coping chapter 8

8 Systems Dynamics

In the long run, the only sustainable source of competitive advantage is your organization's ability to learn faster than its competition.

From the dust jacket of The Fifth Dimension

Thinking about phenomena as a system has not been around for a long time, perhaps for the past 50 years. This should strike you as strange, even improbable, yet it is true. Terms such as feedback are relatively recent additions to our language. It took a long time to understand that despite the hints implied in electrical, hydraulic, and pneumatic circuits, there were such things as systems, and they are all around us in profusion. There have been a succession of systems approaches which have gained attention, and then more or less waned. Among the first was General Systems Theory. Then there was Cybernetics, another relatively high-level model which still finds vogue in some information warfare circles, and of which the Russians have been, and remain, especially fond. The third in this succession can loosely be called Systems Dynamics, to which the name Jay Forrester, who in the 1950s invented core memory for the early main frame computers, is associated. We will, however, focus on its foremost contemporary practitioner, Peter M. Senge of the Massachusetts Institute of Technology.

Peter Senge's book's cover proclaims "Over 400,000 copies in print!" That is a lot by any standard. The Fifth Dimension has grown with companion fieldbooks, guidebooks, and everything short of a "Cliff notes." It is perhaps the single most influential book in circles dealing with private sector management. Here is a description of systems dynamics based on an interview with Peter Senge. (1)

Systems theory is not as gray or mechanical an idea as it sounds. In fact it can be quite lively. One key to systems is nonlinear feedback-and as we've seen, nonlinear feedback can turn the simplest activity into the complex efflorescence of a fireworks display. The systems approach has taken the form of many species of theories that have evolved over the years. There is a general systems [tradition] pioneered by the late Ludwig von Bertalanffy; the cybernetic tradition begun by Norbert Wiener; and the servomechanistic or engineering tradition represented by MIT systems theorist Jay Forrester.

In its various forms and hybrids, the systems idea has been infiltrating virtually every discipline. Departments of systems have sprung up in universities all over the world.Nobel prize economist Herbert Simon announced in 1978 that he had abandoned traditional economic theory and was converting to information and systems theory. However, despite the enthusiasm, the systems approach has yet to prove itself as more than a clever new way of looking at things.

Above Peter Senge's desk at MIT's Sloan School is pinned a drawing by his young daughter. It is a swirling spasm of lines, a portrait of chaos, on which she has printed in a preschool hand, "Daddy at work." Chaos and uncertainty are indeed part of the work Senge does at the Systems Dynamics Group.[which has] taught dozens of corporations and municipalities to deal with management problems through "nonlinear" modeling.

We all have countless models in our heads about how things work. "If the car starts to skid, turn your wheels in the direction of the skid"-that's a model. "Spare the rod and spoil the child"-another model. Some of our models involve feedback but generally not the kind of iterated (positive) feedback that makes for nonlinearity. In business and economics the theoretical models used for planning have traditionally been linear. "Increase the sales force and we'll increase the number of sales," or "Take the growth rate for the last five years and project it for the next five years after compensating for population declines."

But linear models are notoriously unreliable as predictors, which is their usual function. Forecasts don't work out. The population suddenly starts to grow or moves to another part of the country or starts buying less of a product because of some unforeseen reason, such as a gas crisis. Attempts to make predictions suffer a chaotic fate. The predictions fail because the models can't take in the whole of how the elements in sensitive dynamical systems interact. Systems Dynamics' answer to this modeling dilemma was to make the essence of the model nonlinear and to shift the emphasis away from prediction.

Nonlinear models differ from linear ones in a number of ways. Rather than trying to figure out all of the chains of causality, the modeler looks for nodes where feedback loops join and tries to capture as many of the important loops as possible in the system's "picture." Rather than shaping the model to make a forecast about future events or to exercise some central control, the nonlinear modeler is content to perturb the model, trying out different variables in order to learn about the system's critical points and its homeostasis (resistance to change). The modeler is not seeking to control the complex system by quantifying it and mastering its causality; (s)he wants to increase her "intuitions" about how the system works so (s)he can interact with it more harmoniously.

Thus the development of the systems model exemplifies the shift that the science of chaos and change is making from quantitative reductionism to a qualitative holistic appreciation of dynamics.

How is a qualitative model made? When they work in complex organizations such as corporations, System Dynamics modelers try to identify the written and mental concepts the people in an organization are using when they do their work, the organization's rules and policies, the actual behavior of people in the organizational setting, the organizational structure, its purpose, and numerical data such as how many people are working and when they work. The goal is to see what kinds of loops these elements form. The process of making a nonlinear feedback model is itself a nonlinear feedback process. See Fig. 8.1.

