[ CogSci Summaries home | UP | email ]

Herbert A. Simon (1969). The Sciences of the Artificial (First Edition), MIT Press

  author =       "Herbert A. Simon",
  year =         "1969",
  title =        "The Sciences of the Artificial",
  publisher =    "MIT~Press",
  address =      "Cambridge, Massachusetts",
  edition =      "first",
  note =         "\iindex{Simon, H. A.}",

Author of the summary: David Furcy, 1999, dfurcy@cc.gatech.edu

Cite this paper for:


The thesis of the book is that certain phenomena or entities are "artificial" in the sense that they are contingent to the goals or purposes of their designer. In other words, they could have been different had the goals been different (as opposed to natural phenomena which are necessarily evolved given natural laws). Chapter 1 tackles the following issue: Since artifacts are contingent, how is a science of the artificial possible? How to study artifacts empirically? Chapter 4, on the other hand, deals with the notion of complexity. This is necessary because "artificiality and complexity are inextricably interwoven."

Detailed outline

Chapter 1: Understanding the Natural and Artifical Worlds

Natural science is very familiar to us (especially physics and biology) but the world around us is mostly man-made, artificial. It evolves with mankind's goals. So science must encompass both natural and goal-dependent (artificial) phenomena. Chapter 1 discusses how to relate these two. There are two perspectives on artifacts, synthetic vs. analytic. The science of the artificial is really the science (analytic or descriptive) of engineering (synthetic or prescriptive).

Artifacts are Fulfillment of purpose involves a relation between the artifact, its environment and a purpose or goal. Alternatively, one can view it as the interaction of an inner environment (internal mechanism), an outer environment (conditions for goal attainment) and the interface between the two. In this view, the real nature of the artifact is the interface. Both the inner and outer environments are abstracted away. The science of the artificial should focus on the interface, the same way design focuses on the "functioning".

Simulation is the imitation of the interface and is implied by the notion of artificiality. Simulation can also be viewed as adaptation to the same goal. It can be used to better understand the original (simulated) entity because simulation can help predict behavior by making explicit "new" knowledge, i.e. knowledge that is indeed derivable but only with great effort. Simulation is even possible for poorly understood systems by abstraction of organizational properties.

Computers are organizations of elementary components whose function only matters. They are a special class of artifacts that can be used to perform simulations (in particular of human cognition). They can be studied in the abstract, namely using mathematics. Yet, they can and must also be studied empirically. Their study as an empirical phenomenon requires simulation (example of time-sharing systems). In conclusion, the behavior of computers will turn out to be governed by simple laws, the apparent complexity resulting from that of the environment they are trying to adapt to.

Chapter 4: The Architecture of Complexity

In this chapter, the author notices that complexity is a general property of systems that are made of different parts and that the emergent behavior is hard to characterize.

The first part of the chapter argues that complexity takes the form of hierarchy and that hierarchical systems evolve faster than nonhierarchical ones. Very generally, a hierarchy is a recursive partition of a system into subsystems. Examples of hierarchies are common in social, biological, physical and symbolic (e.g. books) systems. In biological systems, it is argued that hierarchical systems evolve faster because the many subsystems form as many intermediate stable stages in the process. Similarly in the problem solving activity, mainly a selective trial-and-error process, intermediate results constitute stable subassemblies that indicate progress.

The second part of the chapter argues that hierarchies have the property of near decomposability, namely that (1) the short-term (high-frequency) behavior of each subsystem is approximately independent of the other components and (2) in the long run, the (low-frequency) behavior of a subsystem depends on that of other components in only an aggregate way. The example of cubicle and room temperature in a building is covered. Other examples are common in natural and social systems.

The last part of the chapter deals with system descriptions. It is argued that the description of a system need not be as complex as the system due to the redundancy present in the latter. Redundancy results from the fact that there are only a limited number of distinct elementary components. Complex systems are obtained by varying their combination. Also, the near decomposability property can be generalized to the "empty world hypothesis" that states that most things are only weakly connected with most other things. Therefore, descriptions may contain only a fraction of the connections. There are two main types of descriptions. State descriptions and process descriptions deal with the world as sensed and as acted upon respectively. The behavior of any adaptive organism results from trying to establish correlations between goals and actions.

In conclusion, a general theory of complex systems must refer to a theory of hierarchy. And the near decomposability property simplifies both the behavior of a complex system and its description.

Summary author's notes:

Back to the Cognitive Science Summaries homepage
Cognitive Science Summaries Webmaster:
JimDavies ( jim@jimdavies.org )
Last modified: Thu Aug 12 09:53:42 EDT 1999