MODELLING MEANING, NOT VARIABLES: TOWARDS AN
INTERPRETATIVE MODELLING OF SYSTEM DYNAMICS
Rueylin HSIAO
Assistant Professor
Department of Decision Science
NUS Business School, National University of Singapore
FBA2 (#03-14), 17 Law Link
Singapore 117591
Tel: (65)-874-3074 Fax: (65)-779-2621
Email: rueylin@nus.edu.sg
Nature of the submission: Plenary session Paper
Submitted to the International System Dynamics Conference 2001 Atlanta, Georgia,
USA July 23-27, 2001
11June 2001
MODELLING MEANING, NOT VARIABLES:
TOWARDS AN INTERPRETATIVE MODELLING OF SYSTEM
DYNAMICS
Abstract
This study suggests a rethinking of qualitative system dynamics modelling. The results
highlight that “interpretative modelling” is a useful way to enhance the use of system
dynamics when encountering a situation coloured by social, cultural and political
factors. The paper examines the problem embedded in the current use of system
dynamics and proposes a three-level analysis (process, influence diagram and frame) to
show how interpretative modelling can be attempted. It argues that, for qualitative
inquiries, researchers need to consider interpretative modelling that emphasises more on
surfacing meaning rather than on building variables.
Keywords: interpretative modelling, knowledge elicitation, process theory,
influence diagrams and frame analysis
Introduction
Since its inception, system dynamics researchers have developed a strong pool of
knowledge to understand better non-linear problems though quantitative simulation of
feedback loops. Recent developments in this field come to recognize the limit of
quantification and propose to use qualitative methods (e.g. combine system dynamics
and soft system method) to produce policy insights (Coyle and Alexander, 1997).
However, scant attention thus far is paid to examine the underlying worldview rooted in
both use of system dynamics.
This study therefore aims to explain why current applications of system dynamics are
entrapped by a positivistic worldview, an assumption which seeks to generalize
“objective” theories through testable hypotheses. As shown in this study, researchers
may inevitably oversimplify the organizational realities by embracing the positivistic
assumption. Three major problems are addressed: (1) the problem of knowledge
elicitation (Does the system dynamics model represent the “real” problem under
investigation? Or does it merely reflect the modellers’ bias?), and (2) unit of analysis
(Should researchers focus on finding constructs for hypotheses-testing? Or should they
analyse the meaning of participants’ mental models?). These two problems seek to
sensitise researchers to a social-scientific modelling of system dynamics.
The paper thereupon proposes an analytical framework seeking to respond partially to
these three challenges. The framework assumes an interpretative worldview and
considers a three-level analysis with a combinational use of process theory, influence
diagrams and frame analysis. An in-depth field study of IT failure is used for illustration.
First, the sequences of events are traced to offer a process story, identifying the patterns
that explain how problems emerge and evolve over time. Secondly, these process
patterns are used to build a system dynamics model, with a focus on “meaning” (how
participants perceive these problems). Thirdly, a frame analysis is used to examine how
the mental models (of participants) contribute to the problems exhibited in the system
dynamics model. Finally, the limitations of this three-level method are reflected.
Conceptual Basis
Although more and more researchers begin to employ qualitative system dynamics to
understand policy dilemma (e.g. Sterman et al., 1997), few critically examine the
underlying assumption of system dynamics method. A common understanding of
qualitative system dynamics is to include “soft” constructs such as trust, leadership, and
motivation. System dynamics researchers rarely consider the “worldview” behind the
use of non-linear modelling. It is also the dominant view that a rigorous, scientific
system dynamics study must be accompanied by statistical justification and correct
mathematic formula embedded in the model. This view of qualitative system dynamics
is nonetheless bias, if not wrong. The qualitative research discipline of organizational
science might be useful to provide some hints to system dynamics researchers.
To provide a conceptual basis for this proposal, one question is proposed to offer a lead
for the intellectual debate: i.e. should system dynamics be described as a
hard/deterministic system approach? The purpose of this discussion is not to provide
the “right” answer of what qualitative system dynamics should be. Instead, it aims to
suggest an altemative way to consider a qualitative method of deploying system
dynamics.
Should system dynamics be described as a hard/deterministic system approach?
The question is to revisit Lane’s (2000) defence on system dynamics as a deterministic
system approach. Lane (2000) argues that researchers mistreat system dynamics as a
hard science because there is a lack of theoretical understanding. There are four major
“accusation” of system dynamics modelling. First, system dynamics is a nave method
in assuming that future events can be prophesised. Second, system dynamics assumes
complete control of the decision of human agents. Third, system dynamics assume that
there are only cause-effect laws, ignoring issues derived by human subjectivity. Fourth,
system dynamics is just another form of system engineering which is operationally
austere and coercive. Lane (2000: 18) concludes that the misunderstanding is due to
poor communication between system dynamicists and other system practitioners.
Although Lane (2000) argues in length to counter these accusations, unfortunately, few
qualitative researchers might agree with his view. The reason is neither about the
methodology adopted nor the measurement taken. The poor communication perhaps lies
in the paradigmatic gap between a positivistic and interpretative worldview (for detail
description of paradigm and worldview, see Blaikie, 1993; Burrell and Morgan, 1979).
In order not to fall into abstract debates, I will try to illustrate this paradigmatic gap by
discussing three issues relating to system dynamics modelling. I will then revisit why
current system dynamics is considered as under positivistic paradigm.
The first issue is related to the problem of knowledge elicitation. This is to ask: does the
system dynamics model represent the “real” problem under investigation? Or does it
merely reflect the modellers’ bias? Often, it is the trained system dynamics modellers
(or consultants) who will interview informants and decide what the problem is about,
although they claim to be neutral. One manager from client site (a petroleum company
in UK doing system dynamics exercise) interestingly called it “Apotheosis of Model
Building”. This refers to the criticism of the God-like role of researchers as modellers.
