I think of causality as generally linked to our notion of invariants. We can arrive conceptually at causes in physics because we have "laws of physics", which

I think of causality as generally linked to our notion of invariants. We can arrive conceptually at causes in physics because we have "laws of physics", which are invariant over all conceivable conditions (except in the most extreme cases). Once you lay down your assumptions/invariants, the rest of the system can be "controlled" - intervene upon X and observe Y. If the system is truly invariant over repeated trials, like a computer algorithm or highly controlled experiment, then you can deduce some form of causality. As in a formal proof, causality necessitates a "starting point", an axiom, an invariant, from which all else follows. In biology and many other systems, you simply don't have many invariants at all. The entire topology is variable, constantly "shifting" as a single node changes. How can you ever control or reproduce such a system to deduce what causes what? It's possible, but as the author suggests, only with great difficulty, frequent missteps and irreducible imprecision.