Central Argument

How can we develop a science of things that are designed rather than given by nature?

A clock, a business firm, a computer, and a human decision-maker are very different systems, but all can be studied as systems adapting their internal capabilities to external conditions in pursuit of goals.

Simon argues that we can understand these systems through the relationship between goals, inner environment (capabilities), and outer environment (task conditions).

When adaptation is successful, the environment largely determines behavior. When it is imperfect, the limitations of the internal system become important.

This provides a common framework for understanding economics, cognition, design, and complexity.

1. Behavior Often Reflects the Environment

An ant crossing a beach takes an irregular path around obstacles. The path looks complicated, but the ant may be following relatively simple rules.

Much of the apparent complexity of its behavior comes from the beach, not the ant.

Similarly, different artifacts can exhibit similar behavior despite having completely different internal mechanisms. Birds and airplanes fly; mechanical and electronic clocks tell time.

The crucial object of study is the adaptation of a system to its environment.

2. Adaptation Is Constrained by Limited Rationality

Traditional economics often assumes people choose optimal actions. Simon argues that the real difficulty is discovering those actions with limited information and computation.

In the traveling salesman problem, finding the shortest possible route through many cities is well defined but computationally difficult. Rather than examine every route, we use heuristics to find a sufficiently good one.

Satisficing means finding an alternative that meets aspiration levels, instead of proving it is the best possible alternative.

This distinction shifts attention from ideal outcomes to the actual processes of thinking and decision-making.

3. Knowledge and Representation Make Search More Effective

Expert chess players recognize meaningful board configurations built from experience. Their advantage is not simply that they calculate more possibilities, but that stored knowledge guides their search.

The way a problem is represented also influences how difficult it is to solve. Simon discusses alternative representations of games and computer-aided design systems that make constraints explicit.

Intelligence depends not only on computational capabilities, but also on the knowledge and representations available to those capabilities.

4. Coordination Can Reduce the Need for Individual Intelligence

Markets coordinate large numbers of individual decisions through prices. Simon cites experiments showing that even traders following minimal rules can produce market-clearing behavior.

Organizations achieve coordination differently, through shared assumptions, communication, authority, and decentralized decisions.

Markets and organizations are information-processing arrangements that help people coordinate despite bounded rationality.

This helps explain why elaborate collective behavior need not require a single actor capable of understanding or calculating everything.

5. Design Is a Systematic Search for Possible Worlds

Simon argues that engineering, medicine, architecture, and management all share a central activity: devising actions that change existing situations into preferred ones.

Design requires evaluating alternatives, searching for feasible solutions, managing limited resources, and choosing useful representations.

But social design introduces an additional complication: the goals themselves may change through the design process.

In Pittsburgh’s redevelopment, the initial work opened possibilities for further changes that were not fully anticipated at the beginning.

Design is not always about executing a fixed plan. It can be an adaptive process of discovering what is possible and desirable.

6. Hierarchy Explains How Complexity Can Be Built and Understood

Two watchmakers assemble equally complex watches. Tempus builds each watch as one continuous assembly and loses his progress when interrupted. Hora builds stable subassemblies, losing little progress from interruptions.

Hora is dramatically more efficient.

Simon uses this example to explain why hierarchical structures are so common: stable intermediate structures make the evolution of complex systems much more feasible.

When subsystems interact more strongly internally than externally, a complex system is also easier to analyze. Simon calls this near decomposability.

Hierarchy therefore helps explain both the creation and comprehensibility of complexity.

Synthesis

The book develops three connected arguments.

Understanding: Artificial systems can be studied scientifically through their goals, internal capabilities, and environments, even when their specific constructions are contingent.

Intelligence and design: Because adaptation is limited by knowledge and computation, intelligent systems rely on search, heuristics, memory, representation, and coordination.

Complexity: Hierarchical organization and stable subsystems make complex systems more feasible to construct, evolve, and understand.