In most situations, although statistical justifications have demonstrated the objectivity of
the model, it nevertheless assumes that modellers have a final decision on which
variable should be included and which causal relationships should be specified. Most
experienced managers would wonder how one could play the role of God in deciding
what should be the “final model”, even the method of group model building is used
(Vennix et al., 1996). In a real life context, this is often a matter of power struggle
between modeller and key stakeholders. As this is not merely a technical but also a
social issue, it frequently requires reconciliation rather than measurement.
Moreover, there is another issue of inter-subjectivity (Mitchell, 1983). The system
dynamics modellers bring in a subjectivity based on their expertise background (e.g. if
the modeller is from technology discipline, he/she will tend to consider solution from a
technology viewpoint). The informants interviewed, from different departments, also
bring with them different kinds of subjectivity, if not mention the vested interest. The
interaction of modellers and informants again create inter-subjectivity. In this case, will
the mathematic formula truly reflect the complex interaction of subjectivity involved?
Therefore, system dynamics can be considered as a “hard science” if it fails to address
the issue of knowledge elicitation from a qualitative angle.
The second issue is about the unit of analysis. This is to ask: Should researchers focus
on finding constructs for hypotheses-testing? Or should they analyse the meaning of
participants’ mental models? A typical process of system dynamics modelling includes:
(1) defining the problem by collecting data, (2) defining constructs to represent the
problem, (3) formulating hypothesis about the reciprocal causal relationship among the
constructs, (4) building and optimising the model, (5) analysing problem through
simulating the system dynamics model, and (6) identifying an optimal solution through
simulations. Such a build-test-solution process seems too good to be true. We might
wonder: how can this systematic build-test-solution process guarantee an “optimal
answer”? Most companies encounter complex problem coloured by organisational
politics, culture, and power. It requires researchers to investigate qualitative concepts
such as “trust” and context-sensitive constructs such as “leadership” (a leader will
influence how a problem can be solved in an organisation). Without getting an in-depth
understanding of the issue under investigation, researchers might in fact come to a
solution that oversimplify the problem. In particular, it would be difficult to get an
in-depth understanding if researchers do not understand the problem perceived by the
stakeholders. Without recognising this, researchers might create a model that has no
meaning to the context-specific problem.
Should system dynamics be described as a hard/deterministic system approach? The
answer I am afraid is “yes”. According to the yardstick of social science, the current
system dynamics modelling effectively embrace a “positivistic worldview”, a paradigm
(way of thinking) that consider the world as ordered universe made up of atomistic,
discrete and observable events. The positivism view regards true knowledge to be
tepresented by universal laws: i.e. only which can be observed can be regarded as truth
science (Blaikie, 1993). It holds that knowledge is derived from sensory experience by
means of experimental analysis. Science is to gain predictive and explanatory
knowledge of the external world. In contrast, interpretivism view sees “reality” (true
science) as the product of processes by which social actors together negotiate the
meanings for actions and situations (Blaikie, 1993). “True knowledge” thus is be
derived from the everyday social world in order to grasp the socially constructed
meanings, and reconstructs these meanings in social scientific language.
According to Waring (1996), hard systems approach relates to those situations in which
human behaviour is perceived to play a minor role, even though many people may be
involved in the system. People are assumed to be objective in any situation. Hard
systems approach refers to attributes perceived to be quantifiable, predictable and
relatively undisputed. It involves a set of tacit assumptions on the part of
problem-solvers, which may be summarised as followers:
« The existence of the problem may be taken for granted.
« The structure of the problem can be simplified or reduced so as to make its
definition, description and solution manageable.
« The reduction of the problem does not reduce the effectiveness of the solution.
e Anoptimal or superior solution does exist
e The selection of the optimal solution is through a rational process of
comparison.
These assumptions indicate that a hard system view of problem-solving involves a very
detailed examination of the system experiencing the problem. For hard systems analysis
to be effective, there will also have to be a large measure of agreement concerning the
overall goal. The role of human actors is assumed to be that of passive objects amongst
whom consensus exists. The major criticism of hard systems thinking concems its
deterministic view of social systems which sees individuals performing deliberate acts
and imparting subjective meanings. Indeed, if we take a laymen approach to look at
system dynamics, we might wonder how valid is it by building a mathematic formula or
assigning a 1-10 scale to constructs such as “love”. Perhaps, we want to know more
about the meaning of love rather than measuring the quantity of love.
Therefore, even soft system method or soft OR is used to support system dynamics
modelling (e.g. Coyle and Alexander, 1997; Lane, 1994; Lane and Oliva, 1998;
Rosenhead, 1989), it still does not change the positivistic assumption. Not until this
paradigmatic issue is addressed, researchers might not be able to communicate to each
other about what is the true model to represent the true knowledge of the selected
phenomenon. To explore how qualitative system dynamics can work, we need a new
way to experiment with system dynamics and take a new perspective to consider the
influence of human actors. The next session makes such an attempt.
Research Method
Case Selection
The case is an international firm - FoodInc (disguised name) - in a consumer product
industry based in Asia. The selection of this case is based on the principle of theoretical
sampling (Eisenhardt, 1989; Yin, 1989) for two reasons. First, FoodInc has continuously
invested millions of dollars working with various consulting firms on large-scale
IT-enabled change projects. However, the top management remains baffled by the
dilemma of ineffective investment in IT-enabled change. Secondly, these changes have
resulted in the resignations of many senior staff, and the problems at the operational
level seem to be escalating. Divisional managers in general feel that the company’s
overall capability in dealing with these problems has gradually dwindled. Thus, the case
provides a rich context for studying IT failure characterised by high causal and
human-induced complexity. In this case, the critical events within the organization
(based on the Diary Division) are used as the focus of analysis.
