A system is what its loops make it do next. The most counter-intuitive insight of systems thinking is that the structure of a system — the network of variables that influence each other — predicts its behaviour far more reliably than the intentions of the people inside it. The thermostat, the predator-prey ecology, the financial bubble, the office political dynamic, the climate, the human body — each is dominated by a small number of feedback loops that determine whether the system stabilizes, oscillates, accelerates, or collapses. Once the loops are seen, the behaviour stops being mysterious.
Negative feedback (also called balancing) loops resist change and produce stability around a setpoint. A thermostat: room cools below 20°, heater turns on, room warms, heater turns off. A predator-prey balance: prey numerous → predators eat well → predator population grows → prey population falls → predators starve → prey recovers. Most homeostatic biology runs on negative feedback: blood glucose, body temperature, electrolyte balance, breathing rate. Positive feedback (also called reinforcing) loops amplify change and produce exponential growth or collapse. Bank runs: depositors withdraw → bank weaker → more depositors withdraw. Viral epidemics: infected people infect more → more infected people. Climate albedo: ice melts → less reflectivity → more warming → more ice melts. Social-media outrage cascades: indignant post amplified → more indignant responses → topic hijacks the feed. The general analytic move is not to ask 'who caused this?' but 'what are the loops, and which is dominant right now?' Systems often flip between regimes when a different loop becomes dominant — economies can shift from stable equilibrium to runaway crisis when a positive-feedback loop (e.g., debt-deflation) is activated. Time delays in loops produce oscillation: predator-prey populations cycle, business cycles oscillate, biological hormone systems pulse.
Climate change is dominated by positive-feedback loops (ice-albedo, water-vapor amplification, methane release from permafrost, loss of carbon sinks) that may be approaching tipping points beyond which the system cannot be returned to the previous equilibrium by human action. Social media amplifies positive-feedback loops in opinion formation, polarization, and viral content; the platforms' algorithms are themselves loop-dominated systems whose emergent behavior surprises even their designers. Generative AI creates new feedback loops: model output enters training corpora, gets re-amplified, with effects on future model behavior that are not yet well understood. Donella Meadows's leverage points (1999) — the most-cited essay in systems thinking — argues that the highest-leverage interventions in any system are at its loop structure, not at the variable levels, because changing the loops changes the long-run behaviour.