In personal relationships and social settings, there are numerous unspoken rules of conduct. Sociologists know that making certain rules explicit can actually reduce their effectiveness. It seems that in many cases, we know that we are following rules, but don't like to be reminded of that fact. Imagine a third party standing by, narrating your conversation with an acquaintance a couple seconds before it happens (this is possible because your greetings are so predictable: "Hey Peter, what's up?" "Not much, you?" etc...). This would greatly annoy most of us for highlighting the rule-based nature of our actions.
Another example: you've just said something perhaps slightly more callous than you should have, and the charming lady sitting across from you is no longer smiling. In fact, she's holding back tears. As she begins to cry, you offer sincere apologies for having said what you did. But what if, just before the tears began, she had said, "That wasn't a very nice thing to say--I'm going to cry now so that you apologize to me"? The outcome would likely be somewhat different. Even though both parties involved are aware of the rules of engagement, so to speak, having those rules made explicit can seriously detract from their effectiveness.
Designing systems to be clear and explicit is a truism of modern engineering. But as technology becomes more social--from today's IM and social networking sites, to tomorrow's agent-based systems and AI--how might this affect the design and use of devices and services? And if you think that this is irrelevant because computers are explicit by nature, why is it that we should adapt to the (lack of) social graces of our tools, rather than vice-versa? I think there is most likely a whole level of design thinking beyond our current horizon, because we're bound within software and hardware habits that were never very human to begin with. If we define "usability" only in terms of mouse clicks and monitor resolution, we're still designing for engineers rather than humans.