Data Collection
The data collection traced the developmental path of IT introduction, following
interviews from front-line staff to the top management team and tracing the value chain
activities horizontally (ie. from R&D to customer service). The overall interview
scheme is illustrated in Table 1. The data collection is designed to trace the change
incidence retrospectively, including a real-time intensive field visit (spanning one month
from August to September 1997), a series of follow-up semi-structured interviews, and
another site audit (spanning three weeks in A pril 1998). In addition, during the site visit,
the researcher also attended many of the departmental lunch meetings in order to
appreciate the problem and context. Informal talks to product managers and front-line
staff also helped to understand the organizational climate that contributed to the IT
failure.
R&D Production Sales/Retailers Headquarters
Management team 18
Divisional Managers}1 2 3:
Middle Managers |1 4 15
Frontline workers {1 6 28
Sub-total 3 12 46, 8
[Total 169 persons
TABLE 1. Interview Scheme for Data Collection
Two main sources of information were collected. First, process data were collected with
reference to content and context over time (Pettigrew, 1990, 1997). This is mainly
concemed with retrospective tracing of different change initiatives. Secondly, data
conceming the subjective interpretations of key stakeholders actors was collected,
including the perceptions of the top management, Strategic Planning Division (SPD, a
key policy designer), IT Division, and product managers (they are users mostly based in
the Diary Division). This data was used to understand how conflicting frames lead to
resultant actions (Schén and Rein, 1994). Data were gathered through semi-structured
interviews (around 2 hours for each interview), participant observation in intemal
meetings, document study (intemal archives and consultancy reports), and two group
interviews.
Data Analysis
Based on interpretivism tradition (Walsham, 1995), three particular techniques are used
in the data analysis: processual analysis (Pettigrew, 1990, 1997), system dynamics and
frame analysis (Schon and Rein, 1994). First, processual analysis involves the use of
ethnographic narratives to capture the organisational dynamics and a process map to
understand how problems evolve and accumulate over time. The detail documentation
of ethnographic narratives is provided in a working paper (Hsiao, 2000), while the paper
sums up only the key events in the case analysis.
Second, influence diagram is used for system dynamics modelling (Coyle, 1996;
Wolstenholm, 1990) to understand the underlying causal pattem of change. However, it
should be noted that the modelling process used in this study is slightly different from
the current quantitative (e.g. Sterman, 1989) and qualitative approach of modelling (e.g.
Wolstenholm, 1990). The quantitative modelling approach, in principle, follows the
positivist paradigm and emphasizes the measurable factors of a system, without paying
sufficient attention to the complexity of human interaction. This approach is more
concerned with the production of a universal framework for prescribing remedial
actions. For example, if analysts control the “morale factor’ in the system dynamics
model, the system performance will achieve certain optimal outputs.
On the other hand, qualitative modelling stresses the collection of behavioural data,
which aims to understand how the dynamics of the problem evolved. Nevertheless, the
current qualitative approach seems to focus more on the construction of a system
dynamics model and less on the description of problem in organizations. This study
attempts to adopt qualitative modelling approach by incorporating a “thick description”
of the social dynamics involved (Geertz, 1973). The purpose is to examine the detail of
human interactions in the context of IT-mediated organizational change. In this way, the
influence diagram model is used mainly as an interpretative device to add to the
explanatory power of the case study. This feedback loop analysis, provided by influence
diagram modelling, offers an effective way of representing the reciprocal relationships
of the problem under investigation.
Third, the analysis revisits the processual data and system dynamics model in order to
reflect on how problems are caused by the conflicting frames of human actors. This
helps to understand the root cause of IT failure in this particular case. The research
framework shown in Figure 1 explains three levels of analysis in this study.
Typical Questions Level of Understanding Research Methods
How does the firm
react to this problem Events situated in
over time? Contexts
Processual Analysis
What kinds of (Pettigrew, 1990, 1997)
patterns of events
seem to be recurring? Process of Events
Influence Diagrams
What are the Analysis
underlying causes (Coyle, 1996)
that create the
patterns? Underlying Patterns <
What are the stated Frame Reflection
and unstated frames (Schén and Rein, 1994 )
which generate the
underlying pattern?
Conflicting Frames | ¢——J
FIGURE 1. The Research Framework for Data Analysis (based on Kim 1992)
Validation Issues. The issue of validity raised in this study is complex. It involves
modelling subjective cognition, which gives rise to the problem of inter-subjectivity. In
order to minimise the impact of the researcher’s own bias and key actors’ “attributional
egotism” effect (see Brown, 1998: 52, which refers to the phenomenon wherein actors
offer self-serving explanations for events, attributing favourable outcomes to their own
effects, and unfavourable outcomes to external factors), reflective interviews and group
meetings were used to incorporate key actors’ comments. Triangulation is also achieved
through the use of multiple informants (in different divisions and ranks) and data
sources (retrospective vs. real-time data and field data vs. archive data). In addition,
some key informants were interviewed 2-3 times to examine the coherent of their claims.
Their personal career in the company is mapped and compare to the overall process of
change. This helps to validate their accounts of change events.
However, the researcher retains the final decision about mapping the change process
and model building, with reference to the hidden agendas informed by site visits and
informal conversations. Five guiding questions are repeatedly raised in different ways to
informants in order to achieve data triangulation (Forrester, 1993). These questions are:
(1) What was done in various kinds of past problems? (2) What are the self-interests of
social actors? (3) Where are the influential power centres in the organisation? (4) What
could be done in various hypothetical situations that have never happened? (5) What is
being done to help solve the serious problems facing the company?
