Sometimes, it makes sense to count a single feature more than once. Minsky provides a nice example in which "we seem to see two complete arches, despite the fact that there aren't enough legs to make two separate arches."
"The double-arch problem also offers a choice of description styles. If you plan to build several separate things, you'd better keep count carefully or take the risk of running out of parts. But if you do that all the time, you'll miss your chance to make one object serve two purposes at once." This is easier to do by focussing upon functional descriptions instead of structural descriptions. "This doesn't mean that functioanl descriptions are necessarily better. They can make it hard to keep track of real constraints; hence they have a certain tendency to lead toward overoptimistic, wishful thought."
(NB: It is amusing to me to relate this last point to the "biological vacuum" of classical cognitive science, or, at the opposite end of the spectrum, to Paivio's "house of cards" attack on the nonempirical nature of computational theories).
Pearl Street | Society Of Mind Home Page | Dawson Home Page |