Drawbacks to Causal Loops

Causal loop diagrams (CLDs) and systems archetypes can be extremely effective for surfacing mental models and discussing potential solutions. However, they cannot effectively analyze the dynamic consequences of those solutions. As MIT professor John Sterman points out, CLDs and archetypes are working hypotheses about the interconnections in the real system, but they are not precise enough representations to be used for predicting the outcomes of decisions.

How, then, can managers examine the possible consequences of their actions and make better informed decisions? In order to vigorously test hypotheses, we must translate CLDs and archetypes into mathematical simulations. But for a non-modeler, this can be a daunting process. What resources are required? What does a typical modeling project look like? What can be learned from other companies’ experiences? Fortunately, a recent publication addresses these questions.

Modeling for Learning Organizations, edited by John Morecroft and John Sterman, is a collection of 18 essays by leading practitioners and theoreticians in the field of system dynamics. It provides practical tips, techniques, and case studies that demonstrate how modeling can be used to support learning in organizations (see “What’s Inside”). This unique volume not only provides the reader with an in-depth look at the tools of system dynamics, but it also discusses several critical modeling processes that can be used to test causal hypotheses. The varied essays included in this volume are grouped into four main sections: an overview of the model-building process, summaries of business applications, methods for sharing insights from the modeling process, and descriptions of software and modeling tools. In addition, real-world illustrations that show how system dynamics has been used effectively in a variety of corporate settings are sprinkled throughout the book.

Model-Building Process

Simulation models can greatly aid decision-making by providing a way to assess the dynamic consequences of a set of potential policies. Once managers have mapped policies of the system into mathematical equations, the computer model can show the outcome of that particular set of assumptions. The results of these simulations often illustrate the counter-intuitive nature of complex systems, which is not usually apparent in the CLDs or archetypes. By providing a way to assess the dynamic implications of various alternatives, simulation models can enhance managers’ effectiveness in designing policies and decisions that create genuine, long-term solutions.

The first part of the book explores the process of model building through essays such as John Morecroft’s “Executive Knowledge, Models, and Learning.” In this chapter, Morecroft explains how models can be used to capture, filter, and organize knowledge, and discusses how they can be used by managers to facilitate experimentation and learning.

In “Policies, Decisions, and Information Sources for Modeling,” founder of system dynamics Jay W. Forrester goes one step further to explain how models can be developed to enhance actual decision-making. Forrester defines decision-making as the process of converting information into action, and breaks it down into three distinct stages: determining the actual condition, determining the desired condition, and developing a plan to achieve the desired condition. Because such plans are generally based on specific policies, Forrester argues that if managers gain a better understanding of the dynamic consequences of their policies, they will be able to make better decisions.

Real-World Applications

The test of a good model or methodology is whether it can help make real improvements in organizational performance. In part two of the book, the editors provide a collection of case studies that describe how system dynamics modeling was used to help managers gain a better understanding of how to create high-leverage change in the systems in which they operate (see “Modeling in Action: Case Studies from the Field”).

One of my favorite selections here is “Systems Thinking and Organizational Learning: Acting Locally and Thinking Globally in the Organization of the Future,” by Peter Senge and John Sterman. This chapter describes how the Hanover Insurance Company used a learning laboratory to help its managers come to the realization that “external” problems — such as an increasing number of claims requiring litigation — were actually the result of internal policies and decisions.

Hanover Insurance’s experience began with a desire to use the tools of system dynamics to explore issues involved in claims management. First, a model was created by a team of claims managers (facilitated by outside consultants) to map the managers’ mental models of the claims management system. This model was then incorporated into a management flight simulator that could be used by a non-technical audience to conduct their own exploration of the issues. Eventually, the flight simulator was incorporated into a work-shop that allowed groups of claims managers to come together and explore the dynamics that affected their performance as an organization.

One counter-intuitive insight that came out of the workshops was that in their efforts to restrain costs by aggressively managing the number of adjusters, the claims managers were inadvertently raising their costs. The reason? By not having enough skilled adjusters on hand to assess claims accurately, the level of service quality declined, thus contributing to escalating settlement costs. Senge and Sterman not only discuss Hanover’s experience, but also draw some general conclusions from that setting about how others can design effective learning environments.

Learning from Models

As the managers at Hanover Insurance discovered, the model builders themselves learn a great deal about the system under study during the modeling process. But many teams struggle with how to share this experience with a wider group of people. In part three, learning laboratories are explored as a means to transfer learning from the core modeling team to others in the organization. Learning laboratories are a form of simulation-based training workshops in which managers use system dynamics models to help them challenge and improve their mental models. But how effective are they?

