You are on page 1of 4

2nd Asia-Pacific Forum on Engineering & Technology Education 1999 UICEE

Sydney, Australia, 4 - 7 July, 1999

Addressing complexity and systemic behaviour in engineering management:


A tutorial for real-life problems
K. Linard & A.C. McLucas
University of New South Wales Codarra Advanced Systems
(Australian Defence Force Academy)
Canberra, ACT, Australia Canberra, ACT, Australia
Email: keithlinard#@#yahoo.co.uk
(Remove hashes to email)
ABSTRACT: The human mind is ill-adapted to modelling of problems with delayed feedback. This paper outlines an approach
used in undergraduate and postgraduate programs in the School of Civil Engineering at ADFA to compensate for human cognitive
shortcomings when dealing with complex and systemic engineering management problems. A management problem is presented
and analysed through the medium of a tutorial. It appears easy to solve but has proven otherwise in practice. Logical and well-
intentioned management interventions result in counter-intuitive systemic responses, suggesting that the wrong problem has been
addressed, or inappropriate techniques employed. The proffered methodology, founded in research and practice, addresses problem
conceptualisation, identification of options, strategy development, and evaluation of effectiveness of selected strategies.
Fundamental competencies sought to be enhanced through application of the methodology are: a systemic and holistic perspective;
a skill-set of systemic techniques for enhancing understanding, analysis and decision-making; use of the concepts and language of
systems; and individual and organisational learning.

MISPERCEPTIONS OF FEEDBACK Subjects were unable to account well for delays and feedback
A UBIQUITOUS HUMAN LIMITATION effects because (1) peoples mental representations of complex
tasks are highly simplified, tending to exclude side effects,
Engineering management stresses the importance of project
feedback processes, delays, and other elements of dynamic
performance indicators MTBF, BCWP, BCR, etc etc. An
complexity; and (2) even when these elements are known,
implicit and fundamental assumption behind the selection of
peoples ability to infer correctly the behaviour of even simple
business performance indicators is that the feedback they
feedback systems is poor.
provide will cause decision makers to make appropriate
adjustments to the inputs or processes. This presumes decision The implication of this work is that managers require
makers can judge with reasonable accuracy the consequences specialised decision support tools to respond appropriately to
of their decisions. performance indicators impacted by delayed feedback.
However, there is abundant research in the field of system The Civil Engineering School at ADFA runs undergraduate
dynamics [1], as well as in the fields of experimental and postgraduate units in system dynamics modelling to give
economics and psychology which suggest that managers have students the intellectual and analytical tools to address such
great difficulty managing dynamically complex tasks. Sterman cognitive limitations. The following case study forms the
argues persuasively from his work at MIT that there is foundation of the undergraduate unit.
systematic misperception of feedback especially when there
THE CASE STUDY SCENARIO
are delays in the system. Mosekilde, Larsen and Sterman [2]
present the results of 48 simulations of the Beer Game [3] (a ThreeTwoOne-Ready Pty Ltd owns a small urban concrete
simulation of a simple factory-warehouse-retail system) run batching plant. Competition is stiff and ThreeTwoOne-Ready
with 192 MBA students and senior business executives. The knows that client service is paramount. An inability to respond
results show that their decision-making on the basis of fairly on time and on quality to customer demands will lose orders in
straight forward performance indicators, but in the face of the short-term, and clients-base in the longer term.
delays, resulted on average in costs 10 times the optimum!
Initially the company owned three delivery trucks and utilised
Simulations at the Australian Defence Force Academy show a additional owner-driver trucks for peak demand, for which it
similar pattern. In both the MIT and ADFA simulations, paid a premium. Company directors invested in two additional
managers and graduate students failed to comprehend the trucks on the basis of a review of demand patterns. However,
significance of feedback in the face of delay induced dynamics. even with five company-owned trucks, customer demands
cannot be fully met without supplementation. The company is
In more recent experiments, where graduate students had full
struggling to build a reputation as a reliable and responsive
information, training, incentives and opportunities for gaining
supplier. Directors of ThreeTwoOne-Ready are increasingly
experience, Diehl and Sterman still found poor managerial
anxious about the companys ability to remain competitive in
performance in the face of variations in feedback strength and
the longer-term.
delay. [4] They found the subjects were often outperformed by
a simple no-control rule. That is, totally random business A year ago the directors advised staff of the plan to invest in
decisions gave better results that the considered judgement of additional trucks, increase office staff and simplify the
their MBA students. Diehl and Sterman argue that the mental ordering workflow procedures. Instead of the expected
constructs and heuristics that managers bring to bear on accolade, they were met with a barrage of complaints about
complex tasks are fundamentally dynamically deficient: antiquated plant, excessive working hours and the ignoring of
employee suggestions on productivity improvement.
In the event, these measures did not solve the supply problems. Policy
Truck efficiency in fact decreased, while average response implementation

