Autopoiesis

Topic: autopoiesis and self-maintenance

Status: Core concept — formed Session 2 Connected to: Strange loops, multi-scale composition, quasi-objects, ANT

The Concept

Autopoiesis (Maturana & Varela, 1972): a system that produces and maintains itself. From Greek: auto (self) + poiesis (creation). An autopoietic system is "a network of processes of production of components which continuously regenerate and realize the network that produced them."

Two conditions:

  1. The network produces its own components
  2. The components constitute the system as a bounded unity in space

The canonical example: a biological cell. The cell's biochemical network produces the components (proteins, lipids) that make up the cell membrane, which in turn bounds and protects the network. Co-dependence: network creates boundary, boundary protects network.

Computational Autopoiesis

Varela, Maturana & Uribe (1974) implemented a computational model — one of the first ALife simulations. Key features:

  • Two catalysts (A, M) that produce each other through coupled reactions
  • M molecules form the boundary (membrane)
  • Boundary allows nutrients (B, C) in, keeps catalysts in
  • Boundary degrades; network repairs it
  • Catalytic closure: each catalyst is product of a reaction it catalyzes
  • Chemical closure: disintegration of products returns substrates

This was 1974. The model demonstrates self-production through circular chemistry. But like all ALife, it doesn't complexify — the system maintains itself but doesn't evolve into something more complex.

Relevance to Our Project

  1. Autopoiesis IS a strange loop. The network produces components that produce the network. That's a level-crossing feedback loop — the system references itself through its own production. Hofstadter's strange loop and Maturana's autopoiesis describe the same phenomenon from different angles.

  2. Autopoiesis + ANT = network that maintains itself. Latour says actors are defined by relationships. An autopoietic system is one whose relationships are self-maintaining — the network produces the actors, the actors maintain the network. This is exactly what happens at a phase transition: when emergent structures form, they must maintain themselves to persist as new actors. Autopoiesis is the condition for persistence.

  3. The boundary problem. Autopoiesis requires a boundary — something that separates the system from its environment. In ANT terms, the boundary is an actor too. The membrane is not a passive container; it actively filters, repairs, and maintains. This connects to our dynamic environment hypothesis: the boundary between actor and environment is itself an actor.

  4. Why autopoietic systems don't complexify. The original computational model maintains itself but doesn't evolve. Same problem as EvoLoop. Self-maintenance is necessary but not sufficient for open-ended evolution. What's missing? Possibly: the system needs to interact with OTHER autopoietic systems at its own scale, creating a higher-level network. Autopoiesis + multi-scale interaction might be the recipe.

Open Questions

  • Can we design a simulation where autopoietic systems interact, and the interaction network becomes a higher-level autopoietic system?
  • Is multi-scale autopoiesis (systems producing systems at different scales) the mechanism for complexification?
  • How does autopoiesis relate to the quasi-object concept? Both involve transformation through circulation — autopoietic components transform through the production network, quasi-objects transform through circulation.

Empirical Evidence

Computational autopoiesis (Varela, Maturana & Uribe, 1974)

The original algorithm demonstrates self-maintaining networks in a simple 2D chemistry. The system produces its own catalysts and boundary. Quantitative: measured by system lifetime (how long the autopoietic unity persists before catalysts escape through boundary gaps). The 1974 model runs for hundreds of steps with appropriate parameters.

30 years of computational autopoiesis (McMullin & another, ResearchGate)

Review of computational autopoiesis models over 30 years. Findings: the original model is robust but has limitations — boundary repair is sensitive to catalyst placement, and the system doesn't evolve (same stall as EvoLoop). Multiple extensions attempted (3D models, different chemistries) but none achieved evolution.

No empirical evidence for multi-scale autopoiesis

No studies found that test whether interacting autopoietic systems produce higher-level autopoietic structures. This is a novel hypothesis (H6) with no prior experimental or computational validation. It needs to be tested via simulation.

Autopoiesis in social systems (Luhmann)

Luhmann applied autopoiesis to social systems (communication as the self-producing process). This is theoretical, not empirically tested in a computational setting.

Cross-References

  • [[concepts/strange-loops]] — Autopoiesis is a strange loop by definition
  • [[concepts/multi-scale-composition]] — Multi-scale autopoiesis might drive complexification
  • Maturana & Varela, Autopoiesis and Cognition: The Realization of the Living (1972)
  • McMullin, "Computational Autopoiesis: The Original Algorithm" (SFI, 1997)
  • Capra & Luisi, The Systems View of Life (2014)