Elements of a cybernetic epistemology: Adaptive systems that can develop system‐specific behavior

Abstract

Purpose

The purpose of this paper is to investigate necessary cybernetic structures that allow complex adaptive systems to develop system‐specific behavior.

Design/methodology/approach

Following Holland's concept of “adaptive agents”, it is argued that the development of system‐specific forms of goal‐oriented behavior requires a decision to deviate from some default behavior and to trigger any new one, and a mechanism to evaluate the goal‐orientation of this new behavior. Using a functional approach cybernetic structures are developed that are able to carry out these two tasks. Then these structures are added as subsystems to the structure of a simple one‐level adaptive system.

Findings

The paper finds that a hierarchical adaptive system can recognize with a higher level controller, if lower level decisions lead to an insufficient degree of goal‐approximation and can use preprogrammed higher level decisions to intervene on the lower level to trigger new system‐specific actions. An additional controller can evaluate the “success” achieved with these new actions and can select the “best” actions found, i.e. the behavior leading to the highest degree of goal‐approximation.

Practical implications

The paper shows necessary cybernetic structures that are seen as core of all complex adaptive systems able to develop system‐specific behavior. It is suggested that the underlying basic concept of “success” understood as a degree of goal‐approximation holds for any adaptive, learning or otherwise improving endeavor.

Originality/value

The paper is the second in a series of three on a cybernetic theory distinguishing system capable of preprogrammed adaptation, system‐specific adaptation, and learning. It shows necessary cybernetic structures that a system can develop individual actions.

Citation

Nechansky, H. (2010), "Elements of a cybernetic epistemology: Adaptive systems that can develop system‐specific behavior", Kybernetes, Vol. 39 No. 4, pp. 553-569. https://doi.org/10.1108/03684921011036781