive at a collective map is to build a composite map. For instance Eden et
al. (1983), working in a consulting mode with client groups, extract individual's cause maps,
‘add’ these together and in the group process try to arrive at a joint cognitive map. This
method, however, is not very appropriate to establish organizational learning from model-
building. This is because it basically is a group model-building process itself. In order to
establish organizational learning this way, one would first have to construct a joint map with
the group (i.e. the pretest stage). Next the group would have to go through a model-building pro-
cess and finally, it would again have to create a joint group map (i.e. the posttest stage). Hence,
the group would have to go through a joint model-building process three times. Apart from the
fact that this is very time consuming, measurements and the model-building process would
most probably be heavily interfering. This would severly impede the accurate establishment of
organizational learning effects from model-building itself.
A third way to establish a collective map would be to assemble individual cognitive maps by
linking these through common concepts in individual maps. Under this procedure one derives
individual cognitive maps before and after the model-building process, identifies common
elements in these maps and link these through common concepts.
This procedure to arrive at an organization's policy map is for instance followed by Hall
(1984). The cognitive maps of different departments in the Saturday Evening Post are linked
through common concepts occurring in these departmental maps. This in turn produces the
organization's policy map. As Hall points out, each of the departmental maps "...represents
the causal paths from policy variables (...) through intervening variables to the department's
goals.” (Hall, 1984, 914). What constitute goal variables for one department often are policy
variables for another. And as Weick and Bougon point out: "This common relevance of con-
Bite is what ties these departments together into an organization." (Weick and Bougon, 1986,
11
Weick and Bougon also point out that: "Cause maps can be coordinated with relatively little
shared understanding, a characteristic that is important to emphasize given the current em-
phasis on shared beliefs in organizational culture. Concerted action is possible where there is
common relevance of two concepts in two cause maps and a double interact to link the maps.”
(Weick and Bougon, 1986, 109-110) If this is supposed to mean that in organizations individual
or department's cognitive maps do only have to overlap marginally, i.e. only contain a few
common concepts, we disagree with them. As Hall's example clearly suggests policy making
at The Saturday Evening Post was based on only part of the total organization's map (i.e. the
map of the dominant coalition), which largely ignores important processes going on in the or-
ganization and its environment, contained in the maps of the other departments. Hall points
out that the Saturday Evening Post went bankrupt because "...the policy elite of the old
Saturday Evening Post seemed to be oblivious to the recursive relationships that tightly coupled
System Dynamics '90 1207
readers, advertising sales and magazine pages." (Hall, 1984, 923) In other words there was no
overall view which guided the policy making process. Hence, although Weick and Bougon
may be right from the point of view of organizational knowledge representation, from the point
of view of organizational performance a marginal overlap between individual cognitive
maps is not a sufficient condition for organizational survival, as Hall's example reveals.
And this is particularly true as organizations become more complex.
Using cognitive maps to establish organizational learning: from sharing mental models to
integrating mental models
A number of authors have shown that, particularly in large organizations, different persons
might have different views of the organization's problems and the way to solve these. Each or-
ganizational member perceives the organization and its environment in his own way, de-
pending on such factors as problem complexity, cognitive limitations and styles, selective per-
ception, position in the organization etc. (cf. Hedberg, 1981). The existence of different view-
points within organizations tends to be promoted by the coming of the organization of the fu-
ture, the so-called knowledge-based or information-based organization (Drucker 1988,
Applegate et al., 1988; Malik 1989). Information technology forms the foundation of the future
organization. It provides organizations with new opportunities and new threats. In order to
stay competitive, organizations will have to adapt. That is, information technology will have
consequences for the organization's structure and the formal regulations as well as for the
members of an organization. More specifically, the organization's structure will be flatter be-
cause middle management will be cut out. The disappearance of middle management will be
accompanied by a shift from traditional departments towards task-forces or teams (cf.
Galbraith 1973). In order to benefit optimally from opportunities provided by information tech-
nology, highly educated and qualified personnel is required. Members of knowledge-based
* organizations will be specialists and it is their autonomy that causes problems of control. In
* solving this problem an organization "...needs a view of the whole and a focus on the whole to
be shared among a great many of its profesional specialists, certainly among the senior
ones." (Drucker 1988, 51). In other words, organizational learning as assumption sharing
(Shrivastava, 1983), as the proces of building and modifying the view of the whole is at the heart
of the future organization. Without organizational learning, without the construction and
changing of theories-in-use the future organization will have problems surviving.
