Summary
An anticipatory system, as defined by Robert Rosen (1985), is a system that contains an internal predictive model of itself and its environment, and uses the predictions of that model to take antecedent actions in the present. This contrasts with purely feedback-driven (reactive) systems, which respond only to past events. The defining property is that the model must run faster than real time, enabling the system to act before the predicted state occurs.
Key Points
- Feedforth (feedforward) vs. Feedback: Feedback is past-driven (event → detection → response); feedforth is future-driven (model predicts → present behavior adjusts).
- Internal model runs faster than real time: The system must be able to simulate possible futures at a speed that outpaces actual system evolution.
- Self-modeling is constitutive: The predictive model must encompass the system's own dynamics, not merely the environment.
- Formalized via (M,R)-systems: Rosen’s (Metabolism, Repair) systems exhibit closure to efficient causation, which naturally embeds predictive self-models.
- Contrast with allostasis: Anticipatory systems align with Sterling’s allostasis model (predictive regulation) as opposed to homeostasis (reactive error correction).
Concepts
- Feedback (past-driven): Observes an event, detects a deviation, then triggers a corrective response (e.g., thermostat). Relies on history.
- Feedforth (future-driven): Uses a forward-running model to forecast outcomes and adjusts behavior before the predicted event occurs (e.g., predictive cruise control).
- Faster-than-real-time simulation: The internal model must generate predictions faster than the actual system evolves, otherwise the system can only react.
- Self-model: The model must include the system’s own state transitions and interactions, not just external factors.
- (M,R)-systems: Rosen’s formalism for biological systems that achieve closure to efficient causation — a circular causal structure that inherently produces predictive models of the system itself.