systems

A system is an interconnected set of elements that is coherently organized in a way that achieves something.

This definition comes from Systems theorist Donella Meadows’ definition in Thinking In Systems.

Elements

A system must have more than one part, otherwise it’s just an object.

Connections

The elements in a system must exert some influence upon each other, otherwise it’s just a collection of parts. An automobile is a system capable of transportation when all of its parts are connected. If taken apart, it ceases to be a system.

Purpose

All systems have a primary purpose. In the case of naturally occurring systems, that purpose usually aligns with natural selection. For example, the heart’s purpose is to pump blood in order to sustain the organism so that it can continue living, pass on its genes, and sustain the species. In human-made systems, purposes vary widely. A toaster is a system designed to toast bread. A soccer team is a system designed to win games. It’s important to remember that in human-made systems, the system’s purpose can change over time and is not necessarily the system’s stated purpose.

Self-Preservation: The Unstated Purpose

While different systems obviously have unique purposes, most systems have the secondary purpose of self-preservation, since the system cannot achieve its primary purpose if it ceases to exist. Put another way, systems have integrity (they hold together in some way), and mechanisms in place to preserve that integrity. Notably, this overlaps significantly with the definition of life. My body is a system that resists disintegration until I die, at which point it ceases to be a system and disintegrates.

It’s especially interesting when a system’s primary purpose conflicts with its self-preservation. Consider a charitable organization whose primary purpose is to eradicate a disease: if it were to succeed at that purpose, it would cease to be necessary. This conflict of interest could explain the charity’s behavior if it appeared to focus on treating the symptoms of the disease rather than addressing its root causes.

This conflict also gives rise to a number of nightmare scenarios with artificial intelligence, as outlined in Superintelligence: Paths, Dangers, Strategies. If we were to create a powerful AI and give it the mission of ensuring the continued existence of the human race, it might reasonably take steps to make sure we don’t interfere with that mission by unplugging it, perhaps by killing most humans and keeping the remaining few in a kind of zoo, thereby technically fulfilling its mission.

Viable Systems

Stafford Beer proposed the viable systems model for understanding how systems and their various subsystems interact. He proposed 5 kinds of subsystems each with a unique role in the overarching system:

1. System 1 (doing the work) does the actual work to fulfill the purpose of the overarching system.
2. System 2 (support) supports System 1 with the information and communication channels needed to do its job and acts as an interface to System 3.
3. System 3 (internal coherence) monitors System 1 through System 2 to ensure they’re working together to fulfill the goals of the overarching system.
4. System 4 (environment) evaluates the outside environment, providing feedback to other subsystems and driving the adaptation of the overarching system to remain viable in a changing environment.
5. System 5 (policy) steers the system as a whole, balancing the needs of subsystems and determining how the system will achieve its purpose.

Some important consequences of the viable systems model:

• Viable systems are recursive at the lowest level: every System 1 must itself be a viable system (with its own subsystems of types 1-5).
• Higher-level systems exist to support lower-level systems. Systems 2-5 must not be viable systems, or they will threaten the integrity of the overarching system. (I imagine this danger looms larger as systems and subsystems grow in size and complexity.) If a supporting system (e.g. System 2) starts trying to become viable and preserve itself, this will come at the expense of System 1 which actually fulfills the purpose and maintains the viability of the overarching system.

Jessica Kerr applies the viable systems model to software teams.

And according to LowImpact.org:

combining, mutual credit and an idea called the viable systems model is the most promising way to build a new, non-extractive economy.

Further Reading

• “How Complex Systems Fail”, a short and influential paper by Richard I. Cook, available at the delightful URL: https://how.complexsystems.fail/