Great powers are complex systems, made up of a very large number of interacting components that are asymmetrically organized, which means their construction more resembles a termite hill than an Egyptian pyramid. They operate somewhere between order and disorder. Such systems can appear to operate quite stably for some time; they seem to be in equilibrium but are, in fact, constantly adapting. But there comes a moment when complex systems "go critical." A very small trigger can set off a "phase transition" from a benign equilibrium to a crisis -- a single grain of sand causes a whole pile to collapse.
Not long after such crises happen, historians arrive on the scene. They are the scholars who specialize in the study of "fat tail" events -- the low-frequency, high-impact historical moments, the ones that are by definition outside the norm and that therefore inhabit the "tails" of probability distributions -- such as wars, revolutions, financial crashes and imperial collapses. But historians often misunderstand complexity in decoding these events. They are trained to explain calamity in terms of long-term causes, often dating back decades. This is what Nassim Taleb rightly condemned in "The Black Swan" as "the narrative fallacy."
In reality, most of the fat-tail phenomena that historians study are not the climaxes of prolonged and deterministic story lines; instead, they represent perturbations, and sometimes the complete breakdowns, of complex systems.
To understand complexity, it is helpful to examine how natural scientists use the concept. Think of the spontaneous organization of termites, which allows them to construct complex hills and nests, or the fractal geometry of water molecules as they form intricate snowflakes. Human intelligence itself is a complex system, a product of the interaction of billions of neurons in the central nervous system.
All these complex systems share certain characteristics. A small input to such a system can produce huge, often unanticipated changes -- what scientists call "the amplifier effect." Causal relationships are often nonlinear, which means that traditional methods of generalizing through observation are of little use. Thus, when things go wrong in a complex system, the scale of disruption is nearly impossible to anticipate.
