An Empirically testable Theory of Causality

Gerhard Schurz (Düsseldorf) gives a talk at the MCMP Colloquium (12 Jan, 2012) titled "An Empirically testable Theory of Causality". Abstract: Is the concept of causality a cognitive illusion without empirical content (as Hume taught us), or does it have a cognitively and empirically valuable function? This is the central question of this talk. I argue that should be understood as a theoretical concept, in analogy with "force" in Newtonian physics. The difference is only that 'causality' does not beong to a particular scientific discipline, but to a transdisciplinary theory. Causal-effect relations explain and/or predict certain (in)stability properties of probabilistic dependencies, namely screening-off and linking-up. I develop a theory of causality, based on causal graph theory, and prove theorems that demonstrate that this theory has empirical content, i.e., excludes logically possible probability distributions.