Visually group the author and the comment
Weblog plausibility: the top tep design mistakes

No confirmation dialogs or pages, please

Like I've said before, the fact that Google Web Accelerator makes "delete this resource" links dangerous does not mean that you should be going around adding confirmation dialogs or confirmation pages to your web applications. Don't do that, it's bad design.

Quoting from the material of an UI course at our university:

When the user is executing a task, he is always sure. It is not effective to ask confirmation beforehand. Users get used to the confirmation boxes and "are you sure" questions. At the moment, he is (nearly) always sure.

Afterwards, the user may find out that he has done something he would not have wanted to do. At this point, confirmation boxes or any boxes don't help. Instead, you should design an undo facility.

(Consult your favorite UI guru for more details. Alan Cooper's books are a good starting point, for example.)

We saw a great demonstration of this principle from a video clip taken from a usability test. We saw a person that used some archaic old software to play around with pictures and files. Played at normal speed, there was nothing suspicious going on. But when they showed what really happened, frame by frame, we saw how a confirmation dialog popped up on the screen and the user dismissed it by pressing enter. All this happened in just a couple of frames. It is impossible to imagine how the user could have had made a conscious decision to press OK to the dialog.

Avoiding confirmation dialogs is just one of the many principles or design techniques in the field of interaction design, but it is absolutely central in the sense that it's tightly coupled with perhaps the most important principle: minimize navigation.

I personally don't know whether this all means that you actually can't produce a great user experience with web applications (as of now), but if it indeed does, that means that we shouldn't make everything a web application! (Gosh, what a heretic I am.)