“Experimentation in Learning Organizations: A Management Flight Simulator Approach,” an essay by Bakken et al., examines how well-suited learning laboratories are for transferring learning, and offers ways to enhance managers’ ability to transfer learning from the workshop to the workplace. In the study, managers participated in two Ant learning labs — one in which they were responsible for making decisions based on anticipated world demand for oil tanker transportation, the other in which they made decisions based on the anticipated demand for office real estate. The simulators were designed so that, although the two settings were different (oil tankers vs. real estate), the underlying model was essentially the same.

For the most part, managers who had been exposed to one simulator did perform better on the other simulator, indicating that there was some transfer of learning from one setting to the next. More importantly, the more times a player went bankrupt (indicating that he or she used the simulator in a more “exploratory” mode to gain a better understanding of the underlying dynamics), the better they were able to transfer insights from one game to another. This points to a critical value of simulation modeling: participants may take risks in this “safe” environment without having their decisions pose a threat to their organizations, all the while accelerating their learning process so that their decisions back on the job will be better informed from a systems perspective.

Software

Part four includes descriptions of many products available from the major vendors of system dynamics software. It also describes how these products can be used to create models that aid organizational learning. More than just a listing of software, this section provides the beginning modeler with a set of resources for learning more about the modeling process. For example, in “Software for Model Building and Simulation,” Steve Peterson illustrates how one software product was designed to help the modeler navigate the various stages of model building — from conceptualization and model construction to model simulation analysis and communication. This section aids managers in identifying the kind of software that will best suit their application needs, while at the same time offering further insights into modeling design and execution.

Over all, Modeling for Learning Organizations provides an excellent balance between system dynamics theory and practice. These essays give the systems thinker both an introduction to the academic underpinnings of system dynamics, as well as many illustrations of how this work can be applied in the context of organizational learning.

After reading this book, managers will be better prepared to decide if mathematical modeling can add value to their systems thinking efforts. If so, the annotated bibliography can help them decide where to go next for further study and practice. Though the book is somewhat theoretical and tends toward an academic tone (some of the vocabulary may be challenging for those with no prior introduction to system dynamics modeling), Modeling for Learning Organizations is an essential resource for any manager who wants to go further into studying the science behind the learning organization.

W. Brian Kreutzer is a system dynamics author. Educator, and practitioner. He is currently working as a senior system analyst for International Computers and Telecommunications and is pursuing his masters degree in training and development at Penn State University.

Modeling for Learning Organizations (Productivity Press. 1994) is available through Pegasus Communications. Inc.

Editorial support for this article was provided by Colleen Lannon-Kim.

The post Modeling for Learning Organizations appeared first on The Systems Thinker.

Using a computer model in the resolution process alters the character of the adversarial exchange. A system dynamics model can provide an objective, “transparent” view of a complex, emotional situation. Instead of posturing, bluffing, and attempting to intimidate each other, the disputing parties find themselves challenging, debating, and even agreeing on the model’s assumptions. By making the exchange less emotional and more objective, models substantially increase the likelihood of productive negotiations and equitable settlement of disputes.

Simulations can show what would have happened if certain events or conditions had not occurred. For example, how much better would the cost and schedule performance of a project have been if the customer had not ordered certain design changes? This “would have” capability is invaluable for isolating the impacts of the factors which are the subject of the dispute.

Also essential for quick and fair dispute resolution (and for dispute avoidance) is the capability to compute the full impacts of events and conditions, including so-called indirect or “ripple” effects. A system dynamics model can represent the complex network of cause-and-effect relationship through which such effects propagate. Further, the model can fully quantify the costs of indirect effects, tracing the impacts back to specific events and conditions. It is therefore possible to isolate the effect of each event and condition and show their cumulative impacts.

What if…?

A key aspect of contract claims and many other business-related disputes is evaluating the performance of management. The “what if…?” capabilities of a system dynamics model facilitates this process. For example, simulations can compare the performance of a design and construction program based on the decisions management actually made versus the results if management had made certain decisions differently. Would costs have been lower if management had slipped the schedule earlier? What if management had expanded the workforce instead of relying on overtime?

One successful example of system dynamics modeling in dispute resolution was the Halter Marine dispute, a bitter legal battle that spanned half a decade and cost millions of dollars.

The Halter Marine Dispute

In 1979 Halter Marine, a highly respected Gulf Coast shipbuilder agreed to build a unique (and somewhat experimental) vessel called a “Catug” for Amerada Hess Corporation, one of the world’s major oil empires. The vessel consisted of a tug and a barge joined with a ridged connecting device. Halter was responsible for building the tug portion according to plans drawn up by outside architects and delivering it to Maryland to be joined with the barge.