Model-building with organizational client groups exactly aims at sharing mental models
(Senge, 1989). It gets participants to exchange mental models and integrate these into one
overall perspective, revealing the organization's view on its environment. However, ex-
change and integration of various mental models is not enough. As we have seen several
authors point out that; people act largely according to their cognitive maps (Axelrod, 1976; Hall,
1984; Weick and: Bougon, 1986). Hence, in order to improve organizational learning and
adaptation to the environment individual cognitive maps have to be restructured in the model-
building process: Individual mental models need to be enriched with elements from mental
models of other members in the organization in order to expand their view of the organization
and its environment. One might expect that this will lead to actions of organizational
members which are (1) more aligned to actions of other organizational members and (2) more
in line with the interest of the whole organization.
Hence, from the empirical point of view of organizational learning this would mean that or-
ganizational learning would have taken place if cognitive.maps of organizational members
after the model-building process reveal more overlap, i.e. are more similar to each other than
before the process. For instance overlap in the use of concepts to describe the problem, or in the
kind of relationships deemed important. If measured twice, the degree of similarity between
the individual's cognitive maps in an organization would be larger after the model-building
process than before.
The study conducted by Vennix (1990) provides partial support for the hypothesis of greater
similarity between individual cognitive maps as a consequence of model-building. As stated
i
i
4
1
i
}
‘
{
Hl
i
H
i
1208 System Dynamics '90
above one of the indicators for the quality of the cognitive maps is the precision of concepts.
This is amongst others measured by the proportion of concepts which can be directly related to
one of the concepts in the simulation model during the coding process of the policy notes. In the
posttest the percentage of concepts, which could be directly related to one of the concepts in the
simulation model by the coders, is significantly higher than in the pretest. This, although pol-
icy notes from pretest and posttest were distributed randomly over coders. In the pretest this
percentage is 87%, in the posttest this percentage is 92% (N=77; p < .001). Stated differently, re-
search subjects tend to think more in terms of the concepts used in the simulation model, which
indicates that their maps have become more similar to each other. A comparable conclusion
can be drawn from considering the type of relationships between variables included in the
cognitive maps. Currently this question of similarity of cognitive maps is studied in more
depth by reanalyzing the data from the experiment using a number of criteria with which this
similarity can be established. We think of such criteria as similarity in the use of policy vari-
ables, relationships, delays etc.
SUMMARY AND CONCLUSIONS: CONSEQUENCES FOR BUILDING MODELS WITH
CLIENT ORGANIZATIONS.
In this paper we have argued that model-building with client groups can be considered to in-
duce an organizational learning process. Seen from the empirical perspective there is a vari-
ety of problems in trying to establish organizational learning empirically. First there is the
distinetion between the individual and the organizational level. Second there is the problem of
how to establish organizational learning from individual cognitive maps. We have shown
that it is not so much the changes in individual cognitive maps per se, but rather the similarity
between individual cognitive maps which should be used as an indicator for the degree of or-
ganizational learning due to a model-building process with a client group. The question now
is what this means for the process of building models with client groups. In our view there are
several important conclusions that can be drawn from this.
First, this view has profound consequences for the process of eliciting and mapping knowledge
from model-building processes. Richardson et al. (1989) have pointed out that there exist a va-
riety of techniques to elicit knowledge from individuals and groups. To arrive at a choice be-
tween these various techniques the authors introduce a number of criteria with which a selec-
tion from these techniques might be made. One of these criteria is learning form model-build-
ing. As the authors point out: "The process of eliciting and mapping knowledge to build system
dynamics models is iterative--through successive cycles of refinement the ultimate model
gradually appears. (...) So knowledge elicitation is not simply a process of uncovering a fixed
body of knowledge and representing it. Participants learn, as their mental models are re-
shaped by discussion and interaction.” (Richardson et al. 1989, 353) The authors conclude that
as @ consequence:not only written documents and individuals must be included as a source of
knowledge for model-building. Rather groups will always have to be included since it is
through discussions in these groups that mental models become shared, improved and possibly
converge towards each other. Group discussions in model-building should thus provide ample
opportunity for each of the participants to learn from others in the model-building process. In
this sense the model-building process is likely to go as slow as the slowest learner.
