An attractor defines a stable system: a system is a pattern in its State Space. But patterns persist for different reasons. A pendulum’s pattern persists because nothing disturbs it; an agent’s pattern persists because the system itself works to restore it – it registers deviations and acts to return, so that its Lyapunov function is an outcome of goal-directed behaviour rather than a given of the dynamics.
An agent, then, does not merely have an attractor: an agent is an attractor together with the activity that maintains it from within.
This makes “attractor” and “agent” two descriptions of one phenomenon
- from the dynamics-eye view, a region of state space with a basin and pull;
- from the agent-eye view, a system with Information Closure – a self-model, goals, a boundary it holds.
Neither description reduces to the other; which is apt depends on whether you are asking how the surrounding landscape behaves or how the pattern sustains itself.
Three corollaries:
- An attractor maintained from within cannot be installed from without. Imposing the configuration reproduces the pattern without the closure that sustains it – the shape of the agent without the agent. This is the deep reason Revolutions try to force systems into imaginary attractors and fail: what they try to install is precisely the kind of attractor that can only grow.
- Collective agents are attractors of their social systems. A movement that acts coherently is a pattern the wider Social System has fallen into, maintained from within by practices of mutual commitment. Cultivating such an agent therefore means shaping the landscape – its constraints and affordances (Affordances contribute to attractors) – until the pattern can maintain itself; it never means assembling the pattern directly.
- An agent’s attractor is a strange attractor. Since Living Systems stay near-critical and avoid settling into rigid attractors, the pattern an agent maintains is a way of moving, not a fixed configuration – states always similar, never the same. An agent that tries to hold itself or its world at a point has begun to die.
References
- Juarrero (2000), “Dynamics in Action: Intentional Behavior as a Complex System” (a system’s identity as the signature distribution of its dynamics)
- Friston (2017), “The mathematics of mind-time” (goal-directed systems as actively seeking their attractor states)