Beyond Correlation: Redefining Causation Through Robustness and Resilience to Perturbation

Correlation and causation are often treated as interchangeable yet describe different relationships. Correlation quantifies how variables co-vary, while causation denotes a directional influence by which one variable determines another’s state. Classical causal inference assumes that where causation occurs, correlation must follow, an assumption formalized as Faithfulness. However, Faithfulness fails in many biological and physical control systems like hormonal regulation, neural homeostasis and ecological feedback loops, which function by counteracting disturbances rather than amplifying them. Causation may therefore operate without producing observable co-variation, causing correlation to vanish and revealing the limits of conventional statistical approaches that rely exclusively on correlated change. We introduce an information-based definition of causation, conceived as preservation of informational structure against disturbance. A variable is considered causal when its influence decreases uncertainty in another variable exposed to unpredictable inputs, thereby maintaining order under noise. Using numerical simulations of feedback and feedforward systems, we showed that strong causal interactions can be reliably detected even when correlations between variables are negligible or negative. Our simulations revealed also reductions in conditional entropy and delayed oppositions between control and outcome, providing quantitative evidence of stabilizing causation hidden to traditional correlation-based measures. Unlike regression, structural equation modeling or transfer entropy, our approach revealed compensatory and self-maintaining dynamics operating through feedback, nonlinearity and temporal delay. By unifying causal inference and control theory, our agenda reframes stability as an active expression of causal power and enables the detection of hidden causal architectures in physiological homeostasis, neural stability, ecosystem resilience and engineered feedback systems.