Case Analysis
Context
The case is based on the study of IT failure in a consumer products company - FoodInc
(with its headquarters based in Taiwan). The globalisation challenge has brought about a
series of changes in FoodInc. Since its establishment in 1967, the company has grown
into an international organization which employs 6,200 people in the core businesses,
owns 52 factories around the world, and has strategic alliances with over 70
internationally known firms (up to 1997). Its business scope ranges from animal foods
(e.g. stock-feed), consumer foods (such as plain flour, meat, frozen foods, and
beverages), chain stores, distribution, construction, electronics, semiconductors, and
financial services to leisure enterprises.
The company’s aim was to integrate its core competencies to achieve successful
globalisation. The management team intended to upgrade the legacy information
systems in order to support future business growth. With the assistance of various
consulting firms, FoodInc invested in a series of IT initiatives during the period
1989-1998. Internally, the firm had two key change agents - SPD (Strategic Planning
Division) and Information Division - to facilitate the transformation for over 72
business divisions. However, these initiatives were not entirely beneficial; rather, they
seemed to create more trouble throughout the organization. The empirical investigation
is mainly based on the introduction of IT in Dairy Product Division.
Process
The events are summarized into 11 episodes to illustrate the meaning interpreted by key
stakeholders in the implementation process (see Table 2). In each episode, the emerging
conditions of context (first column) are explained and the dominant actors’ perceptions
are “interpreted” (second column). This reflection of the frame of references of social
actors helps to understand why particular actions are resulted (third column). The
purpose of this analysis is to show that reciprocal causality is derived not necessarily as
“rational”. The structural constraints may be “irrational” and “emotional”. Finally, the
outcome of these frame-induced actions is provided in the fourth column. The 11
episodes help to enhance our understanding of the situations in the context of IT failure
in FoodInc. The process story is to assist the building of system dynamics model.
However, the qualitative system dynamics model is used more as an interpretative
device to deepen the understanding of IT failure problem, rather than a predictive model
to forecast behaviour under structural control.
Episode
Contexts and situations
‘Dominant actors’ perceptions
Resulting actions
(Outcome
1
[The outsourced ISD project was
abandoned (1985). Operational
\bottleneck was seen as a key issue.
ISPD: ISD must be managed in-house.
IT Dept. was expanded (became ID). The
WANG system was used for hardware platform
land COBOL language was used for ISD.
Users complained that the problem stemmed
from the new IS. Operational bottleneck was still)
unresolved. ID gained more power over SPD.
2. [The top team pressured SPD to ISPD: Users’ complaints are only SPD concentrated on pushing ISD schedules. | Users’ complaints were continuously aroused.
resolve operational bottleneck. temporary. ID eoineentrated on haniweane and software here was sill no sign of productivity
Users’ complaints were mounting. ID: We must prove our worth; IS would |integration for ISD, paying more attention to P
work well within the parameters of a _|system coherence.
\coherent IT infrastructure.
3 |A new CEO proposed an initiative: |SPD/ID: Smart-work culture could be —_|SPD urged ID to expand the local exploitation of| Users became less tolerant and more resistant to
to transform the old hard-work lachieved by introducing smart machine _|IS into a company-wide implementation. the IS. ID expedited the ISD schedule but did
Iculture into a smart-work culture __|(i.e. the IS). not consider the redesign of the outmoded
(1989). process.
4 Users’ resistance was mounting. |SPD: Something must be done to show SPD installed groupware to demonstrate how |The use of groupware further promoted hostility
CEO pressured SPD and ID to lways of implementing “smart-work”. _|smart-work could be achieved. and distrust among users toward SPD. However,
a . ‘ the feeling was not recognized by SPD. ID was
deliver results. ID: To smooth complaints, we must first |To gain sympathy from users, ID developed a _|tio4 up in fixing the problems generated b
’ confidence. \data-mining system to help users retrieve data ip peng te Dl 9 y
Sar USEES, COI lfrom POS system breakdowns. Under the pressure to
‘Users: ISD only means more workloads. . develop several software applications at the
|Users considered groupware was a distraction same time, the workload of ID staff increased
from their work. and IS quality suffered.
5/6 Old culture persisted in FoodInc. User: The IS was not useful; itis merely |Users (product managers) were engaged in |Another round of ISD interviews added to users’
Users’ skepticism towards smart
machine was rising.
(There was a lack of senior product
managers in the consumer goods
lindustry in the Asia Pacific region.
doing the wrong thing faster. SPD/ID
lwere spending money on entertainment.
ID: Better technology was needed.
\product failure problems.
ID focused on system coherence and upgraded
IT infrastructure (1993/94); later, ID was
lengaged in data conversion and software
lredesign.
workloads. As the workloads were increased
consistently and the career systems remained,
‘many product managers left and joined
FoodInc’s competitors. Conflict and distrust
between users and SPD/ID were aroused.
Conventions: SPD = Strategic Planning Division; ID = Information Division; IS = Information Systems; ISD = Information Systems Development; IT = Information Technology.
Table 2. Tracing the Frame-induced Conflicts in Foodinc.
Episodes}
\Dominant actors’ perceptions
(Resulting actions
(Outcome
va IID: We needed to demonstrate our technical ID decided to shun Window-based applications Users decided to ignore ID’s ISD efforts.
competence to gain user confidence. because of system consistency (1992); ID later
User: IS would not solve productivity problems; Concentrated on system migration (1998-94).
|PC-based application is a better solution. |User started to use end-user applications (e.g. Excel).
8 \User: We did not get any productivity improvement |User decided to bypass ID and hire its own (Distrust arose between users and ID.
from the ISD. rogremmners Divisional rectors sent junior staff to System quality suffered further because
ID: Why didn’t they appreciate the importance of : jof poor user inputs; most users gave up
system coherence? We need to enhance ID began several user-communication programs; ithe use of IS in response to the outbreak
communication. llater, ID was again engaged with the tasks of data __|of system problems.
lconversion.
