Thursday, July 10, 2003

Thoughts on Spandrels

No one said "spandrels" up on stage, although Kevin Werbach told me he had considered it.

But they probably should have; it's an interesting and useful idea.

Architecturally, a spandrel is the triangular space created between (for example) two adjacent archways (that is, bordered by the right side of the leftward arch, the left side of the rightward arch, and the ceiling). Steven Jay Gould used this as a metaphor for something in evolution that's created as a side effect. No one (in Gould's formulation, although perhaps this isn't good architectural criticism) planned or intended the spandrel; it's an inevitable side effect of putting two adjacent arches in a design. Similarly, in evolution, a vast amount of phenotype isn't necessarily driven purely by natural selection; it might have evolved only because some other, possibly distantly-related, feature was selected for, and that feature had side effects. This is so for things like body plans; we often assume that a body (of, for example, a dog) is plastic to evolution. If evolution (or, more to the point for dogs, breeders) favored long snouts, then voila, the snouts would get longer over the generations. Now this is true as far as it goes, but lengthenieng the snout can't be done without making other changes in the shape of the skull and even the rest of the body. So a long snout might, for example, come packaged with sharper teeth. Those sharper teeth are the spandrels.

This whole argument is meant to be a counterbalance to fundamentalist adaptationists, who insist that every feature in an organism must have evolved that way because it offered some kind of selective advantage in evolutionary history, often requiring acrobatic feats of explanation. The idea of spandels significantly lifts the burden of adaptationist justification. (Of course, spandrels can then evolve into a useful feature in their own right, just as architectural spandrels became elaborately decorated).

The application of this idea to software or policy is useful. What features do we have not because they were useful on their own merits, but because of some other thing that we needed? The QWERTY keyboard, of course, which was created to slow down typists so they wouldn't jam the machine and "guard bands" in spectrum allocation because our technology wasn't sophisticated enough to resolve multiple signals. Certainly there must be others, but of course it's often hard to see them as such.