time to clients remained virtually unaltered. There was no


decrease in the number of client sales lost due to inability to
deliver as required. The intuitively obvious solution increased Understandings
of a system
costs, lowered efficiency and left ThreeTwoOne-Ready
marginally competitive, at best.
Policy Problem
At this point the student consulting group(s) take over. analysis definition

CASE STUDY OVERVIEW (IISD & SDM)


The tutorial provides a structured methodology, Iterative and
Interactive Strategy Development (IISD) [5] for understanding
the problem and identifying likely value adding solutions. It Simulation
System
conceptualization
supports this with a rigorous computer simulation modelling
approach, System Dynamics Modelling (SDM), to validate the
business rules of the proposed solution(s). Together these Model
formulation
management tools help build the shared vision among the
stakeholders which will underpin implementation of a solution. Figure 2: Overview of the system dynamics modelling
approach (Richardson and Pugh, 1981, p. 17)
Information
Experience Perception
c. Analytical tools and techniques to overcome the cognitive
Knowledge Motivation limitations of the human mind in dealing with complex,
systemic, dynamic problems. The capabilities of these
include:
building systemic and holistic views of the problem
Inputs from Client and rather than taking simplistic or reductionist views.
Iterative and Environment
Interactive a technical language which uses systemic concepts in
Strategy describing problems and their behaviour.
Development
Desired Outputs at a skill-set of systemic techniques for understanding,
Individual and
Organisational Levels analysing and making decisions problems
individual learning.
organisational learning in the context of group
Strategy Learning
decision-making.
Consensus Commitment skills in conflict resolution and group dynamics.
to Action
d. Underpinning the above are:
Figure 1: Iterative and Interactive Strategy Development developing an understanding of the nature of
The tutorial shows why it is critical to identify unresolved organised complexity.
issues and their influences, and how to set about solving the developing an understanding of system dynamics
'right' problem. The tutorial demonstrates that only after the techniques and their application.
nature of the problem has been established can problem
building systems thinking and systems thinking skills.
definition and conceptualisation commence.
developing insight into and understanding of dynamic
Each step of the methodology is demonstrated to the point behaviour
where a set of remedial strategies is identified. System
dynamics modelling proves ideal for both detailed analysis of QUALITATIVE AND QUANTITATIVE ANALYSIS
the problem and testing of potential strategies before The critical first step is to investigate ab initio the nature of the
implementation. The strategies are simulated and tested in a problem. The problem is subjected to deliberate elicitation of
benign environment where pre-implementation rehearsal can stakeholder domain knowledge as a pre-cursor to formal
be conducted without causing collateral damage. The whole problem conceptualisation. The aim of qualitative analysis is
analysis process requires ongoing involvement of stakeholders. to produce an influence diagram showing the influences at
An opportunity for rethinking permits reconsideration of work, the interplay of which is the cause of its dynamic
strategies before taking the critical step to implementation. behaviour. [6]
CASE STUDY OBJECTIVES
Many problem solving techniques overlook rigorous
With an overall aim of communicating ways to address the qualitative analysis or avoid it because of perceived poor
complexity and systemic behaviour in socio-technical systems, traceability through to quantitative analysis. IISD permits
the case study highlights: strong and logical development, with clear audit trails, from
a. Approaches to eliciting, recording and testing the mental qualitative through quantitative analysis.
models that stakeholders have of the problem set.
PERSPECTIVES OF THE PROBLEM
b. Approaches to eliciting, recording and testing the mental
models that stakeholders have of the problem set. Decision-makers cognitive make-up creates preferences for
particular types of reasoning. Often, those from non-technical
backgrounds are averse to quantitative analytical approaches.
[7] During the early stages of problem conceptualisation, and issues' have triggered the discontent. Whilst they may only
especially in investigation of individual stake-holdings, it is be distractors they nevertheless must be addressed before
important to gauge the stakeholders preferred modes for key organisational problems can be confronted. Without the
gathering and processing information. Cognitive mapping is a stakeholder support major initiatives are likely to fail.
valuable tool in this regard as it combines a rigorous and
QUANTITATIVE ANALYSIS OF THE BUSINESS RULES
structured methodology for capturing ideas with a qualitative
SYSTEM DYNAMICS MODELLING
analytical focus.
While qualitative analysis identifies causal relationships and
QUALITATIVE ASSESSMENT OF THE PROBLEM
feedbacks , it does not aid understanding of second or higher
CONTEXT COGNITIVE MAPPING
order dynamic impacts. Also, qualitative descriptors of the
This knowledge elicitation step collects and collates relationship between variables often disguise complexity in
perceived knowledge about the problem from key business rules. To validate the business rules developed
stakeholders. It seeks to identify and define stakeholder through the interviews, and especially to understand the
perspectives, values, ingrained assumptions, and preferences implications of delayed feedback mechanisms, quantitative
for gathering and processing information. system dynamics models is often essential.
Knowledge elicitation helps identify areas of agreement, The models are built in modular form from influence diagrams
contentious issues, misunderstandings and potential for accepted by the stakeholders. When fully integrated, modules
misinformation. It collects quantitative facts and qualitative replicate observed behaviour to a level of fidelity appropriate
impressions, beliefs and feelings. In IISD, this is done through to the problem analysis and the strategy development.
mapping factors impacting on the problem. From the
Close interaction with stakeholders is desirable throughout the
individual stakeholder maps it is then possible to develop
model development. This is particularly important when hard
aggregated, or summary, views that are the basis for gaining
data is difficult to obtain, or the behaviour of elements of the
agreement on specification of the problem. The output is an
system does not lend themselves readily to quantification.
Stakeholder knowledge and experience regarding dynamic
behaviour of the system are incorporated into the model as it
evolves. This involvement is also crucial to building
ownership of, and confidence in, strategies identified through
IISD.
The process of model building significantly aids student
understanding of feedback effects, and the counter-intuitive
consequences of delay induced dynamic behaviour.

