How Are the Sciences of Complex Systems Possible?

To understand the behavior of a complex system, you must understand the tangled web of interactions between the system’s many parts. The web is tractable in systems that are somewhat decomposable, meaning that the interactions influence only weakly the short-term behavior of the parts. But how to handle complexity in systems that are nowhere near decomposable? Science’s principal tool for dealing with non-decomposable systems is a variety of probabilistic analysis that I call epa, exemplified by statistical physics and population genetics. I show that epa’s power to deal with complex bundles of interactions derives from an assumption that appears to be false of non-decomposable complex systems, in virtue of their very non-decomposability. Yet epa is extremely successful. I aim to find an interpretation of epa’s assumption that is consistent with, indeed that explains, its success.