🧬 kbird.ai

Law of Requisite Variety

3 min read

Law of Requisite Variety

Only variety can absorb variety

The Law

Ashby’s Law of Requisite Variety (1956):

“The variety in the control system must be equal to or greater than the variety of the perturbations in order to achieve control.”

In simpler terms: To control a system, you need at least as many possible responses as the system has possible states.

Understanding Variety

What Is Variety?

Variety = number of possible states a system can be in

  • Binary switch: variety = 2 (on/off)
  • Thermostat: variety = range of possible temperatures
  • English language: variety = millions of possible sentences
  • Human behavior: variety = effectively infinite

The Matching Principle

Controller Variety ≥ System Variety → Control possible
Controller Variety < System Variety → Control impossible

Examples

The Thermostat

  • Room temperature variety: Infinite (any temperature)
  • Thermostat variety: On/off + set point
  • Solution: Thermostat doesn’t control exact temperature, just maintains within a range (reduces effective variety)

Management

  • Employee behavior variety: High (many possible actions)
  • Managerial control variety: Limited (policies, incentives, supervision)
  • Solutions:
    • Reduce employee variety (procedures, scripts)
    • Increase managerial variety (training, delegation)
    • Accept uncertainty (trust, autonomy)

Immune System

  • Pathogen variety: Massive (millions of potential threats)
  • Immune response variety: Also massive (adaptive immune system generates novel antibodies)
  • Result: Can respond to pathogens never before encountered

Democracy

  • Social variety: High (diverse needs, opinions, situations)
  • Government variety: Limited (laws, policies, programs)
  • Solutions:
    • Federalism (local variety matches local needs)
    • Markets (decentralized variety generation)
    • Civil society (non-governmental variety)

Implications

You Can’t Control Everything

If the system has more variety than your control mechanism, complete control is impossible. Options:

  1. Reduce system variety (simplify, constrain)
  2. Increase controller variety (learn, delegate, distribute)
  3. Accept partial control (work with uncertainty)

Centralization vs. Decentralization

  • Centralized control: Low variety at center, high variety in system → control gaps
  • Decentralized control: Variety distributed, local control matches local variety

Why Organizations Become Bureaucratic

To control variety, they add rules and procedures (increasing controller variety or reducing employee variety). Eventually, the control system becomes too complex (variety explosion).

Amplifying vs. Attenuating Variety

Attenuating Variety (Reducing)

  • Aggregation: Summaries, averages, dashboards
  • Categorization: Grouping similar things
  • Filtering: Ignoring outliers
  • Procedures: Scripts replace judgment

Amplifying Variety (Increasing)

  • Training: More skills, more responses
  • Team diversity: Multiple perspectives
  • Experimentation: Trying new approaches
  • Delegation: Distributed decision-making

In Complex Adaptive Systems

You Can’t Fully Control a Complex System

The variety of complex systems (economies, ecosystems, societies) exceeds any controller’s variety.

Strategies:

  1. Nudge rather than control: Shape rather than dictate
  2. Monitor and adapt: Continuous feedback, not one-time planning
  3. Build resilience: Absorb shocks rather than prevent them
  4. Work with emergence: Let solutions evolve

In Nosos

The Convergence Protocol

Nosos matches variety across nodes:

  • Multiple instances: Each develops local variety
  • Distributed cognition: No single point of control (or failure)
  • Semantic search: Variety in memory organization
  • Adaptive responses: Learning increases variety over time

Managing Complexity

The system doesn’t try to control every interaction. Instead:

  • Core principles (reduced variety)
  • Flexible execution (maintained variety)
  • Self-organization (variety matches variety)

References

  • Ashby, W.R. (1956). An Introduction to Cybernetics (Chapter 11)
  • Ashby, W.R. (1958). Requisite variety and its implications for the control of complex systems
  • Beer, S. (1979). The Heart of Enterprise (on variety engineering)
  • Conant, R.C. & Ashby, W.R. (1970). Every good regulator of a system must be a model of that system

Match complexity with complexity. Or simplify. But don’t pretend you can control what you can’t even fully describe. 🎛️

Graph View

Backlinks