are interconnected such that one agent’s actions changes the context for other agents. Such systems have been the focus of intense study across a variety of scientific fields over the past 40 years (see Waldrop, 1992; Lewin, 1992; Wheatley, 1992; Kelly, 1994; Gell-Mann, 1995; Zimmerman et al., 1998; Brown and Eisenhardt, 1998). A major center of such research is the Santa Fe Institute, which includes several Nobel Prize winners among its faculty and associates (see Gell-Mann, 1995, p. xiv). Examples of systems that have been studied as a CAS include the human body’s immune system (Varela and Coutinho, 1991); the mind (Morowitz and Singer, 1995); a colony of social insects such as termites or ants (Wilson, 1971); the stock market (Mandelbrot, 1999); and almost any collection of human beings (Brown and Eisenhardt 1998; Stacey, 1996; Zimmerman, et al. 1998).
The study of such systems reveals a number of properties. Although the list below is not a comprehensive description of the field, it illustrates some key elements of a way of thinking about complex organizational systems such as health care.
Adaptable elements. The elements of the system can change themselves. Examples include antibiotic-resistant organisms and anyone who learns. In machines, change must be imposed, whereas under the right conditions in CAS, change can happen from within.
Simple rules. Complex outcomes can emerge from a few simple rules that are locally applied.
Nonlinearity. Small changes can have large effects; for example, a large program in an organization might have little actual impact, yet a rumor could touch off a union organizing effort.
Emergent behavior, novelty. Continual creativity is a natural state of the system. Examples are ideas that spring up in the mind and the behavior of the stock market. In machines, new behavior is relatively rare, but in CAS it is an inherent property of the system.
Not predictable in detail. Forecasting is inherently an inexact, yet bounded, art. For example, in weather forecasting, the fundamental laws governing pressure and temperature in gases are nonlinear. For this reason, despite reams of data and very powerful supercomputers, detailed, accurate long-range weather forecasting is fundamentally not possible. However, weather forecasting (and forecasting in general in any CAS) is bounded in the sense that we can make generally true statements about things like the average temperatures in a given season and place. The behavior of a machine is predictable in detail; it is just a matter of more study (reductionism). In a CAS, because the elements are changeable, the relations nonlinear, and the behavior creative and emergent, the only way to know what a CAS will do is to observe it.
Inherent order. Systems can be orderly even without central control. Self-organization is the key idea in complexity science (Kaufmann, 1995; Holland,