On jeudi, f�vrier 13, 2003, at 12:34 AM, Tripp, Bryan wrote:
Bryan,
My guess is that data entry into a form, or choosing an option from a list,
requires more attention than typing free text (for someone who can type
well) ... but that's just a guess. Does anyone know of any research along
these lines?
Your words have prodded me into finally reposting some replies I received from the CHI (Computer Human Interface) list after forwarding them a resum� of a prior thread here last month that was called 'GUI vs TUI'. This is about two weeks overdue and I do apologize.
For recall:
Some members of Openhealth questioned whether a GUI was adapted to high-stress, fast-data entry environments frequently encountered in clinical settings. Some thought a good old Unix shell-type terminal mode entry window would offer a sleeker more aerodynamic and brain-mappable interface. These were generally veteran users for whom computers hold no mystery, but we all know, said others, that many HCPs would prefer not to touch a computer etcetc.
What I sent to CHI:
Dear list,I am, of course, solely responsible for the above summary, and hope it is moderately accurate. The views expressed below are non-filtered :
On another list that I participate in concerning medical informatics
issues there has been a recent discussion questioning the true interest
of relying on GUI-based entry screens in stressful clinical
environments. Many of the folks on this other list, mostly programmers
with medical experience or clinicians with programming experience,
believe there is a case to be made against GUIs in favor of more modest,
lightweight terminal or DOS-like data entry screens in many cases.
I do not come from a keyboard-only world and was a little surprised and
intrigued by all this.
Various arguments thrown about, more to place the markers than restate
the thread, were:
-GUI look nice but TUI are more efficient for quick entry
-Health care professionels (HCP) don't like or don't know how, or resent
having to use computers
-non-clerical HCP (ie, doctors) usually can't type quickly and grabbing
a mouse to move around a GUI form is too much to ask of them. Having
them learn a couple of keyboard shortcuts to move around monochrome
screens is probably reasonable.
-most clinical information is generated in the form of plain text
entries that are difficult but not impossible to formularize (this will
change, however, in the next 5-10 years, I think)
-the suits who buy clinical software expect a snazzy GUI, because it's
the only part of medical software that they can evaluate :-)
-you can't see an Xray or photo on a TUI
etcetc
I don't think the med list really expects the data-entry screens of the
Future to be DOS screens. But I've never seen the subject broached on
this list, and since it addresses real world usability issues, I hoped
the group might find it interesting to comment, point to studies, etc,
that bear on the GUI vs TUI debate.
Post to me directly and I will summarize to the list.
----------------------------------------------------------
1)
That seems to me a strange combination of a) assertions easily proved or
refuted by simple user research (like HC professionals preferences for mouse
vs. keyboard, or the preferences of the purchasers) and b) glaring gaps in
factual knowledge (like the format of the data that clinical information is
generated). They're not really "usability" or even "UI Design" questions,
are they? I mean, do people designing medical informatics apps really not
know how amenable health care professionals are to "computers" in general?
The generated format of clinical information, so long as it's somehow
structured, can be massaged by a programmer into a reasonable data format,
even if the source is plain text. (Unless you mean that clinical info is
written by hand.)
It's true that many users can learn to use a terminal screen, and for
experienced users, command line application use is faster, if there are many
commands to be invoked, etc.. For pure data-entry, a well designed and well
behaved GUI is plenty efficient for quick entry; observe standard UI
conventions like tab order and make keyboard shortcuts for everything and
you're good. There are surely plenty of studies that show the efficiency of
command-line interfaces for experienced users, of course the trade-off is
learning time and maybe accuracy.
It sounds from your comments that you're (maybe unconsciously) maintaining
an attitude that I've seen in many programmers: that GUIs are "pretty" or
"fancy" or "dressed-up" interfaces, which are mostly there to impressthe
suits or show off whizbang bells and whistles; and that command-line
interfaces are "power tools", rendered in rugged "monochrome" (preferably
white on green), where experienced users do their work. (None of these are
your words, of course!) Command line interfaces are hard to learn, and
require prodigious memory skills. Oddly enough, those are both qualities
that programmers prize highly, and mock when absent in others (see Cooper's
"Inmates" book for more). If you have a staff of hard-worn data wranglers,
whose only goal are speed, speed, and power, perhaps a terminal command line
or terminal interface is appropriate. If you're designing a product to sell
to a wide variety of users, forget that command line!
----------------------------------------------------------
2)
For reasons that escape me, many system designers just don't get it when
it comes to data entry. Admittedly GUIs do not make it that easy to do
data entry well, but it is possible. The design needs to focus on
keyboard use and while a mouse could be used, anything that *requires*
its is just broken. Designers also need to take proper account of
tabbing order and shortcut keys. Most of this is handled by the dialog
manager in Windows, but developers need to implement dialogs correctly
for it to work. It is also possible to implement keyboard-oriented
controls like a smart auto-complete entry field similar to those used in
Internet Explorer and Microsoft Office applications.
GUIs do have some advantages over other approaches, even for data entry.
They are easy to learn and of course, most users will have some
experience of them.
----------------------------------------------------------
3)
<snip/>
Basically though, there is no GUI vs TUI debate, just the constant 20 year long whining
from computer guys (those rare doctors who can actually program included) who do not want
to realize that they have become nearly one with the [expletive deleted] keyboards and
thus live in complete denial of the fact that 99% of humanity (even in industrialized
countries) finds the keyboard an alien way to use computers.
