Complexity: It’s Not as Simple as Bill Easterly Thinks It Is (It’s Simpler)

Folks over at @aidwatch have been getting into writing about complex systems these days. I’m not sure I know what they’re talking about. And I don’t think they do either.

Here’s Bill Easterly writing today in the Guardian’s Poverty Matters blog:

A popular topic in the aid blogosphere this week was not about Haiti or Ivory Coast or south Sudan but about complex systems, i.e. systems that cannot be reduced to a simple mathematical or statistical model, where actions often have unintended effects. [Link included as in the post]

Now, admittedly, this is just one sentence. But still, it would be difficult to come up with a more poorly informed summary of the nature of “complex systems” than the one Easterly offers here. 
Why? The reason is that the core insight of the study of complexity–be it deterministic chaos, cellular automata (e.g. John Conway’s marvelous “game of life“) or agent-based modeling in general, to cite just a few of the many variants that loosely define this domain of study–is this: systems that are not just reduced to, but actually defined by, simple mathematical models, have the potential to generate extremely…well, complex behaviors. The classic example is the logistic map, an extremely simple function whose dynamics are highly complex:
Anyhow, the key point is that such complex systems, while entirely deterministic (that is, lacking any random element) generate behaviors that are indistinguishable from those of stochastic systems (systems driven by a random component). (When mapped in “state space” they also yield beautiful fractals…but that’s another story.) Before the study of deterministic chaos (and with it “extreme sensitivity to initial conditions” a.k.a. the “butterfly effect”), determinism and randomness were understood to be opposites. So understanding that there were significants domains in which they were indistinguishable from one another was a pretty big deal.
The caricature of “complexity theory” that Easterly offers is in line with a body of work by Austrian economists that has sought, with greater and lesser desperation, to connect Friedrich Hayek’s famous notion of “spontaneous order” in economics systems to that of “emergence” and “self-organization” in complex systems. This was a failed undertaking from the outset, nearly twenty years ago now. It is also one that I have some confidence Hayek himself would have judiciously avoided. The reason is that the problem that concerned Hayek (and von Mises before him) was not the emergence of  complexity in the absence of randomness, but rather the impossibility of calculation in the presence of randomness (and its siblings–noise, and informational decay). 
An illustration: In the “The Impossibility of Socialist Calculation,” Hayek assails the now-forgotten Oskar Lange for making following assertion in an attempt at critiquing von Mises: “The administrators of the socialist economy will have exactly the same knowledge, or lack of knowledge, of the production functions as the capitalist entrepreneurs have.” This is what Lange said; Hayek characterizes the claim, with his characteristic light touch, as “a blatant untruth, an assertion so absurd that it is difficult to understand how an intelligent person could ever honestly make it.”
Now, let’s pause to consider. If the essence of the problem faced by administrators of socialist economies was deterministic chaos as described above (the essence of unpredictability in complex systems) then Lange’s statement would not be absurd. It would actually be correct: under circumstances of deterministic chaos there is no way the entrepreneur could have better information than the planner. But that’s not Hayek’s objection. He does not believe that Lange has failed to grasp the essence of extreme sensitivity to initial conditions. Indeed, his point is not fundamentally about dynamics at all. What Hayek is really talking about is the heterogeneity and localization of information. In other words, Lange is wrong because the entrepreneur has unique and specific information to which the planner does not have access:

The individual entrepreneur will not possess or require knowledge of general production functions, but he will currently learn from experience how at any given time variations in the qualities or the relative quantities of the different factors of production he uses will affect his output. This information relevant for and possessed by each entrepreneur will be very different from that possessed by others. To speak of the aggregate of such information dispersed among hundreds of different individuals as being available to the planning authority is pure fiction.

Given this, what do you think Hayek would say about a model of economic dynamics in which every economic actor not only possessed the same ability to observe local conditions, but followed the exact same deterministic rules? That is the essence of a complex system! But it is the opposite of a model built upon the heterogeneity of economic actors, and the impossibility of reducing their choices to simple models in the aggregate. In other words, a complex system is the opposite of “a system that cannot be reduced to a simple mathematical or statistical model, where actions often have unintended effects” (ref. above).
Now, since ours is a country in which liberty prevails, I suppose it’s OK to employ the term “complex systems” to refer to things that are its opposite. Why not? It’s all just silly science stuff anyway. So, henceforth, I am sure that Easterly will allow the same license to others who chose to reinvent technical terms–say “effectiveness,” “confidence,” “evidence”–in ways that similarly suit their fancy. Sound good?