A second consequence that might be drawn from this paper is that model-building with client
groups should be an ongoing activity rather than provide the one and final solution to a prob-
Jem. Organizational learning itself is an ongoing process (Shrivastava, 1983). Mental mod-
els will continue to be reshaped during the process of carrying out organizational tasks. In
carrying out the day to day tasks mental models will probably become dissimilar again
through time. Hence, it might be important to repeat the process of model-building at regular
time intervals, using the ‘previous’ model as the starting point for the discussions.
Since organizations continually interact with their environment, a third point is that the
model-building process should allow to direct attention to developments in the organization's
environment and exogenous events which possibly affect the organization's functioning. The
12098
model-building process should aim at reducing uncertainty with regard to future develop-
ments in the organization's environment.
REFERENCES
Applegate, L.M., J.I. Cash, D. Quinn Mills. 1988. Information technology and tomorrows
managers. Harvard Business Review, 66 (6): 128-136.
Argyris C. & D.A. Schin. 1978.Organizational learning: a theory of action perspective,
Reading (Mass.).
Axelrod, R. 1976. Structure of decision: the cognitive maps of political elites, Princeton (NJ).
Bakken, B. 1989. Learning in dynamic simulation games: using performance as a measure.
In: P.M. Milling, E.O.K. Zahn (eds.), Computer Based Management of Complex Systems,
Proceedings of the 1989 International Conference of the System Dynamics Society, Berlin
etc., Springer: 309-316.
Beer, S. 1979. The heart of enterprise. Chichester: Wiley.
Berger, P. and T. Luckmann. 1979. The social construction of reality. Norwich: Fletcher &
Son Ltd.
Bougon, M., K. Weick, D. Binkhorst1977. Cognition in organizations: an analysis of the
Utrecht jazz orchestra. Administrative Science Quarterly, 22: 606-639.
Brehmer, B. 1989. Feedback delays and control in complex dynamic systems. In: P.M.
Milling, E.0.K. Zahn (eds.), Computer Based Management of Complex Systems,
Proceedings of the 1989 International Conference of the System Dynamics Society, Berlin
ete., Springer: 189-196.
Daft, R.L. and K. E. Weick. 1984. Toward a model of organizations as interpretation sys-
tems. Academy of Management Review 9 (2): 284-295.
Dalton, G.W., P.R. Lawrence and J.W. Lorsch. 1970. Organizational structure and design.
Homewood, IL: Richard D. Irwin, Inc. and the Dorsey Press.
Dorner, D. 1980. On the difficulties people have in dealing with complexity. In: Simulation
:__ and Games, 11 (1): 87-106.
Dérner, D., H.W. Kreuzig, F. Reither, T. Staudel (eds.). 1983. Lohhausen: vom Umgang mit
Unbestimmtheit und Komplexitét, Bern/Stuttgart/Wien.
Drucker, P. 1988. The coming of the new organization. Harvard Business Review 66 (1): 45-
53.
Eden, C., S. Jones, D. Sims. 1983. Messing about in problems: an informal structured
approach to their identification and management, Oxford/New York etc.
Emery, F.E. and E.L. Trist. 1965. The casual texture of organizational environments.
Human Relations, 18 (1): 21-32.
Etzioni, A. 1969. A Sociological Reader on Complex Organizations.New York: Holt, Rinehart
and Winston, Inc..
Fiol, C.M. and: M:A. Lyles. 1985. Organizational learning. Academy of Management
Review 10 (4): 803-813.
Ford, J.D., W.H. Hegarty. 1984. Decision makers’ beliefs about the causes and effects of
structure: an exploratory study. Academy of Management Journal 27 ( 2): 271-291.
Forrester, J.W. 1961. Industrial Dynamics, Cambridge (Mass.).
Geus, A.P. de, Planning as learning. 1988.Harvard Business Review, march-april: 70-74.
Gould, J.M.-1989. Research Summary, D-Memo D-4021, System Dynamics Group, MIT.
Hall, R.I. 1984. The natural logic of management policy making: its implications for the sur-
vival of an organization. Management Science, 30 (8): 905-927.
Hall, R.L, P. Aitchison, W.L. Kocay, H. Li. 1989.The organization's policy maps: types of
maps, methods for soliciting and recording them, and techniques for analysis. Working
document, Dep. of Business Administration, Univ. of Manitoba, Winnipeg, Canada.
