Skip to content

Emergence & Downward Causation

Thoughts on the relationship between parts and wholes in complex systems.

1. What Is Emergence?

Emergence describes the appearance of properties of a whole that cannot be found in any of its parts taken individually.

A single neuron does not think.
A single bird does not "flock."
A single ant does not optimise a path.

And yet: from the interaction of many parts, phenomena arise — thinking, swarm formation, path optimisation — that do not exist at the level of the individual parts.

"More is different."
— Philip W. Anderson, 1972

2. Weak vs. Strong Emergence

The philosophical literature distinguishes two readings:

Weak Emergence Strong Emergence
Claim Macro-properties are in principle derivable from micro-rules, but surprising and hard to predict Macro-properties are in principle not reducible to micro-rules
Example Turbulence patterns from Navier-Stokes Consciousness (?)
Status Widely accepted Philosophically controversial
Relevance for simulations High — all models in this repo show weak emergence Unclear — no simulation model (yet) shows strong emergence

All simulations in this repository are weakly emergent: the macro-phenomena (synchronisation, swarm formations, Turing patterns) follow deterministically from the local rules — but they are surprising, non-trivial, and not readable from the individual rule alone.

3. Epistemic vs. Ontological

A related distinction:

  • Epistemic emergence: We could derive the macro-level, but it is too complex, so we need new levels of description (thermodynamics, ecology, psychology).

  • Ontological emergence: The macro-level exists in a way that is not fully determined by the micro-level — there is genuine "new causality."

Most natural scientists implicitly work with epistemic emergence: temperature is nothing other than average kinetic energy, but it makes sense to speak of "temperature" because it facilitates predictions.

Whether there is ontological emergence beyond this remains an open question — and one of the deepest in the philosophy of mind.

4. Downward Causation

If macro-properties are emergent — can they then act back upon the parts?

The Strong Argument

The swarm formation determines where the individual bird flies.

In the Boids model, the global formation acts back on each individual: the cohesion vector points toward the centre of mass of the neighbours, which is a property of the collective.

The Counter-Argument

There is no mysterious "downward force." What acts on the individual boid are the positions of its neighbours — everything remains at the micro-level. The "swarm formation" is merely a description of these positions.

Position in This Repository

We take a pragmatic stance:

Downward Causation is a useful explanatory pattern, even if it may be fully reducible to micro-level interactions.

  • In ecosystem-regulation/: the global density feeds back on the local birth probability.
  • In meta-learning-regime-shift/: the global prediction error modulates the individual agent's learning rate.
  • In boids-flocking/: the local neighbourhood structure creates global formations, which in turn determine the neighbourhood.
  • In reaction-diffusion/: macroscopic concentration patterns determine the local reaction rate.

In every case, we can observe: describing at the macro-level makes the system more comprehensible, even though causality is technically micro-local.

5. Connection to the System Intelligence Index

In the System Intelligence Index, emergence appears on three levels:

  1. Predictive Power (P): A system's internal model captures emergent regularities — not atoms, but patterns.

  2. Regulation (R): Homeostasis is a macro-property. The "target variable" (e.g. population density) does not exist at cell level — it emerges.

  3. Adaptive Capacity (A): Meta-learning alters how a system learns — an effect of the performance level on the learning level.

The SII framework itself is an attempt to capture the degree of emergent intelligence — without claiming that intelligence is anything other than a macro-phenomenon.

6. Open Questions

  • Can emergence be compared across simulation models? (Is a swarm "more emergent" than a cellular automaton?)

  • Is there a connection between degree of emergence and information measures (integrated information, mutual information)?

  • Can Downward Causation be operationally defined — e.g. as: "The behaviour of a micro-element is better predictable when one knows the macro-level"?

  • How does emergence relate to complexity? (Not every complex system is emergent, and not every emergence requires high complexity.)

7. References

  • Anderson, P. W. (1972). More is Different. Science, 177(4047).
  • Bedau, M. A. (1997). Weak Emergence. Philosophical Perspectives.
  • Kim, J. (1999). Making Sense of Emergence. Philosophical Studies.
  • Kauffman, S. (1993). The Origins of Order. Oxford University Press.
  • Tononi, G. (2004). An information integration theory of consciousness. BMC Neuroscience.

This essay is intentionally informal and exploratory. It accompanies the simulations in this repository and is meant to invite further thinking.