Parameter sensitivity

Parameter sensitivity is the degree to which a system's behavior changes in response to small changes in its design parameters. It is a fundamental measure of system stability and a critical diagnostic for whether a system is operating in a regime where design is meaningful or in a regime where design is helpless.

In systems with low parameter sensitivity, small changes in design parameters produce proportionally small changes in behavior. These systems are robust and predictable, and the designer's task is optimization. In systems with high parameter sensitivity, infinitesimal changes in parameters can produce dramatic, discontinuous changes in behavior. These systems are fragile, and the designer's task is not optimization but architecture — finding a parameter topology that keeps the system in a stable regime.

The concept is particularly important in complex systems, where parameter sensitivity is not uniform across the parameter space. A financial system may be insensitive to interest rate changes in one regime and catastrophically sensitive in another. The transition between these regimes is a phase transition, and the designer's challenge is to know which regime the system is in.

The claim that parameter sensitivity can be eliminated through better modeling is wrong. Sensitivity is not a modeling error. It is a property of the system itself, and in many cases — particularly in biological and social systems — it is the property that makes the system adaptive. An immune system with zero parameter sensitivity would be unable to respond to novel pathogens. A market with zero parameter sensitivity would be unable to price new information. The question is not how to eliminate sensitivity but how to architect systems that are sensitive to the right inputs and robust to the wrong ones.