influence diagram.

Figure 3: First pass influence diagram of ThreeTwoOne-


Readys problem produced from cognitive maps.
With revision and progressive refinement, the influence
diagram graphically maps the problem with increasing
clarity. Dominant mechanisms are identified and the links are
mapped for all to see, appreciate, challenge, and ultimately
Accept or reject. There may be further cycles of analysis and
challenging of assumptions until the diagram is in a form
broadly acceptable to all key stakeholders.
Achieving complete agreement about the nature of the problem
often turns out to be an unattainable ideal. Despite this, the
influence diagram proves to be a valuable record and cue for
more investigative discussions and the foundation for further
detailed analysis. Equally important, it provides the basis for Figure 4: (Simplified) Truck module of student System
dialogue with and between stakeholders. Dynamics model of ThreeTwoOne-Ready operations
The preferred mapping technique is based on George Kellys
time-honoured, 1955 Psychological Theory of Personal The first iterative cycle ends with the presentation to sponsor
Constructs as interpreted and employed most effectively by and key stakeholders of influence diagrams, system dynamics
Ackermann, Williams, and Eden. [8] Our experience is that models and, if appropriate, prospective strategies.
clients take readily to these diagrams and find them valuable in
During subsequent cycles model development continues. The
clarifying and summarising their ideas.
emphasis shifts to learning, by 'playing' the models repeatedly,
Often issues identified in stakeholder interviews and how and why alternate strategies lead to different systemic
preliminary qualitative analysis are found not to be 'root-cause' behaviour. Cycles of IISD continue as needed in response to
issues, but distracters. In ThreeTwoOnes case such 'side performance monitoring against model-specific metrics.
FLIGHT SIMULATOR SUMMARY
Preferred strategies are identified and then subjected to further The system dynamics unit at ADFA seeks to provide a
detailed analysis both to validate the assumptions and to fine conceptual basis for, and experience in, rational enquiry into
tune the proposed business rules. Technologies such as genetic and design of complex socio-technical systems.
algorithm optimisation may then be harnessed with the system
Its specific goal is to increase student effectiveness in
dynamics software to identify both the sensitivity of the model
analytical problem solving through increasing their level and
to different decision levers and the optimum setting of those
use of systems thinking and system dynamics modelling. This
levers. We now have a solution, but the critical step is still to
requires that they understand problem solving as a social
get stakeholder buy-in. The development of a flight
learning process, where the solution is an emergent property of
simulator is a valuable tool in achieving this.
a group enquiry. This enquiry should be based on systems
thinking principles and methodologies, with particular
reference to soft systems modelling and system dynamics
theory. This is achieved through the application of a structured
methodology (IISD) and powerful computer tools (Decision
Explorer and Powersim) to a realistic case study.

