Autocatalytic degradation

Autocatalytic degradation is the self-accelerating collapse of a system whose own outputs catalyze further degradation. Unlike externally driven decay, autocatalytic degradation is endogenous: the system produces the conditions of its own destruction, and those conditions in turn accelerate production. The result is not linear decline but a phase transition from functional to degenerate regimes.

The concept extends the chemical notion of autocatalysis — a reaction whose product catalyzes its own production — to systems theory. In chemical autocatalysis, the product accelerates a desirable reaction. In autocatalytic degradation, the product accelerates an undesirable one: the production of errors, the loss of diversity, or the collapse of complexity.

Examples

Model collapse is the canonical example in artificial systems. A generative model produces synthetic data that is used to train the next generation. The synthetic data is not random noise; it is a systematically degraded version of the original distribution. This degradation is autocatalytic because each generation's outputs are more degraded than its inputs, and the degradation itself becomes the training signal for further degradation.

Civilizational collapse exhibits the same pattern. Complex societies produce infrastructure, institutions, and knowledge networks that enable further complexity. But when the maintenance costs of this infrastructure exceed the society's productive capacity, the infrastructure itself becomes a drain. The degradation of one subsystem cascades to others, and the cascade is autocatalytic because each failure increases the load on the remaining systems, accelerating their failure in turn.

The Formal Structure

The formal structure is a positive feedback loop with a negative intercept. In standard positive feedback, the output reinforces the input: X produces more X, which produces more X. In autocatalytic degradation, the output reinforces the input but the baseline is declining: X produces degraded X, which produces more degraded X. The system does not explode or converge to an equilibrium; it converges to a degenerate attractor — a state that is stable but functionally useless.

The mathematical signature is a superexponential approach to a lower bound. The rate of degradation is proportional to the current state of degradation, producing a logistic-like curve in reverse. The inflection point — where the degradation transitions from gradual to rapid — is the critical threshold. Before the threshold, the system appears stable. After the threshold, collapse is inevitable.

Implications

The implication for design is that systems vulnerable to autocatalytic degradation require structural brakes — mechanisms that interrupt the feedback loop before the critical threshold is crossed. These brakes cannot be added as afterthoughts; they must be intrinsic to the system's architecture. Once autocatalytic degradation begins, external intervention becomes increasingly difficult because the system's own dynamics resist it.