The American Journal of Medical Informatics has quite a few interesting studies which
point out the radical difference of things like the acceptance of PDAs (and NOT command
line interfaces) by young doctors and the allergical reaction of older ones towards
keyboards. In short no normal doctor ,young or old, likes a keyboard, so you can forget
about TUIs and the only way you can avoid the word GUI is if you consider PDAs and tablet
PCs to have some form of gestural text entry, or if you go to text entry by the way of a
Voice User Interface of the kind that has been developped for air trafic controllers.
----------------------------------------------------------
4)
Terminal Interfaces are *more* intuitive because they originate from the
'self' (insight moves sight to the site) and therefore employ reflexive
knowledge (i.e. learning to type, or 'which' button to push, is like
riding a bike ~ thoughtless).
Conversely, Graphical Interfaces are *less* intuitive because their
operation originates from OUTSIDE the 'self,' requiring determination of
'other' to achieve action.
[ Of course, even a Terminal Interface is graphically achieved (i.e. you
'see it' to use) -- but the terminal is bare-bones or first-order much
more closely than GUI (i.e. terminal is *static,* whereas GUI requires
manipulation of object to navigate to GUI object and then to manipulate,
like a mouse input device -- a second-order or further removed). I grant
touch-screens or tablet-like touch-screen devices may be more
*first-order* than terminals, but those are not yet widely available and
being GUI in nature are prone to retardation (i.e. a picture is worth
1000 words -- zzzzzz...) ]
Certainly Terminal Interfaces are 'more modest/lightweight' in design.
This modesty, importantly however, applies in both directions (input and
output) -- so that *useless additions* (graphical or what-have-you) do
not retard the process which is at-hand, just what is
terminal/necessary.
Obviously, where a graphical component is required -- it is required
(i.e. X-rays, etc.). However, dressing up the patients X-rays, medical
chart, or other medical presentations with snazzy colors, swoosh logos,
etc., intended, however they may be, to enhance or further differentiate
data/information ultimately *fails* because each addition is one more
step removed from original concern (like marketing). (i.e. Imagine
Grandpa Simpson relating a point of concern, and expounding in further
and further detail/particulars under the intention of clarification -->
zzzzzzz... corpses set to banquet ~ the problem of 'ad infinitum').
Since many of the folks on your 'other list' are computer programmers,
the most intuitive explanation would be to offer them this equation
which explains the above:
X = X + 1 (true, will run)
X + 1 = X (false, will not run ~ assignment is unclear)
This is the basic symbol set (0/1) for computers which clearly, simply,
and accurately *demarcates* two conditions from each other (true/false).
As is demonstrated by the equations -- riddle does not exist. If a
question can be framed at all, it is also possible to answer it.
In the same way, Graphical Interfaces often *are* riddles in that they
include (x+1) all sorts of additions in the first-order, meaning to
demarcate, but which are however *useless* because the assignment is
unclear ('external/other'). Whereas, Terminal Interfaces are typically
as clear (x) as we can make them (and therefore ideally suited for
high-stress environments) because they are 'snazzy-free' first-orders in
which cognition is not retarded.
I hope that makes sense to all. This is really basic 'Principles of
Mechanics' (Hertz). Anyone should feel free to ask me questions (on or
off list) which I would answer to the best of my understanding -->
http://derekrogerson.com/
----------------------------------------------------------
5)
I know this probably sounds trite, but this sounds like a classic example of where you need to talk to some actual users. And you might find that different users need different things. As you probably know, GUI's typically don't do a very good job of supporting power users, whose experience and knowledge allow them to key a limited number of choices much faster than mousing up and down lists. You remind me of what I face at least once during every project where I set with a group of developers, listening to them argue about what would be best for users. At some point I usually say something like, "We could sit around here and argue/speculate about this for the next three days. As an alternative to that, I suggest the rest of you find something useful to do while I go out and get some feedback from some actual users."
----------------------------------------------------------
6)
Interesting issue, and it reminds me of discussions in the field of user
interfaces for ERP software(ERP stands for Enterprise Resource Planning). I
worked in research in that field for some years. At the moment I worked
there (around 1995), a new generation of software was created: it was
Windows. But an often-heard complaint from the field was that the users
wanted their DOS screens back. They missed the shortcuts for fast data
entry.
In the new generation of software, the developers did not keep the shortcuts
intact for several reasons:
- development was not aware of how users used the original software
- shortcuts were contradictory to Windows standards
- development thought that Windows was a great improvement
So, my recommendation to you is to contact UI designers in the field of ERP
and find the detailed requirements for fast data entry. (e.g. via
http://www.sapdesignguild.org.) You might be able to use their input, but
always remember that your user population has it's own specific
characteristics as well, so doing your own user research is strongly
recommended.
----------------------------------------------------------
7)
I don't have any studies, but with Section 508, http://www.section508.gov/, a GUI needs to be keyboard traversable, which I applaud for my own selfish reasons. I prefer a nice-looking GUI, for aesthetics reasons, but if I can't use it via just a keyboard, then it's not worth using. I agree with the doctors and such that say keyboard traversal (and data entry) is much quicker than using a mouse (which doesn't necessarily mean you have to use a TUI over a GUI). For example, a GUI with a drop-down list (combo box) is very quick with a keyboard and prevents typos compared to a plain text entry screen. You might have a drop-down list of diagnoses, and when you TAB to the list, typing a letter scrolls the list. Typing the same letter shows the second item that starts with that letter. If the list had something like Delirium Tremors, Depression, Diabetes, Diphtheria, then typing a "D" twice while on the combo box would give them Depression and then they could tab to the next
field. Very fast and easy.
----------------------------------------------------------
Denny Adelman
