Ideally, users will prefer a system that's faster, easier to learn,
and that allows them to make fewer mistakes.
That's not always the case.
Sometimes we could build a system that has good Speed, Efficiency, Learn-ability and
Memorability, but the users prefer it less than a system that has
worst usability on all the other measures.
This has actually happened.
There's an example from an older paper,
from The University of Maryland, where users were shown two map systems.
The first one, shown here, has an overview window in the corner.
In this case showing a map of the state of Washington, with a little context window
that shows what's zoomed in, in the main part of the screen.
This is one of the interfaces users we're given, and
the other interface would start with the full state zoomed out like this and
users could actively zoom in much like you can do with Google maps now.
Users were given a set of tasks, to do with these maps.
Their Speed and Efficiency were measures, and the users were
also given a questionnaire to fill out at the end to indicate their preferences.