Enterprises as Complex Systems

M.L. Kuras

Enterprises are "complex systems." Until recently, such a characterization would have been no more than an admission that enterprises are hard to comprehend. Today, however, to say that something is a "complex system" is beginning to mean something very specific. Complexity theory is providing a new way of thinking about the relationships among components of a complex system.

The study of complex systems began to gain real momentum roughly 20 years ago. There are now several institutions (for example, the Santa Fe Institute and the New England Complex Systems Institute) and an international conference devoted wholly or principally to the concept.

The field of complex systems includes a number of distinct mathematical and conceptual tools. "Chaos Theory" focuses on the sensitivity of deterministic non-linear systems to initial conditions. "Catastrophe Theory" focuses on abrupt interruptions in certain kinds of continuous functionalities. (Cellular) "Automata Theory" focuses on the implications of simple local rules repetitively applied. Neural nets, genetic algorithms, and agent-based modeling are additional tools in this field.

Multiscale analysis is another such tool. It calls attention to the fact that the perceived behavior of a system depends upon the "scale" of the perspective used to study it. Not all patterns of behavior are visible at every level of scale; but some are. Patterns that are apparent when certain aggregates are considered seem to disappear when their parts are considered separately, or when just the relationships among specific individual parts are considered. Such patterns can only be associated with the interdependencies among all of the parts taken collectively. This helps to explain why the behavior of biological organs like hearts and lungs cannot be deduced or inferred from even a complete understanding of the behavior of individual cells—even if all of their cells are understood in this way. And it's why "an enterprise" cannot be fully appreciated simply by closely watching what every single person in it does as a person—either manually or with the aid of automation.

There are two fundamentally different ways to approach the study of systems, complex or otherwise. One is purely analytic. The other is developmental or synthetic. In the analytic case, the focus is on understanding what already exists as a system. In the synthetic case, the focus is on creating (or developing or synthesizing) a system. However, it is no great leap to observe that efforts to synthesize benefit greatly from analysis.

Traditional systems engineering focuses on the synthesis of systems. An important approach to the synthesis of systems today is known as systems engineering. The "traditional" variant of systems engineering relies on a recursive or fractal model of systems. The whole is exactly the sum of its parts; the parts can always be resolved into smaller parts that can be treated in an identical fashion if necessary. Traditional systems engineering relies on a closed, complete, precise, and a priori specification of desired behaviors (including outcomes). The current notion of a "system of systems" is an assertion that this recursive model can also be scaled upward indefinitely to address ever more elaborate systems.

Traditional systems engineering can be understood as an expression of a "Newtonian" world-view. Everything works like a clock. No matter how elaborate, everything can be resolved into arrangements of interlocking gears of varying sizes and orientations governed by fully predictable laws. Build and position its gears just so, wind it up, and then it runs exactly as expected—forever. Traditional systems engineering is focused on capturing and detailing desired outcomes, then working backward to the proper gears and their interconnections, and then finally and progressively detailing, realizing, and integrating these parts until the desired outcomes have been achieved.

The emerging field of complex systems suggests that this notion is incomplete. Enterprises, for example, just don't behave like clocks.

Of course, since the time of Newton, many additional discoveries have been made. These include, for example, stochastic processes, relativity, quantum mechanics, and evolution. While it would be presumptuous to suppose that the implications of these ideas are fully understood today, it is also unquestionably true that a fuller and more accurate understanding of the real world must take these ideas into account. To the extent that these and other new ideas bear on the analysis and synthesis of systems, the study of complex systems is an attempt to update our world-view to account for these ideas.

In the traditional view, an enterprise is a system. It is complex only to the extent that it has a large number of parts—ultimately people and their tools. Given sufficient time and effort, an enterprise's workings should be completely resolvable and understood as the workings of these people—either directly or through progressive groupings of these people. Conversely, any desired enterprise can be constructed solely from the behaviors of many individual people if the enterprise's outcomes are properly and thoroughly characterized beforehand. A major difficulty with this view is that increasingly the behaviors being sought from enterprises are highly interactive with their environments (which can only be described, not specified), are frequently stochastic and not fully predictable beforehand, and whose internal structures change over time. A full and complete specification (analytical or synthetic) of modern enterprises is not just impractical—it is actually impossible in the traditional sense.

Real complex systems (real enterprises) are not closed and static. There is insufficient time, even with maximum theoretical efficiency, for an enterprise to specify its own behaviors and outcomes before they change, even just as a consequence of that specification process.

The assertion that an enterprise is a complex system implies that if it is desired to synthesize or modify an enterprise, a fractal or recursive approach will not suffice by itself. Additional and complementary methods, combining an analytical multiscale approach with a synthetic evolutionary methodology, should augment traditional systems engineering. Rather than focusing exclusively on the realization of a single specification (which remains a valid and useful thing to do up to some level of complexity), evolutionary engineering first presumes that synthesis of a single complex system will proceed in multiple concurrent development tracks. It provides for the identification and application of developmental precepts as a framework for continuous contextual discovery among these parallel tracks (rather than relying only on a priori specifications) as well as the formulation of generalized goals and constraints at multiple aggregate levels (identified through multiscale analysis). The precepts must be independently applicable in each parallel developmental track. It must be possible to apply these precepts repetitively and locally while specific local objectives are also pursued. However, the violation of constraints imposed on the multiple parallel development efforts in the aggregate will terminate or restart individual efforts. This decoupling of the development into parallel tracks enables larger systems to be constructed without a priori specification of decomposition and behavior. Synthesis is allowed to proceed so long as the constraints are not violated, and the generalized measures of progress are seen, not only if certain preconceived outcomes are eventually achieved.

The implications that will flow from an understanding of an enterprise as a complex system will affect the management, organization, and engineering practices of modern enterprises, as well as the outcomes of these enterprises.

For more information, please contact M.L. Kuras using the employee directory.

Solutions That Make a Difference.®
Copyright © 1997-2013, The MITRE Corporation. All rights reserved.
MITRE is a registered trademark of The MITRE Corporation.
Material on this site may be copied and distributed with permission only.

Privacy Policy | Contact Us