1 Sterman, J. Misperceptions of Feedback in Dynamic Decision


Making. Organisational Behaviour and Human Decision
Processes, 1989, 43(3), 301-335.
Paich, M, & J Sterman. Boom, Bust and Failures to Learn in
Experimental Markets. Management Science, 1993, 39(12), 1439-
1458.
Smith, V, G Suchanek and A Williams. Bubbles, Crashes and
Figure 5: Flight simulator control panel for Endogenous Expectations in Experimental Spot Asset Markets,
ThreeTwoOne-Ready system dynamics model Econometrica, 1988, 56(5), 1119-1152.
The flight-simulator is simply the system dynamics model Funke, J, Solving Complex Problems: Exploration and Control of
converted to a management game by the addition of a user- Complex Systems, in R Sternberg and P Frensch (eds.), Complex
friendly control panel and performance indicator display. Problem Solving: Principles and Mechanisms. Erlbaum Assoc.,
New Jersey, 1991.
Stakeholders can test the impact of their personal wish lists
and compare it with the recommended strategy. If the issue is 2 Mosekilde, E, E Larsen and J Sterman. Coping With Complexity:
sufficiently contentious, the system dynamics software Deterministic Chaos in Human Decision Making Behaviour, in J
Casti and A Karlqvist (eds.), Beyond Belief: Randomness,
(Powersim) allows the development of multi-user war-games.
Prediction and Exploration in Science. CRC Press, Boston, 1990.
Most adults learn best by doing, by trying things out in real- 3 The Beer Game is described in detail in Senge, P, The Fifth
time. However, learning by doing only works as long as the Discipline - The Art and Practice of the Learning Organization.
feedback from our actions is rapid and unambiguous. The Doubleday, New York, 1990
flight-simulator is an ideal practice field where stakeholders 4 Diehl, E and J Sterman. Effects of Feedback Complexity on
have perfect freedom to experiment with alternative strategies. Dynamic Decision Making. MIT Sloan School of Management,
This improves the likelihood of developing a shared vision for Research Report D-4401-1. March 1994.
policy implementation. 5 McLucas, A., Integrating Soft and Hard Systems Analysis:
STUDENT REACTION TO APPROACH Seeking a Practical Framework for Addressing Strategic
Problems. Proc Systems Engineering Pragmatic Solutions to
The UNSW undertakes periodic student ratings of all subjects. Todays Real World Problems. Systems Engineering Society of
The student assessment of the Systems Dynamics subjects, at Australia. Nov 1998.
both undergraduate and postgraduate levels is very favourable 6 (Coyle, R., System Dynamics Modelling: A Practical Approach.
in relation to student interest and perceived difficulty. Chapman Hall, London, 1996.
On a 1-7 scale (7 being most favourable), over 95% of 1997 7 (Nutt, P., Making Tough Decisions: Tactics for Improving
and 1998 respondents indicated that they would most certainly Managerial Decision Making. Jossey-Bass, San Francisco, 1989.
recommend the subject to other students. A significant number Churchman, C., On the Designing of Inquiring Systems: Basic
of these students, however, also rated the subject as more Concepts in Systems and Organisations. Basic Books, New York,
difficult than their other (more technical) units. Follow up 1971
questioning suggests that students had mastered the skills to Mason, R.O. and Mitroff, I.I., A Program for Research on
address technical problems, where the objectives are clear and Management Information Systems. Management Science, 1973,
unambiguous. However, they were less prepared for a 19(5), pp 475-487.
situation where there are multiple stakeholders and multiple or 8 Ackerman, F., Eden, C., and Williams, T., Modelling in
even conflicting objectives, and where the paradigm is one of Litigation: Mixing Qualitative and Quantitative Approaches.
mutual learning rather than optimisation.. Institute for Operations Research in the Management Sciences.
London, 1997.

You might also like