Image Removed

In modern economic thought, efficiency is paramount. The goal of economic systems, and entities within those systems, is to maximize efficiency. Policymakers are supposed to remove impediments that are making markets inefficient, and cut government inefficiency.

Even those that claim not to buy into such thinking often still do without realizing it. There have been many hundreds if not thousands of papers and articles by critics of mainstream economic thought on how the “efficient market hypothesis” is wrong, and that markets today aren’t efficient or are broken in other significant ways despite their efficiency (e.g. “markets may be efficient but don’t consider human happiness”). However these critics are still buying into the framing of efficiency as king (and thus as the thing to inveigh against). Ultimately these arguments are saying that “if only we were to make change X to the system, then it will be properly efficient.”

A key aspect of photosynthesis is thought to have evolved only once. But that one time was enough, and was such an evolutionary advantage that we’ve had it and relied upon it ever since. However, it’s far from efficient — most plants only use a couple percent of the energy they receive from the sun.

Consider a highly-efficient solar panel farm — panels at just the right angles, tracking the sun, covering every inch of ground. From one perspective this is the pinnacle of technological development: shiny, “green” technology, silent and efficient (much more so than a plant’s leaf). But in ecological terms the landscape is a wasteland: no light is allowed to reach the ground for any other use, and as a result it’s likely next to nothing else will live there. (And we’re already seeing the impact of this — desert ecosystems are hurt where solar arrays are being built, and sometimes the carbon lost from the desert soils as a result is greater than the fossil-fuel carbon offset by the solar electricity production.) The landscape is efficient in a narrow sense, but it’s not clear that that’s what we should want.

Compare that to a forest’s use of sunlight, in which layer upon layer of foliage, from the canopy to the forest floor, take part in a dance of extracting what little they can from the sun’s rays, and sharing that with a huge diversity of species. Those species are arranged in a complex web of interactions such that no species is truly on top of a hierarchy — every species is in the middle of some part of the food web, and the density of connections between them ensures that the forest ecosystem can continue to function as these links disappear and new ones are created.

Consider the structure of network routing on the Internet. (Background: the Internet is decomposed into what are called ASes — Autonomous Systems — each of which is a large organization such as a company or university. Using the routing protocol BGP, ASes announce routes to IP addresses on the Internet to one another, and store the announcements they hear from others. When data packets arrive at an AS, it uses its knowledge of currently available routes to the packets’ destinations to forward them to the next AS along the path to the destination.) While the Internet is thought to be resilient — the apocryphal story is that it was designed to survive nuclear attack — the connectivity of the Internet is not nearly as resilient as is assumed. Specifically, there are roughly a dozen large ASes (the large telecom companies that provide most long-distance connectivity) that the Internet depends upon for its operation. Most paths between two destinations go up to a large AS and then back down to the final destination.

The networking specialist reading this might say, well wait a second — those individual autonomous systems are internally resilient, as they are composed of a large data plane of many thousands of routers. That’s true, and that does provide a degree of resilience — it’s why we don’t see huge outages on a regular basis. But they rely upon a single protocol for the control plane, and a handful of system administrators and programmers to keep that control plane in check. (The network’s data plane consists of the paths through wires and networking devices like routers and switches over which data — packets — flow for delivering useful service. The network’s control plane consists of the systems used to manage the network’s data plane; the control plane may be distinct from the data plane, or it may rely upon the data plane for its own communication.) Regional-scale disasters have had huge ripple effects in Internet connectivity — the Baltimore tunnel fire of 2001, the Taiwan earthquake of 2006 — while simple misconfiguration in a single spot has caused the entire Internet to go down. In each instance, network engineers put in place fixes to deal with the proximate cause of the outage, but the broader issues of a hierarchical, efficient, and un-resilient Internet remain. For the Internet to be truly resilient, in the way nature is, it would likely have to have less hierarchical routing, more backup network links, more diversity in ASes, and as a result would not be as profitable for the large corporations that carry much of the Internet’s traffic.

As a final example, a couple of years ago I looked at the challenge of using nuclear energy in hard times. Operators of such plants frequently ask and receive the go ahead for power uprates — they rely on the fact that each plant was built with some amount of safety margin, and cut into that safety margin so as to operate at a higher power yielding greater profit. (In re-reading my post on nuclear, I’m sad to say that large-scale solar thermal — which I was hopeful about — is looking more problematic than I thought then. Still, small-scale solar thermal works great and is even simpler.) Once again, the temptation to reach for efficiency, which is currently the primary virtue for any organization, trumps other concerns.

It seems to me that we need to openly begin questioning the dogma of efficiency as the main aim for all systems — economic, technological, social — a dogma that permeates modern life in many spheres — political, corporate, academic — even when it isn’t acted upon in every case. Instead, perhaps we can aim for inherent complexity in our systems, just like nature, ensuring resilience over efficiency.