19 SPD: The previous IS failure was due to changing —_|SPD: decided to transfer built-in best practice via__‘[The rising conflicts resulted in users’
luser demands; ID was not capable of dealing with it. |ERP systems (1997) and; SPD set up committees to resistance to the ERP implementation.
. lget top management support. Users felt they were not!
Ex? astens wer ppd in a fae ae ink en ee
|the various industries in the Asia sur
. ipport.
\Pacific Region.
10 Users distrusted ID’s competence|SPD: Best practice transfer would solve all the ISPD pushed the use of ERP software aimed at [The overall organizational climate was
lin delivering viable IS and |technical problems and smooth the complaints from reengineering supply chain processes and transferring |filled with conflict, distrust and
doubted SPD’s intention to lusers. Ibest practices. de-motivation. Product managers’
introduce ERP systems. \Users: SPD is squandering money on projects that are|Users decided to concentrate on product-related Roriloads meressed, Tesla pm set
lincapable of producing major results. Our concern is {problems and ignored IT-related tasks. lindi ‘foals towards The ‘onsultants,
|to resolve product-related problems. Se J .
11 [The increasing staff tumover ISPD: We needed something new and interesting to _|SPD decided to introduce e-Business to enhance the [As more and more senior staff left, the
caused the loss of organizational
|knowledge. The concept of
e-Business became a new trend,
replacing ERP.
lengage users.
|Users: These smart machines (ERP and e-Business)
were just expensive calculators.
|ERP-based reengineering project. Users tumed into
\clandestine resistance; they wanted to protect their
trade secrets” from being computerized into
le-Business.
operational problems continued. Conflict
jand distrust persisted in the
joryanizational climate. The use of IS,
junder the banner of
\“smart-work-via-smart-machine”, was
jsuffocated.
Table 2: Tracing the Frame-induced Conflicts in FoodInc (continued).
Structure
In FoodInc’s case, a system engineer may attribute IT solution backfire to poor system
development. A product manager may blame retailing policies and ineffective processes.
On the other hand, the SPD may prioritise the need to align IT and business strategy. An
organizational development consultant may emphasize the resolution of conflicts
between SPD and business divisions in order to smooth the implementation of change.
Each cause-and-effect inference is right, but the dilemma lies in the synthesis of all. For
instance, if the IT manager implements a better software engineering method, this may
speed up the system development cycle; but it may also cause an increase of staff
workload, leading to more staff tumover and intemal conflicts. In addition, the technical
difficulties may increase, and systems may become even unstable. Furthermore, if the
SPD introduces better consultants to assist the strategic planning, this may provide a
coherent design for integrating IT and business strategy; but it may also escalate the
internal conflicts and delay the remedies in distribution channel, given the context in
FoodInc.
Analysing the Underlying Pattern. To provide a viable process theory, analysts need to
reflect upon the recurring patterns of events. This requires an examination of the
processual data presented in the case and identifies the reciprocal effects of the
context-specific constructs. Feedback loops are used to illustrate the problem of IT
solution backfire from a systemic perspective.
The IT Solution Loop. In the early stage, FoodInc’s inefficient processes incurred
operational bottlenecks and an increase in operating costs. To regain competitiveness,
the SPD initiated a series of IT-enabled change which included distribution information
systems, the WANG hardware systems, in-house developed software, a major upgrade
in IT infrastructure (into Oracle RDBM platform), and supply chain management. The
“IT Solution Loop” (see Figure 2) represents this feedback effect. IT-based solutions are
applied to improve operational bottlenecks. If operational bottlenecks remain, more IT
solutions are needed.
IT Proficiency
among Staff
Quality of User
Participation Level of Trust
Accumulation of
Embedded s
Problems in Feedback
Processes + Effect
Accumulation of toop:2 Conflicts
IT System 7 between SPD
Problems Staff and Divisions
“P™* Occurrence of IT Tumover
Technical +
Problems Job Market +
Demands
\t
va of + Staff Workload
IT Solutions at Division
Level Ht
é Stickiness of
Old Culture
seivion g Product Failures
Loop
Process
. Efficiency
Operating
Costs 0 peraonala
ws oreiona 7
FIGURE 2. The Dynamics of IT Failure in FoodInc.
The First Feedback Effect. The provision of IT solutions unexpectedly increases the
staff workload at a divisional level (see “Feedback Effect Loop 1” in Figure 2). The
level of workload is initially maintained by the firm’s ineffective work practices, which.
include, for example, tedious meetings held regularly at a divisional level and outmoded
administrative processes (e.g. product managers have to share a fax machine to receive
and send orders). The result of the continuous provision of IT solutions is that product
managers have to deal with both their ineffective routine tasks and the added IT-related
jobs. Meanwhile, the level of workload is sustained by two key contextual factors - the
old way of working and constant product premature death (see the two factors
“Stickiness of Old Culture” and “Product Failures”). In consequence, such an
accumulated workload invariably decreases the process efficiency in operations, leading
to more operational bottlenecks.
From the SPD’s standpoint, the remaining bottlenecks demand more IT solutions. For
example, this may mean shifting from IT outsourcing to in-house design, and
introducing a third party consultancy. This effect forms a reinforcing loop (see Feedback
Effect Loop 1) that perpetuates operational bottlenecks, urging the SPD to implement
more IT solutions. Inevitably, this initiates another cycle of workload increase, further
process inefficiency and more operational bottlenecks.
The Second Feedback Effect. “Feedback Effect Loop 2” explains a second reinforcing
effect of the IT-induced dilemma, which explains the conflicting perceptions between
the user side and the supply side. The increased workload leads to rising conflicts and
accumulated distrust between the SPD and users. In addition to the lack of IT-related
knowledge of users, the poor quality of user participation also worsens the problem.
