pardo@june.cs.washington.edu (David Keppel) (01/02/89)
There's an ongoing discussion about the desktop metaphor. I'd like to
recommend the work by Card, Moran, and Newll, published in CACM July
1980, pg. 396, and the later work by Roberts and Moran that appeared
in CACM, April 1983, pg. 265.
They built an analytic model of a user's interaction with a computer.
They focused on keystrokes and small combinations of keystrokes, and
built a model of how a user interacts with an editor. They then
tested several editors and compared the performance of their test
subjects with the performance indicated by their model.
Their performance metrics included:
* Time to complete a given set of editing tasks.
* Frequency/seriousness of errors.
* Time for a novice to learn a set of tasks.
* Range of useful tasks that are possible.
* How hard it is to recall infrequently-used operations.
* How many things at once does the user have to concentrate on.
* How fatiguing is the system.
* How much do user like the system.
Users were categorized by
* Knowledge of the task.
* Knowledge of other editors.
* Motor skills (e.g., typing speed).
* Technical expertise (e.g., programmers vs. nonprogrammers).
* Experience (novice, casual, expert).
Note that there is no single kind of task, and analysis of an
``average'' task will not reflect users who predominantly perform
another kind of task.
An interesting idea (borne out by their experiments) is that two
separate one-keystroke commands take much longer than one
multi-keystroke command IF THE USER THINKS OF IT AS ONE COMMAND.
Another interesting result is that across their sample, the slowness
of local motion (a few characters) with a mouse was offset by the
speed of nonlocal motion (many lines) using the mouse (as compared to
local/global motion using keystrokes).
;-D on ( Finger food for thought ) Pardo
--
pardo@cs.washington.edu
{rutgers,cornell,ucsd,ubc-cs,tektronix}!uw-beaver!june!pardo