Can designs be too clever?

The History of the Button blog has a really interesting post on a seemingly smart, but ultimately confusing elevator design. Is it possible for designs to be too clever for their own good?
Dealing with the user’s ingrained expectations is often a dilemma when trying to innovate: do you create something that is novel and potentially confusing, or do you conform to the user’s expectations? Often, an idea will be a clear improvement on paper, but the user’s habits will in practice prove an overwhelming barrier. This is particularly true when the interaction is almost subconscious, as in the elevator example linked to above.
When a novel design provides significant new value, users may be willing to invest some time and mental effort to learn new behaviors. Many people are willing to learn how to use a navigation system, as it makes life so much easier. Nevertheless, designers should work to minimize these switching costs by providing adequate explanations (preferably in context, not tucked away in an obscure manual), adhering to usability conventions as much as possible, and using methods such as progressive disclosure.
Of course, the most revolutionary designs are those that innovate in a way that is so intuitive that they behave exactly as the user expects them to — even though the interaction is completely novel. Automatic transmissions are a good example of this — they are clearly simpler to use than their alternative. Not to use a tired example, the iPod’s scroll-wheel also springs to mind. It remains to be seen if the iPhone lives up to expectations, though usability expert Bruce Tognazzini has high hopes.
I’ve touched on this subject before: elevator algorithms