"Initially clients are skeptical," Senge says, "'You can't model this; this is not just a system of hard variables. We are talking about innovation, passions of man, all sorts of subtle, unmodelable things.' Their first position is always cynicism. But after a while they get enthusiastic. They see you can model the psychology and the subtler dynamics that go on in an organization. They find that if you can talk about something clearly, you can usually model it, and they get enthused about modeling the subtler dynamics that everybody knows are important."

The tangle of feedback loops is often immensely complex, of course, but the computer can handle that. Nonlinear equations are assigned to the loops to indicate the precipitous things that happen as values are powered up ("loop gains") or diminished.

What is purposely left out of the model are the "historical," or "time-series," data used by linear modelers to compute the ups and downs of past trends the organization has experienced. The nonlinear modeler uses the time-series data not to make the model but to check it. By running the model on the computer, the modeler can see how close his or her picture of the organizational feedback comes to behaving the way the actual organization behaved historically. [This is a good example of meshing, or interweaving, linear and nonlinear techniques, which are superior to either alone.]

One advantage claimed for a good model is that you can change the values in different loops, run the simulation on the computer and see what happens. You can try out a policy change, watch the effect on the system of adding staff or cutting staff; you can experimentally change the relationship of different elements, even gauge the possible result of a difference in employee morale or attitude. Because it's difficult for a human mind on its own to visualize any more than a very few loops, the computer is indispensable to the modeling process.

By studying systems' complex and varied forms, systems theorists have developed a long list of systems' principles. Below are a few, summarized by Peter Buttner, an executive for the Boise Cascade Lumber Company and a former student of Senge's at MIT:

-To permanently change a system you have to change its structure.

-In any given system there are very few "high-leverage points" where one can intervene to produce significant, lasting changes in the overall behavior of the system.

-The more complex the system, the farther away cause and effect usually are from each other in both space and time.

-It doesn't take very many feedback loops before it gets tough to predict the behavior of a system.

-Neither the high-leverage points nor the correct way to move the levers for the desired results tend to be obvious.

-"Worse before better" is often the result of a change of a high-leverage policy in the "right" direction; therefore any policy change that produces better results immediately should almost always be suspect.

Senge, for one, believes that we are only just beginning to understand how to handle such complexity on the social level. He says that when he teaches people how to model systems he starts with "a degree of complexity just within the bounds of your conscious ability" and then escalates the complexity until people dimly grasp the whole without actually being aware of it. He thinks learning to handle complexity means learning to live more intuitively, because intuition is the key to making significant changes in complex systems, helping them evolve, and evolving with them.

"At the deepest level of systems dynamics we are trying to cultivate a unique intuitive/rational sense of when we are getting close to a leverage point. It rarely has any correlation to the symptoms most people focus on, because in a system cause and effect are rarely closely related in time and space."

The point of people immersing themselves in the complexity is, he believes, to liberate their visions. You want to change the system so that it expresses your unique angle on things. But the problem is you can't do that mechanically because your unique angle isn't a reducible item; its more of a feel, a nuance. So to get at a vision, the system has to be approached as a subtle whole. The task, as Senge describes it is obviously not an easy one for minds trained in reductionism. He says that "there's an incredible tell-me-what-I-can-do-so-can-fix-it attitude" that people have about organizations. "We're trying to teach people the systems perspective and part of that is assimilating the ability to grow from acknowledging uncertainty. You're always in an experimental mode. I think it's enormously powerful. It liberates the vision side of things. It also liberates the intellect. In education it lets people operate in a learning mode rather than a fix-it mode, which makes them a hell of a lot more effective intellectually."

However, he admits that while people get insights from systems dynamics, they often don't stick with the process. "I think in the back of their minds is the thought that despite their insights, somewhere along the line they're going to get this reduction, this model of the system which then they'll be able to change mechanically. After a while they see there's no end to this modeling process, the intuitive process, and they get discouraged. The nature of what we're doing doesn't fit with their assumption of a reductionist solution."

[Reproduced by permission of Harper Collins Publishers, Inc. via the Copyright Clearing Center.]

Bear in mind that each successive attempt at systems theory has been launched in the linear domain to probe the nonlinear world, and each got better at it by introducing elements that are more native to the nonlinear environment. Systems dynamics is getting close. Yet, even Senge is concerned that it, like its predecessors, may be on the wane, in which case a more successful and persuasive form is likely to follow. Yet, the struggle will be uphill, because this aid to learning employs higher level models, always operating at the edge of the envelope which nonlinearity will permit, compared to lower-level, more tacit models such as that of Van Creveld's and Perrow's.

The future of higher level models, such as systems dynamics remains problematic, because they seem to continually bump their head on the low ceiling that nonlinearity permits for entry into its domain. Should systems dynamics yet hit on the right dimensions, becoming just tacit enough for the nonlinear aperture, it has the potential to be a powerful Aid to Learning.

Next - Chapter 9