New Classification of Information: A Step on the Road to Interpretability

摘要

Complex systems, such as manufacturing supply chains, are often modeled as a collection of interacting components with information flows between them. These components are frequently responsible for making a wide range of decisions that are implemented using an optimization, heuristic, or control technique. The traditional approach to system performance focuses on the performance of these components. The view has been that to improve the system performance one had only to develop better techniques. In this paper, we argue that inadequate attention has been paid to the relationship between information and system performance. Information has played an important role in the manufacturing systems of the past. It will play a dominant role in the Internet-based manufacturing systems of the future. To better design, engineer, implement, and control these systems, we need a fundamental understanding of information and its effects on system dynamics. This paper contends that we need a new characterization of information, a delineation of its salient properties, quantitative metrics for those properties, methods for computing these metrics, and linkages between these metrics and system performance. We focus principally on the first of these, a new characterization of information, and discuss the implications of suggested characterizations for metrics and their measurement, suggesting some approaches for further research.