From: "John Whelan" <john.whelan@alatto.com>
> Strongly disagree: t9 predictive text input is used by the majority of
users
> in Europe in my experience. It really works very very well and has evolved
> over the years with now very complete dictionaries in all major languages.
The paper Howard references describes a system that works. Notably,
however, lots of people fall back on their cute-but-unstandard systems of
abbreviation instead of predictive input. Why is this?
Anyway, "...the majority of users in [your] experience" sounds slightly
anecdotal. Does someone have hard figures on actual popularity of
predictive input?
> What can be simpler than one key input per letter?
"Things should be as simple as possible, but not simpler."
Can you reduce the whole user experience to Huffman coding?
Claude Shannon estimated that English text has a mere 8-10
bits of actual information per word. So how come we're all
reasonably happy with good ol' typing, which requires an
average of 6 keystrokes per word? Other variables enter in.
Back in the 70s, Xerox was doing lots of human interface work
on copiers. They tried out measures of frustration like these:
- "total time spent on figuring out and correcting mistakes in operation"
- "total number of mistakes"
- "total time spent on task."
Oddly, these correlated very poorly with subjective reactions to
interfaces. People who made more mistakes in, and took longer
to finish up, a particular operation on a particular copier, would
often as not still rate that copier as "more usable."
Why? Because a better measure of frustration turns out to be
more like "how long it took to solve the hardest problem in
a series." Making more mistakes, and taking more total time
backtracking and fixing--but solving each problem faster--could
actually make people feel better about the whole experience.
Admittedly this is kind of like that thing with vacuum cleaners
where the loud ones sell better than the quiet ones, irrespective
of actual cleaning efficiency. But that's people for you.
-michael turner
leap@gol.com
[Howard]
> The idea of predictive text entry of the sort "just press the first letter
> or so and the word pops up"
> comes up often in the field. It sounds so seductive, but it doesn't work.
> Meaning, it ends up
> being much more work for the user, much more mental anguish, and, finally,
> much slower, than
> just entering the word directly. (It can be made to work in very special
> circumstances, my comments apply
> to all the usual approaches which you see in commerce.)
>
> See:
>
> http://www.cs.strath.ac.uk/~mdd/research/publications/00dunlopcrossan.html
>
> for some research in this area.
>
> regards,
>
> hag
>
> Howard Gutowitz
> CEO, Eatoni Ergonomics, Inc.
> www.eatoni.com
Received on Fri Dec 21 15:06:20 2001