This results in more embedded problems of ineffective processes. Once these problems
are designed into the information system, they lead users to report system unreliability
as a result of technical IT problems. They also lead the Information Division to interpret
these problems as system incompetence, thus leading to more changes of IT
infrastructure. In addition, frequent changes of IT infrastructure in the name of system
coherence mean more work on system migration (e.g. on translating data structure from
COBOL to RDBM), and this leads to an accumulation of further IT problems (see the
factor “Accumulation of IT Technical Problems”). Altogether, the SPD and the
Information Division feel a stronger need to resolve the technical problems by
providing more IT solutions.
Another unintended consequence is the intemal conflict among business divisions (the
user side), the Information Division (the supply side) and the SPD (the mastermind
side). Initially, referring to the “Feedback Effect Loop 2”, the effect of the increase staff
workload (at the divisional level) leads to an increase of conflict between the SPD and
users. This has a second reinforcing effect on the staff workload, the conflicts between
SPD and business divisions, the quality of user participation, the accumulation of
embedded problems in processes, and IT technical problems, thereby perpetuating the
system instability. In general, users feel that the SPD’s fruitless IT solutions jeopardize
their performance in sales and interrupt their routine work. Moreover, the continuous
failure of IT solutions results in a distrust of the SPD’s competence in introducing
information systems (see the factor “Level of Trust”). A major consequence of this is
that it invites more conflicts and users are less willing to participate in the design of
information systems (see the factor “Quality of User Participation”). The decreasing
quality of user involvement leads to two major problems.
1. Because users (in particular the senior product managers) are not fully involved
in redesigning processes, they provide only partial information to system
analysts. Many problems are still embedded in these ineffective processes. When
system analysts fail to incorporate these problems into process redesign, these
problems are less detectable. As a result, these process-based problems are
translated into technical problems (see the factor “Accumulation of Embedded
Problems in Processes”). From a user’s perspective, information systems are not
reliable and their instability becomes ever more difficult to tolerate.
2. When users later find that IT consultants are paid astronomically and SPD staff
have abundant resources (to travel abroad, for example), their commitment tums
sour. This leads to their alienation from subsequent projects. Users come to
provide system analysts with the wrong specifications to sabotage the whole IT
initiative.
There is another noteworthy contextual factor: the “Proficiency of IT among Staff”. In
FoodInc, most staff lack IT-related training. This makes it more difficult for users to
articulate their real information needs. Users can only explain old processes (how things
have already been done) to system analysts rather than persuade them to consider the
underlying policies (such as the retailing policy). Furthermore, the participation of
novice staff also undermines the quality of system development. However, system
analysts assume that users can objectively and correctly articulate their system
requirements. These specifications, once designed into various information systems,
will only automate the incorrect administrative processes, thus causing further
IT-related problems. Moreover, system analysts have to spend more time dealing with
these IT-related problems, rather than investigating the fundamental process-related
issues. As a result, when information systems are used in divisions, users often find that
they are unstable, and hence more “IT-related” problems are discovered. Again, users
then report these IT-related (technical) problems to system analysts for further
improvement. This then urges the SPD to provide more IT solutions.
The Third Feedback Effect. The increased workload at the divisional level also has a
third reciprocal effect on FoodInc’s IT dilemma (see “Feedback Effect Loop 3”). The
increase of staff workload makes employees suffer from demoralization and family
pressures. As the job market offers more attractive packages, more and more senior staff
are tuming to competitors. The tumover of senior staff means losing organizational
knowledge, which is equivalent to the loss of years of industrial experience in handling
the supply chain, product management and relationships with retailers. The pressure of
senior staff turnover and the accumulating workload force divisional managers to start
sending junior staff to participate in user requirement meetings in order to alleviate staff
tumover and allocate resources to more urgent problems - product failures stemming
from ineffective departmental coordination.
Hence, solutions that fail to recognize the reciprocal nature of change may lead to more
undetected problems, thus merely shifting problems from one part of the system to
another. In some situations, a solution may become a problem of its own; at worst, a
problem may become buried in the historical context when those who handled the first
problem were replaced by those who inherited the new problem (Morecroft, 1985;
Senge, 1990).
Moreover, understanding the nature of reciprocal causality may often develop
counterintuitive insights by observing the social dynamics in terms of feedback
behaviour (Morecroft, 1985; Richardson, 1991). For example, the enhancement of user
participation may eventually accumulate embedded problems in processes if users have
little IT knowledge. The introduction of another strategic exercise by consultants may
only make product managers more resistant to any changes that are brought in by the
SPD. The introduction of IT solutions may lead to a heavier workload rather than
resolving the operational bottlenecks, if the problem of the old culture and product
failures is not considered. By appreciating the reciprocal causality of a problematic
situation, analysts can effectively reflect on the complex interaction of problems and
sources of dilemmas, thereby producing enduring improvements.
Outcome
There are four major indicators of understanding the IT dilemma in FoodInc. First, the
top management were puzzled by the enormous investment in IT and consultancy
services, which seemed to have little positive impact on the firm’s performance. In the
light of the conflicts among the policy designer (SPD), the IT provider (Information
Division) and users (various divisional managers), the top management seemed to lose
confidence in implementing more changes. In 2001, the company try to implement
another software in the hope of using a better technology to resolve the adoption
barriers.
The second condition was the high tumover of senior staff. As the number of
resignations increased, divisions faced major sales difficulties and suffered from low
morale. This in turn tended to cause further resignations. Divisional managers were very
worried about such a vicious circle. The third symptom related to a more intangible
measure of conflict and morale. The conflict between SPD and divisional managers
seemed to escalate because the two parties disagreed about the allocation of resources.
More and more divisional managers were seeking to implement changes by themselves,
thereby neglecting those organized by the SPD. The fourth problem was in fact even
more worrying. The technical difficulties of the IT systems seemed to rise steadily. The
unsuccessful investment in IT-enabled change led to a loss of top management support.
