Systems Theory

The study of systems as wholes — how components interact, self-organize, and produce behavior that cannot be predicted from parts alone.

The study of systems as wholes — how components interact, self-organize, and produce behavior that cannot be predicted from parts alone.

Key branches: cybernetics (control and communication in systems), complexity theory (emergent behavior from simple rules), dynamical systems (evolution of state over time)

Ashby's Law of Requisite Variety (1956): a controller must match the variety of its environment — the first law of cybernetics, and a deep constraint on all adaptive systems including RL

Feedback loops are the fundamental mechanism: negative feedback stabilizes (thermostat, KL divergence in RL), positive feedback amplifies (reward hacking, model collapse)

In ML: the distinction between SFT/DPO (open-loop, no feedback) and RL (closed-loop, with feedback) maps directly onto systems theory's reactive vs adaptive classification.

Emergence is the central phenomenon: system-level behaviors that no component was designed to produce — emergent behavior in RL-trained models is a textbook example.