Runaway Processes

A runaway process is positive feedback without adequate negative feedback to check it. The system accelerates until it hits a hard limit — resource exhaustion, structural failure, external intervention.

The pattern: small perturbation → amplification → further amplification → explosion or collapse. Bank runs, avalanches, nuclear meltdowns, viral outbreaks. Each increment strengthens the force driving the next increment.

Financial panics are classic runaways. Asset prices fall; leveraged holders must sell to meet margin calls; selling pushes prices lower; more margin calls trigger; the cascade accelerates. The 1929 crash, 2008 crisis, and every bubble collapse follow this structure.

The trigger is often tiny relative to the outcome. A single bank failure can cascade through the system. A rumor can empty shelves. The initial perturbation matters less than the feedback structure that amplifies it.

Climate tipping points are potential runaways. Ice reflects sunlight; as ice melts, darker ocean absorbs more heat; warming accelerates; more ice melts. Permafrost stores methane; warming releases methane; methane traps heat; warming accelerates. These positive feedbacks could overwhelm negative feedbacks that have historically stabilized climate.

What makes runaways dangerous: they can cross thresholds beyond which the old equilibrium is unreachable. The system shifts to a new state. Some runaways are reversible with enough force. Others are one-way doors.

Runaways in human systems often involve social proof. Everyone is buying → buying signals value → more people buy → prices rise → rising prices confirm value → everyone is buying. The information cascade creates its own evidence.

The same structure drives cancellations, fads, and contagions of belief. Each person’s action influences the next person’s perception of what’s true, appropriate, or safe.

Stopping runaways requires acting early, before amplification compounds. Once the cascade is running, the force required to stop it grows exponentially. Fire is easy to stop at ignition, impossible to stop at conflagration.

This is why early warning systems matter more than crisis response. Detection when the system is one standard deviation off baseline is worth more than detection at five standard deviations. The small anomaly that precedes the large disaster is the intervention point.