Graphics Reference
In-Depth Information
makes sense to save time where it's expensive—the user!—rather than where it's
cheap.
The discovery of an effective interface model—the WIMP GUI—to replace
the cryptic mechanisms of the past not only enlarged the market for computing,
but in doing so, enabled further progress by providing processor makers with large
economies of scale: The cost of developing a new machine could be amortized
over many more users.
The modern GUI had its origins in Sutherland's Sketchpad system [Sut63], a
CAD system that used a light pen and many (physical) buttons for its input,
and an oscilloscope for output. It included direct-manipulation tools, selec-
tion by pointing, grouping, constraint-based interaction, and many other ideas
that are being constantly reinvented even now. In the 1970s, at Xerox PARC,
researchers developed the WIMP interface in a form that closely resembles
the modern version (albeit in black and white), for a machine called the Alto.
Much of this design was adopted in the design of the Apple Lisa (and later
the Macintosh). In the decades since its introduction, it's come to dominate
interaction, and has only recently begun to be challenged by new multitouch
interaction and new interface devices like the Wii and Kinect.
While the framework provided by a GUI design like WIMP is a wonder-
ful stepping stone, developing a good user interface is still extremely difficult.
Although trial and error have their place in the exploration of possible designs,
effective designs need testing and refinement, and having a model of the entire
process of interaction, from the machine-dependent side (the pixel position of a
pointer, filtering of pointer tracks to remove noise, etc.) to the human (the user's
mental state, or his or her goals and sense of progress toward those goals, as in
“I'm trying to move this paragraph, and I've succeeded in selecting it ...”) is
critical to both of these. The study of effective interaction is the field of
human-
computer interaction
(
HCI
) [PRS02]. HCI is intensely multidisciplinary, involv-
ing hardware and software engineering, computer and mathematical sciences,
design arts, ergonomics, and perhaps most important, human sciences (percep-
tion, cognition, and increasingly, social interaction), not to mention cultural and
accessibility issues. It is, first and foremost, a design discipline, one where results
are subject to experimentation and validation.
Such usability testing is surprisingly complex. Consider the problem of com-
paring two interface choices: one easy to learn but with limited expressive power,
the other with great expressive power but difficult to learn. A good example is the
choice of function keys versus a mouse for selecting menu items. Function keys
are easy to learn, while using a mouse effectively requires several days of training.
(If you doubt this, try using your mouse with your other hand for an hour. Even
knowing all about the mouse, you'll soon find it's annoying you more than helping
you.) Which is better? The function keys or the mouse? The answer is, naturally,
that it depends: If you're going to be using the mouse for lots of other things as
well, the eventual benefit may be large enough to make it worth learning (and the
immediate benefit may be large enough to motivate you to do so). If doing this
particular task is a one-time-only event, then the simpler interface is almost cer-
tainly better. As a concrete example, Adobe's Photoshop has an enormous user
interface that takes quite a long time to learn completely. As a novice user, it
sometimes seems that everything you do makes the picture worse! But when used