Hart, S.L. 1983. Strategic problem solving in turbulent environments: a description and eval-
uation. Ph.D. dissertation, University of Michigan, Ann Arbor (Mi).
1210 7 System Dynamics ‘90
Hedberg, B. 1981. How organizations learn and unlearn. In Nystrom P. C. & W.H. Starbuck
(eds.) Handbook of organizational design: 3-27. Oxford: Oxford University Press.
Hoogerwerf, A. 1984. Beleid berust op veronderstellingen: de beleidstheorie. Acta Politica 4:
493-531.
Kim, D.H. 1989. Organizational learning and individual learning: where the twain shall
meet ?. Internal memorandum, MIT Sloan School of Management, Boston.
Kraan-Jetten, A. 1986, Reconstructie en evaluatie van beleidstheorieén. Beleid en
Maatschappij 4: 186-196.
Lane, D. C. 1989. Modeling as learning: creating models to enhance learning amongst man-
agement decision makers. Paper presented at the European Simulation Conference,
Edinburgh, sept. 1989.
Leeuw, F.L, 1983.Bevolkingsbeleid en reproduktief gedrag: een studie naar de theorieén over
gedrag die aan beleid ten grondslag liggen, Ph.D. diss., R.U. Leiden.
Leeuw, F.L. 1986. Beleidstheoretisch onderzoek toen en thans. Beleid en Maatschappij 1: 27-39.
Luhmann, N. 1976. A general theory of organized social systems. In: G. Hofstede, M. Sami
Kassem, European contributions to organization theory. Assen/Amsterdam: Van Gorcum
Malik, F. 1989. Strategie des Managements komplexer Systeme. Bern: Maupt.
Meadows, D.H., J.M. Robinson. 1985. The electronic oracle: computer models and social de-
cisions, Chichester/New York, .
Miller, E.J.and AK. Rice. 1967. Systems of organizations. London: Tavistock Publications.
Morecroft, J.D.W. 1988. System Dynamics and microworlds for policymakers. In: European
Journal of Operational Research, (35): 301-320.
Perrow, Ch. 1970. Organizational Analysis: A Sociological View. London: Tavistock
Publications.
Peters, G. 1989. Organisation und Information: iber den Zusammenhang von Interessen,
Kenntnissen und handlungen. Ph.D. dissertation, Utrecht University.
Peters, G., W.J. Scheper. 1990. Organization, knowledge and information: theoretical issues.
Working document, Dept. of Gamma-Informatics, Utrecht University.
Senge, P.M. 1989. Organizational learning: a new challenge for system dynamics. In: P.M.
Milling, E.0.K. Zahn (eds.), Computer Based Management of Complex Systems,
Proceedings of the 1989 International Conference of the System Dynamics Society, Berlin
etc., Springer: 229-236.
Schutz, A. 1973. Collected Papers I. The Hague: Nijhoff.
Shrivastava, P. 1983. A typology of organizational learning systems. Journal of Management
Studies 20 (1): 7-28.
Shrivastava, P. and J.H. Grant. 1985. Empirically derived models of strategic decision-mak-
ing processes. Strategic Management Journal 6: 97-113.
Stata, R. 1989. Organizational learning: the key to management innovation. Sloan School
Management Review 30 (3) .
Sterman, J.D. 1989a. Misperceptions of feedback in dynamic decision making.
izatic behavior and human decision processes, 43: 301-335.
Sterman, J.D. 1989b: Modeling managerial behavior: misperceptions of feedback in a dy-
namic decision-making experiment. Management Science 35 (3): 321-339.
Vennix, J.A.M: 1990. Mental models and computer models: design and evaluation of a com-
Duter-based learning environment for policy making. Ph.D. Dissertation, University of
ijmegen.
Vennix, JAM, L.M. Verburgh, 1989, Organizational learning in a computer-based learn-
ing environment. Research proposal, University of Nijmegen.
Weber, M. 1976. Wirtschaft und Gesellschaft. Tabingen: Mohr.
Weick, KE. 1979. The social psychology of organizing. Reading (Ma): Addison-Wesley.
Weick, K.E., M.G. Bougon. 1986. Organizations as cognitive maps: charting ways to succes
and failure. In: Sims, H.P., D.A. Gioia (eds.), The thinking organization, San
Francisco/London.
