Requisite Variety

Origin. W. Ross Ashby's Law of Requisite Variety (1956), adopted and extensively applied in Soviet organizational cybernetics by Glushkov, Berg, and the OGAS project team; "Only variety can destroy variety."

Mechanism. A regulator can control a system only if the regulator's range of responses matches or exceeds the system's range of disturbances. If the environment has N possible states and the controller has M possible responses with M < N, some disturbances must go uncompensated. The law is a statement about information: control requires a channel from disturbance to response with capacity at least log₂(N) bits. Soviet cyberneticians applied this to critique both over-centralized command economies (insufficient variety at the center) and to design distributed information systems.

Procedure. Enumerate the disturbances the system faces. Enumerate the control actions available. If the action set is smaller, the controller is underspecified: either expand the action set, reduce the disturbance set by buffering or filtering, or accept that some disturbances will not be regulated. For organizational systems: decision-making authority must be distributed to match the variety in the environment; centralization is control under-specification.

Applies to. Organization design, API design, regulatory design, and any system where the designer controls the regulator but not the environment.

Limitations. Variety is counted in distinguishable states, and distinguishability depends on the purpose. A regulator with high variety that does not align with the disturbance dimensions provides no control. Also: variety can be attenuated (reduced by aggregation) or amplified (increased by disaggregation); the law governs the necessary variety at the point of control, not the raw variety in the environment.

Note: This is Ashby's British formulation, but it became central to Soviet organizational cybernetics and the theoretical basis for distributed versus centralized control debates within the USSR.