The architecture of complexity

systems
science
Author

Christophe Beaucé

Published

April 10, 2016

Herbert A. Simon, The Architecture of Complexity, Proceedings of the American Philosophical Society, 1962

Herbert A. Simon, one of the founder of cybernetics, has written in 1962 a ground breaking paper about complex systems: The Architecture of Complexity. It was published in the Proceedings of the American Philosophical Society.

First, Herbert Simon analyses the characteristics of complex systems. He remarks that whatever the nature of the systems (physical - such as a galaxy, natural - such as an animal, social - such as a firm, etc.), complex systems tend to take the form of a hierarchy.

In the general term, complex systems are hierarchic when they are composed of interrelated subsystems, each of them being in turn composed of subsystems until we reach some lowest unitary level for the analysis.

Then, the question arises: why is hierarchy such a frequent property of complex systems ?

Simon answers with a parable: the parable of Tempus and Hora, two competing watch makers. They are both producing watchs from 1000 individual parts, and both of them are subject to interruptions of probability p. When they are interrupted, the watch making process is halted and has to resume to the last stable state in the making process. The watch makers use two different processes. Tempus makes a watch from all the 1000 parts. Hora makes a watch by assembling groups of 10 elements in hierarchy. Simon demonstrates that if the probability of an interruption is 0.01, then it will take in average 4000 time longer for Tempus to complete the same number of watches than for Hora !

Stability is a critical, fundamental characteristics of complex systems. The more they are stable, the more they are successful. Fragile, they disappear. In a Darwinian analogy, this is akin to the survival, not of the strongest, but of the most stable !

So what is the key characteristic of complex systems ? They are stable, they have homeostatis, which means that their variables are regulated so that internal conditions remain stable and relatively constant. To achieve this, their architecture frequently exhibits the property of Near-Decomposability. A Near-Decomposable System is such that the short run behaviour of each component is independant, while the long run behaviour of each component is dependant in aggregate of all the other components.

Interestingly, problem solving has similar characteristics than natural selection. Solutions are discovered by cutting down problems hierarchically until a way out of the maze is found. Problem solving involves the description of states (the world as it is sensed), and the description of processes (the world as it is acted). The translation back and forth from state description to process description is the method by which a solution is discovered, validated or rejected. This is the process of trial and error. This is the way by which science has progressed through centuries.

These are only some of my take-aways from this so rich paper, which is a corner stone of the science of systems.