In our hypothetical bus system, we may occasionally observe a bus running right on time. This isolated occurrence, however, does not alter the prevailing tendency of the whole system to develop timing discrepancies in the course of a day. Such tendencies become particularly powerful if the system is large and complex, as many social systems are. When we apply cybernetic analysis to these systems, we should not become disconcerted or discard the entire analysis just because one relatively minor effect is not observable in a particular society.

Cybernetics also reveals a fundamental difference between centralized and decentralized systems in their capacities for the processing of information. Decentralized organization is exemplified by the human brain, consisting of some 10,000,000,000 neurons. Such systems are highly adapted to pattern recognition and pattern processing, where huge amounts of complex data are involved. At the opposite extreme, early-generation computers, with a limited number of components, exhibit a top-down, centralized organization. Such structures are ideal for sequential single-line processing, such as calculating the first million digits of pi, but they are poorly adapted to processing complex patterns. Some of the more advanced recent computers seek to remedy this weakness through architectures based on a "neural net" model, emulating the decentralized, parallel-processing functioning of the brain. In addition, network structures such as the Internet are successful in large part because control and information flow are dispersed over numerous servers and users across the globe. Let us now examine the implications of such information-processing models for decision-making in a large society.

^Previous