The Sciences of the Artificial (3rd ed. 1996)
by Herbert A. Simon (1916-2001)
This book should be part of the education of anyone concerned with design: engineers, writers, artists, executives, physicians, politicians. Simon shows how our understanding of artifacts must differ from our understanding of the natural world (the concern of the “natural” sciences, even in the naïve form of most non-specialists). The study of the natural addresses how things are; the study of the artificial addresses how things ought to be. Hence the sciences of the artificial are moral, utterly different from the natural sciences.
This report will mention some highlights, special insights of Simon’s that I hope illustrate good reasons for reading the book.
Simon points out that the task of the natural sciences is to show that the world, in its wonder-inspiring variety and detail, is not incomprehensible – but not to destroy the wonder. When the order at the root of the wonderful is discovered, we wonder anew at the way in which the complex arises out of the simple.
For most people, most of our experience is not with the natural world, but with the man-made artificial: the disinfected water we drink, the conditioned air we live in, the clothing we wear, the buildings and roads and vehicles we use, our communication tools, and the symbols transmitted through them.
He proposes four characteristics of the artificial, of which the third seems most important to me: Artificial things can be characterized in terms of functions, goals, adaptation. The goals, of course, are those of the designer, or the customer that the designer serves. Of course, the design of an artifact is constrained in two ways by the natural world: it’s construction must obey natural law, and the environment in which it must operate. Accordingly, one may view an artifact as an “interface” between the “inner environment” of its construction, and the “outer environment” for which it is designed. In many cases, the design is motivated by this interface; the details of the inner environment are often irrelevant, and different artifacts with utterly different internal details (e.g., a mechanical or electronic timepiece) can perform the same functions equally well in the same outer environment. A major part of the activities of design and invention consists of the description of an artifact in terms of its organization and functioning – its interface. The designer often partitions the desired behavior into a system of simpler behaviors. Designing components that perform these simpler behaviors can be easier, and can be relatively independent of the internal workings of the other components.
An artifact is rational, in the sense that its design is the design it ought to have in order to perform the functions intended by its designer in the anticipated environment.
Simon discusses rationality in terms of economics. The two major classes of artifacts here are markets and organizations. (He says that 80% of human effort is within organizations, rather than between them. This has large implications for theories of the efficacy of markets in determining economic behavior.) Individuals interact in certain ways with both types of artifacts. Simon was among the first to point out that humans are incapable of performing the necessary computations to act perfectly rationally in economic systems (formerly a major assumption of economic theory). Rather than rationally optimizing economic behavior, humans “satisfice” in their decision-making, they attempt to make good-enough decisions based on limited information and limited expenditure of time and other resources, satisfying their desires to the extent feasible within fairly strict constraints.
Simon addresses human psychology with a series of experimental results selected to illustrate this thesis:
Human beings, viewed as behaving systems, are quite simple. The apparent complexity of our behavior over time is largely a reflection of the complexity of the environment in which we find ourselves.
Some of the findings he mentions:
- Human long-term memory (LTM) is essentially unlimited.
- Human short-term memory (STM) is limited to the number of items that can be rehearsed in about two seconds, typically five to nine “chunks”.
- The size of a chunk depends on the kind of “template” available to hold information (e.g., telephone numbers, chessboard positions).
- Templates can often be described as structured lists of lists of items (e.g., linguistic analysis of sentences).
- The time needed to move a new item (or chunk) from STM to LTM is about ten seconds.
- The time needed to associate an item with an existing template is about one second.
- The time needed to retrieve an item from LTM, given another item “associated” with it, is about one second.
- Problem solving involves attending to items and strategies in LTM in a serial manner.
- Two kinds of processes are available for problem-solving: recognition (fast, based on recall) and search (slow, based on construction of new templates and following a serial protocol).
- Highly skilled performance (e.g., chess grandmaster, medical diagnosis, music composition) depends on the accumulation, indexing, and practice with using about 50,000 templates of items.
- The effective acquisition of 50,000 templates of items requires about ten years of sustained effort.
- When a domain of knowledge grows so large that it cannot all be learned within a decade, it will split into specialties, or its practitioners will rely increasingly on external memory aids (e.g., books, computer storage).
- Memory contains two kinds of items: data (e.g., structured templates) and processes (e.g., condition-action productions). Expert knowledge can use both redundantly.
Simon spends a lot of time on the utility of hierarchy. It provides an organizing principle for understanding the natural world, and for understanding existing artifacts, as well as for organizing requirements for new artifacts, and the associated components of a design to meet the requirements. He gives arguments for believing that hierarchical designs of single-celled creatures as well as higher organisms must be favored by a satisficing process, such as evolution. Similar arguments apply to human organizations.
Simon applies and illustrates his ideas in the domain of social planning, drawing on observations concerning (among many other things) the US Constitution.
He closes with two chapters on complexity, which draw on many of the principles he discusses earlier.
As I mentioned at the outset, this is an important book. However, it might not be the best possible book addressing these issues. Simon was interested in a broad range of domains, and few of those who would benefit from his ideas will have the background to understand everything he says, or the interest to try. I can see a place for a less-challenging book, or set of books, to present the essence with application to specific domains of interest to readers at the undergraduate level.
Following are some quotes from the book (or the second edition).
