“The
spider has constructed a web of statistical probability, honed to near
perfection by millions of years of evolution.” (Robert J. Hustwit)
Photo by Alfonso Diaz.
Systems involving many interacting variables are at the heart of the natural andsocial sciences. Causal language is pervasive in the analysis of such systems, especially wheninsight into their behavior is translated into policy decisions. This is exemplified by economics,but to an increasing extent also by biology, due to the advent of sophisticated tools to identifythe genetic basis of many diseases. It is argued here that a regularity notion of causality canonly be meaningfully defined for systems with linear interactions among their variables. Forthe vastly more important class of nonlinear systems, no such notion is likely to exist. Thisthesis is developed with examples of dynamical systems taken mostly from mathematicalbiology. It is discussed with particular reference to the problem of causal inference in complexgenetic systems, systems for which often only statistical characterizations exist.