Increasingly, there were signs of a significant rise in the workload of senior staff, the
level of inter-departmental conflicts, and operating costs. The management team
believed that immediate remedies must be sought to resolve operational bottlenecks,
and thus promote productivity.
Discussion: Hard SD vs. Soft SD, a New Perspective
The discussion addresses two objectives: 1) to explore the practicality of applying soft
SD (system dynamics) and 2) to summarise the differences between solving ‘dynamic’
issues through hard SD and soft SD from the author’s viewpoint. The purpose of the
discussion is to suggest ways to bridge the classical use of positivistic SD to that of
interpretative SD modelling (see also Dyer and Wilkins, 1990).!
Hard/Positivistic SD Soft/Interpretative SD
Data Acquisition measurable hard variables (e.g. non-measurable soft variables (e.g.
inventory and revenue) motivation and competitiveness)
quantification in the relationships of alignment among the interactive
variables feedback loops
Model Construction models the world models individual perception
focuses on hard facts/constructs focus on subjective meaning/stories
Model Analysis conducts hypothesis testing to reach identifies dominant logic to obtain
policy recommendation in-depth understanding and leverage
points
Ultimate Concerns generalisable laws transferable insights
aims to achieve optimum portfolio of aims to achieve intellectual efficiency
solutions
Figure 2. The differences between positivistic and interpretative SD modelling.
The first aspect explores the practicality of qualitative SD application. The use of
system dynamics in this project leads to several implications for management. These
may be discussed in tum as follow:
1. Soft SD is an effective way to enhance group intelligence (GI). The synergy
between individuals can be increased significantly using system dynamics model to
trigger strategic debates. In the focus group session with informants representing
| The use of term (qualitative and quantitative) can be confusing. Here I will use soft SD to refer to
interpretative SD and hard SD to refer to positivistic SD.
different interested parties, the SD model helps to consolidate controversial views
which are implicitly embedded among the informants. To discuss safety issues in a
feedback loop manner helps informants to challenge their existing mental
perceptions. One interesting anecdote describes a debate between an pilot and a
aircraft designer. They start with serious argument, accusing each other of causing
threats to safety to and later explore the idea that the problem exists in the
‘structure’ rather than with people. On such an occasion, a group of intelligent
experts may reach an unintelligible conclusion. Qualitative SD can thus be very
useful in bringing together diverse viewpoints.
The strategic debates based on the mental model promote consensus which in tum
enhances the quality of decision making. Informants become more tolerant on
controversial issues through the process of revealing their disagreements. Thus, an
airline manager, through the focus group process, recognises the need to
implement remedial policies to deal with pilot subculture problems. Before
reaching such a consensus, although he reads articles regarding similar issues, he
does not actually recognise that the problem is caused by the subculture, but in
stead emphasises human errors.
Qualitative system dynamics provides to be an effective way for identifying the
source of organisational dysfunction. This is especially true when several
stakeholders are involved in the process. The managerial errors caused by the
interaction of the pilot community, airline companies, aircraft manufacturers and
government agents have previously been discussed but not recognised. The
dysfunction in the structure can be identified through the qualitative SD model in
order to explore the ‘helpless’ syndrome (like beer game) caused by the overall
system structure.
The visualisation of SD models helps to increase the capacity of mental information
processing. Because human brains are not able to process too many interwoven
relationships at the same time, the visualisation used by the system dynamics
method can assist the appreciation of complex problems.
Qualitative system dynamics asks a different type of question, exploring subjective
meaning rather than quantifiable measurement and evaluation. The criteria of an
effective use of qualitative system dynamics depends on the insight that can be
provided by the modelling, but not on the facts derived from the building of
equations and figures. As Richmond (1993) suggests, problems can always be
quantified but can rarely be measured. Modellers need to use mediating variables to
study the system behaviour indirectly.
The second aspect concerns the difference between quantitative and qualitative
approaches to SD modelling. This has been a paradigmatic debate in SD field. The
emphasis of the dominate quantitative SD approach is on experiments with quantifiable
variables, using historical data as reference modes to create equations embedded in
models, and thus to simulate behaviour. Quantitative SD modelling is recognised as the
most convincing approach for management decision makers (see, for example, the work
of Coyle, 1996; Roberts, 1978; Forrester, 1961). However, as Wolstenholme (1990)
suggests, in order to relate SD to a wider audience, the subject of qualitative system
dynamics needs to be further developed in order to capture generic insights from many
SD models in a condensed qualitative form. This qualitative form offers a powerful
means to disseminate insights and enhance learning in relation to complex situations.
The Table below compares these differences in terms of data acquisition, model
construction, model analysis and ultimate concems. Although Figure 8 does not aim not
to explain the whole spectrum of differences, it may nevertheless provide a reference
point for bridging these two approaches in future research.
The third aspect relates the learning of system dynamics to the wider field of OR. At
first, it was not clear to the author have the ‘hard’ OR modelling techniques might be
related to the of ‘soft’ ones. During the process of research, however, the author found
that it was useful to understand these relationships in terms of the nature of problems:
hard vs. soft and static vs. dynamic. Hard problems refer to problems which are
structured and can be well-defined: for example, the material requirement problem in
production systems. Soft problems refers to problems which are ill-defined and cannot
be measured in a clear way as they also involve a lot of external factors (such as oil
crisis, recession or the break of war). For example, the measure of competitiveness in
firms is regarded as a ‘soft’ variable. On the other hand, the difference between the
nature of static and dynamic problems relates to their inherent complexity and effects
over time. For a static problem the effect of feedback depends on the portfolio of input
variables and the formulae built in the process mechanism. The complexity is linear and
can only be traced over a specific period of time. For dynamic problems, the effect is
often interwoven between the interdependent variables and cannot be discerned easily.