Problems surfaced almost immediately. As construction moved forward, Halter discovered many errors and omissions in contract plans and specifications. Scheduling difficulties also arose, causing disputes between the companies. In addition, changes in the scope of the work rendered Halter’s original production, planning, and scheduling obsolete resulting in a significant decrease in overall productivity on the project. Halter management likened efforts to predict the total scope of the project to “shooting at a moving target.”

The last straw was the issue of delivering the vessels. Seaworthiness tests and trial voyages suggested that transporting the tug for delivery would create unreasonable risk to the lives of the crew as well as the physical safety of the tug. Halter decided that it was commercially impractical to transport the tug. When Hess refused to tow the barge to Alabama to join the tug there, Halter filed a lawsuit seeking damages and issued a declaratory judgement that it had no obligation to attempt a delivery voyage. Halter contended that the man-hour overruns and the delays were caused by owner-imposed changes, regulatory body reinterpretations, rework resulting from defects in the plans, and the owner’s failure to perform other obligations required by the contract plans and specifications.

Amerada Hess, denying these allegations, took the position that Halter’s problems resulted from its own incompetency, poor management, bad planning and scheduling, and generally “getting in over its head” in a major construction project for which it was not adequately equipped. It counterclaimed for damages in excess of those claimed by Halter. The stage was set for a major — and expensive — legal confrontation.

Far-Reaching Impacts

Impacts as Halter management began the task of quantifying the impacts on the Catug project of owner-imposed changes, they realized that tracking these impacts would be difficult, if not impossible. The difficulty was partly due to Halter’s accounting system which made it difficult to pinpoint the impacts on an item-by-item basis. But immediate, direct results were only the tip of the iceberg. There seemed to be far larger, indirect consequences that would be even more difficult to quantify. They resulted from changes and delays rippling through the project, producing inefficiencies and rework that were separated in space and time from their original causes. Halter management sensed that there were substantial “ripple effects” from the Catug project, but they were having great difficulty pinning them down.

Against this backdrop, Halter contacted Pugh-Roberts Associates, a consulting firm, about the potential use of System Dynamics modeling to estimate the indirect impacts attributable to the owner’s conduct. Pugh-Roberts built a computer model of the Catug project based on data obtained from Halter, interviews with Halter managers (including their quantifications of how they typically responded to the needs and conditions of projects), and relationships derived from Pugh-Roberts previous shipbuilding models. By the time the model was finished, the once-skeptical Halter attorneys had learned enough about the methodology to respect and trust the integrity of system dynamics. The methodology’s uncannily accurate performance, without midcourse manipulations, in simulating a series of complicated and interrelated phenomena convinced them of its validity.

The “base simulation” of the model correctly and accurately recreated the actual history of the Catug project. Pugh-Roberts then altered the inputs to the model to create a simulation of what would have occurred but for the changes, delays, and other owner actions cited in Halter’s claim. The only differences between the “would have” simulation and the base case were the specific inputs describing Amerada Hess’s actions. In every other respect, the model was the same.

The “would have” simulation was compared to the base simulation in terms of such factors as man-hours expended and vessel completion dates to gauge the impacts of the owner-responsible events and conditions. The differences computed from the two simulations were then “dollarized” and formed the basis of the claim against Amerada Hess submitted by Halter to the court.

The Litigation Process

During the pretrial process, all aspects of Halter’s claim — including the model — were subjected to strenuous examination. Repeated attempts by Hess’ attorneys to discredit the model were not successful. At this point, Halter’s lawyers felt that the ground-work had been laid to argue during the trial that the model should be viewed as the “objective source of correct information.” In other words, the model would be used as more than a way to estimate damages. It could be used to show how something happened which resulted in a situation far different than expected at the outset by the parties.

“In a subtle but significant way, the tide had turned. All of the parties increasingly were using the vocabulary and concepts of the model to express their views — including the judge and his influential clerk.”

In a subtle but significant way, the tide had turned. All of the parties increasingly were using the vocabulary and concepts of the model to express their views — including the judge and his influential clerk. Conclusions from the model significantly influenced the final pre-trial conference. The result was that, on the eve of trial, Halter received a highly favorable settlement.

Lawyers and the legal system increasingly are turning to alternative ways to resolve disputes. System dynamics modeling can play a key role in that process. Lessons learned from using modeling in negotiation may show ways of mitigating the costs of the dispute, and perhaps most importantly, avoiding comparable disputes in the future.

Henry Birdseye Wed is president of Pugh-Roberts Associates, Inc. Rayford L. Etherton Jr. is an attorney with Hand, Arerulall, Bedsole, Greaves and Johnston.

This article was adapted and condensed from the proceedings of the 1990 International System Dynamics Conference.

The post System Dynamics in Dispute Resolution appeared first on The Systems Thinker.