Essence of design
Engineering, medicine, business, architecture and painting are concerned not with the necessary but with the contingent – not with how things are but with how they might be – in short, with design.
Essence of science
The central task of a natural science is to make the wonderful commonplace.
Normative or descriptive
The engineer, and more generally the designer, is concerned with how things ought to be – how they ought to be in order to attain goals, and to function … With goals and “oughts” we also introduce into the picture the dichotomy between normative and descriptive. Natural science has found a way to exclude the normative and to concern itself solely with how things are … Artificial things can be characterized in terms of functions, goals and adaptation.
Characteristics of artifacts
Fulfillment of purpose or adaptation to a goal involves a relation among three terms: the purpose or goal, the character of the artifact, and the environment in which the artifact performs.
Artifact as an “Interface”
An artifact can be thought of as a meeting point – an interface… – between an “inner” environment, the substance and organization of the artifact itself, and an “outer” environment, the surroundings in which it operates.
… advantage of dividing outer from inner environment in studying an adaptive or artificial system is that we can often predict behavior from knowledge of the system’s goals and its outer environment, with only minimal assumptions about the inner environment. … In one way or another the designer insulates the inner system from the environment, so that an invariant relation is maintained between inner system and goal, independent of variations over a wide range in most parameters that characterize the outer environment.
Quasi independence from the outer environment may be maintained by various forms of passive insulation, by reactive negative feedback, by predictive adaptation, or by various combinations of these.
The outer environment determines the conditions for goal attainment – if the system is properly designed, it will be adapted to the outer environment, so that its behavior will be determined in large part by the behavior of the latter…
… it is typical of many kinds of design problems that the inner system consists of components whose fundamental laws of behavior … are well known. The difficulty of the design problem often resides in predicting how an assemblage of such components will behave.
We do not have to know … all the internal structure of the system but only that part of it that is crucial to the abstraction. … if it were not, the top down strategy that built the natural sciences … would have been infeasible. We knew a great deal about the gross physical and chemical behavior of matter before we had an atomic theory, and a great deal about atoms before we had any theory of elementary particles…
… the possibility of building a mathematical theory of a system does not depend on having an adequate microtheory of the natural laws that govern the system components. Such a microtheory might indeed be simply irrelevant.
Complexity a façade
As we succeed in broadening and deepening our knowledge – theoretical and empirical – about computers, we shall discover that in large part their behavior is governed by simple general laws, that what appeared as complexity in the computer program was to a considerable extent complexity of the environment to which the program was seeking to adapt its behavior.
Optimize or satisfice
The decision maker has a choice between optimal decisions for an imaginary simplified world or decisions that are “good enough,” that satisfice, for a world approximating the complex real one more closely.
What a person cannot do he will not do, no matter how much he wants to do it. Normative economics has shown that exact solutions to the larger optimization problems of the real world are simply not within reach or sight. … the behavior of an artificial system may be strongly influenced by the limits of its adaptive capacities.
… to use feedback to correct for unexpected or incorrectly predicted events. Even if the anticipation of events is imperfect and the response to them less than accurate, adaptive systems may remain stable in the face of sizable jolts…
The simplest scheme of evolution is one that depends on two processes; a generator and a test. The task of the generator is to produce variety, new forms that have not existed previously, whereas the task of the test is to cull out the newly generated forms so that only those that are well fitted to the environment will survive.
Hierarchy and the functional paradigm
To design … a complex structure, one powerful technique is to discover viable ways of decomposing it into semi-independent components corresponding to its many functional parts. The design of each component can then be carried out with some degree of independence of the design of others, since each will affect the others largely through its function and independently of the details of the mechanisms that accomplish the function.
Process determines style
… both the shape of the design and the shape and organization of the design process are essential components of a theory of design.
Taxonomy of representation
An early step toward understanding any set of phenomena is to learn what kinds of things there are in the set – to develop a taxonomy.
Design of complex, durable systems
… the characteristics and complexities of designing artifacts on a societal scale The success of planning on such a scale may call for modesty and restraint in setting the design objectives and drastic simplification of the real-world situation in representing it for purposes of the design process.
Human cognition architecture
Human memory is best regarded as an extension … of the environment in which human thought processes take place and not as an increment in the complexity of these processes. What is remarkable about the whole architecture is … that memory enables the system to operate efficiently in a wide array of different task domains using the same basic equipment that it employs to understand and solve Tea Ceremony problems or simple statics problems in physics.
Complexity as hierarchy
… complexity frequently takes the form of hierarchy and that hierarchic systems have some common properties independent of their specific content. Hierarchy … is one of the central structural schemes that the architect of complexity uses.
Evolution of complex systems
The time required for the evolution of a complex form from simple elements depends critically on the numbers and distribution of potential intermediate stable forms.
… hierarchies have the property of near decomposability. Intracomponent linkages are generally stronger than intercomponent linkages. This fact has the effect of separating the high-frequency dynamics of a hierarchy – involving the internal structure of the components – from the low-frequency dynamics – involving interaction among components.
Description of complex systems
… two main types of description … (for) understanding of complex systems … (are) state description and process description. The former characterizes the world as senses; they provide the criteria for identifying objects, often by modeling the objects themselves. The latter characterize the world as acted upon; they provide the means for producing or generating objects having the desired characteristics.