This kind of problem is often controversial and conflict-based. In Figure 9 the
framework is proposed to explain this concept. Although the framework requires further
refinement, it serves as a basis for discussing the implications of OR according to the
varying nature of problems. It also offers a reference point for identifying the
appropriate OR techniques to be applied to the right type of problem.
Conclusion
The paper attempts to offer a new perspective of using soft system dynamics. It
proposes a synthetic method based on interpretative paradigm by employing process
and frame analysis. In this way, system dynamics model is used more as an
interpretative device which aims to convey meaning rather than merely the constructs.
The article proposes a different approach: (1) acknowledge an interpretative paradigm,
(2) conduct ethnographic data collection, (3) see processual pattems of SD from events,
(4) incorporate meaning/stories into the structure, and (5) seek not an optimal solution
but maximum understanding. In this way, the paper suggests that SD has much to offer
to qualitative research discipline as a new way to understand problems characterized by
dynamic causal relationships. Lastly, it should be noted that such an interpretative
modeling aims to complement, and not replace, current system dynamics practices.
References
Blaikie, N. 1993. Approaches to Social Enquiry. Cambridge: Polity Press.
Brown, A. D. 1998. "Narrative, Politics and Legitimacy in an IT Implementation." Journal of
Management Studies 35(1):35-58.
Burrell, Gibson and Gareth Morgan. 1979. Sociological Paradigms and Organisational Analysis:
Element of the Sociology of Corporate Life. London, UK: Heinemann.
Coyle, R. G. 1996. System Dynamics Modelling: A Practical Approach. London: Chapman & Hall.
Coyle, R. G. and M. D. W. Alexander. 1997. "Two Approaches to Qualitative Modelling of a Nation's
Drugs Trade." System Dynamics Review 13(3):205-22.
Dyer, W. G. and A. L. Wilkins. 1991. "Better Stories, Not Better Constructs, to Generate Better Theory: a
Rejoinder to Eisenhardt." Academy of Management Review 16(3):613-19.
Eisenhardt, K. M. 1989. "Building Theories From Case Study Research." Academy of Management
Review 14(4):532-50.
Forrester, J. W. 1961. Industrial Dynamics. Cambridge, Mass: MIT Press.
. 1993. "System Dynamics and the Lesson of 35 Years." Pp. 199-240 in The Systemic Basis of
Policy Making in 1990s, (ed.) K. B. De Greene. London: Kluwer.
Geertz, Clifford. 1973. The Interpretation of Cultures. New Y ork: Basic Books.
Hsiao, Rueylin. 2000. "Why IT-Enabled Change Fails." National University of Singapore Working
Paper .
Kim, D. H. 1992. Toolbox Reprint Series: System Archetypes. Cambridge, Mass.: Pegasus
Communications.
Lane, D. C. 1994. "With a Little Help From Our Friends: How System Dynamics and Soft OR Can Learn
From Each Other." System Dynamics Review 10(2-3):101-34.
. 2000. "Should System Dynamics Be Described As a Hard/Deterministic System Approach?"
System Research and Behavioral Science 17:3-22.
Lane, D. C. and R. Oliva. 1998. "The Greater Whole: Towards a Synthesis of System Dynamics and Soft
Systems Methodology." European Journal of Operational Research 107:214-35.
Mingers, John C. 1984. "Subjectivism and Soft Systems Methodology." Journal of Applied Systems
Analysis 11:85-103.
Mitchell, J. C. 1983. "Case and Situation Analysis." The Sociological Review 31:187-211.
Morecroft, J. D. W. 1985. "The Feedback View of Business Policy and Strategy." System Dynamics
Review 1(Summer):4-19.
Pettigrew, A. M. 1990. "Longitudinal Field Research on Change: Theory and Practice." Organization
Science 1(3):267-92.
. 1997. "What Is Processual Analysis?" The Scandinavian J ournal of Management (Special Issue on
Conducting Process Research) Autumn:337-48.
Richardson, G. P. 1991. Feedback Thought in Social Science and System Theory. Philadelphia: University
of Pennsylvania Press.
Richmond, B. 1993. "Systems Thinking: Critical Thinking Skills for the 1990s and Beyond." System
Dynamics Review 9(2):113-33.
Roberts, Edward B., (ed.). 1978. Managerial Applications of System Dynamics. Cambridge, Mass.:
Productivity Press.
Rosenhead, J. 1989. Rational Analysis for a Problemative World - Problem Structuring Methods for
Complixity, Uncertainty and Conflict. Chichester: John Wiley & Sons.
Schon, D. A. and M. Rein. 1994. Frame Reflection: Toward the Resolution of Intractable Policy
Controversies. New Y ork: Basic Books.
Senge, P. 1990. "Catalyzing Systems Thinking in Organizations." Pp. ?? in Advances in Organization
Development, F. Masarik. Norwood, NJ: Ablex.
Sterman, J. 1989. "Modeling Managerial Behaviour: Misperceptions of Feedback in Dynamic Decision
Making Experiment." Management Science 35(3):321-39.
Sterman, J. D., N. P. Repenning, and F. Kofman. 1997. "Unanticipated Side Effects of Successful Quality
Programs: Exploring a Paradox of Organisational Improvement." Management Science
43(4):503-21.
Vennix, J. A. M., H. A. Akkermans, and E. A. Rouwette. 1996. "Group Model-Building to Facilitate
Organizatonal Change: An Exploratory Study." System Dynamics Review 12(1):39-58.
Walsham, G. 1995. "Interpretive Case Studies in IS Research: Nature and Method." European Journal of
Information Systems 4:74-81.
Waring, Alan. 1996. Practical Systems Thinking. London: Thompson Business Press.
Wolstenholme, E. F. 1990. System Enquiry: A System Dynamics Approach. Chichester: John Wiley &
Sons.
Yin, R. K. 1989. Case Study Research: Design and Methods. Newbury Park, California: Sage.
