#1 3-D perception

[John Dubberley <dubb () navo navy mil>]

The purpose of computing is insight, not numbers.
                                                -Hamming

My corollary: The purpose of computing is insight, not pictures.

All this discussion of on the inabilities of others to perceive 3-D
graphics has become too tempting for me to remain silent. I sympathize
with Dan Karron because he probably gave his project managers exactly
what they asked for when he built his 3-D graphics system.
Unfortunately, what they asked for wasn't what the users actually
wanted even if they were the ones asking for 3-D representations
themselves. This and other recent posting about computer practice
brings up one of my pet peeves. The general lack of adherence to, and
instruction in, the commonly accepted graphical presentation practices
by the scientific and computing societies. As one of my coworkers
states,"If you can't present it in an X-Y graph, you aren't at the
heart of the matter." I am not implying that any particular posters
are guilty of this sin, but they do voice the attitudes of the guilty.

The less complex the presentation the clearer the concept is
transmitted. I don't know how many times I have seen 3-D plots and
color plots misused to snow management and program managers. The less
computer able seem to be the easiest to snow. They are so taken in by
the pictures they gloss over the fact that they don't know what's
going on. (Managers hardly ever know what is going on so they at least
like the entertainment of pretty pictures.) Users on the other hand
tend to without fail say give me X-Y plots, contours, or nothing. This
is understandable they want useful easily read data, not art for their
walls.

Among all the presentations I've had to sit through this summer, the
best paper by far was "The use and misuse of color in the scientific
visualization of acoustic phenomena" by V. W. Sparrow of Penn State.
He stated that any system of visualization adopted by a programmer or
a scientist will always make sense to the creator. The art comes in
when that same level of clarity is to be transmitted to others. This
is where the computer science and research science types need their
humility. Since you are always working for others *you* have failed
when your graphics are not clear. This ties back to earlier appeals on
this network for clarity and less reliance on equations in the writing
of publications. Covering up your own inadequacies by blaming the
audience is no excuse.

At this conference he gave some helpful hints on better presentation. His main
points with a few of his and my commentaries were:

1) 99.5% of all data can be most effectively presented in black and
white.  Adherence to this rule also saves the author those pesky color
page charges most journals bill.

2) ROY G BIV spectrum plots make very pretty pictures. They are also
the worst way to set one's color palette. His example for this was to
quickly ask the audience with no outside info, which was larger?
Yellow or Green. The percentages were almost even. For better palettes
look to the work of Cartographers who have spent centuries working out
the details of presenting 3-D data in a 2-D format. Use shading and
only change color to imply a state change.  For example nearly all
physical geography maps show water as blue and vary water depth by
shading. Above sea level, elevations are represented in shades of
brown and green. Also notice that the boundaries between levels are
emphasized by a line.

3) Never use red and green together. 10% of the male population and 1%
of the female population are color blind. Due to the unfortunate
gender skew of the present scientific population using these colors
together automatically looses 8% of the audience. Also He-Ne laser
pointers will disappear when observed by most of this group.

4) Nearly always a contour plot is more useful than a 3-D perspective
representation. Look at a topographical map. With a sufficient degree
of accuracy the elevation can be determined at any given point. Try
this with a wire grid 3-D perspective. Whoops! The point I want
happens to be behind the hill. Even if the point is visible, it is
usually very difficult to acquire an accurate elevation guess. How
many times in our careers have we all had to blow up a 2-D graph and
read off data points from someone else's document? This becomes
impossible with wire grid perspective plots. If wire grids or shaded
surfaces were better at presenting the data even at a higher cost,
they would have been commissioned by the various geographic
organizations. One can argue that surface elevation changes are near
flat making them more applicable to contour maps than perspective
maps. I would challenge this by mentioning cave maps. They are never
perspective maps either.

The kicker to this presentation was that the next presenters in this
visualization session ended up being excellent examples of how not to
present data. They were showing off their fluid dynamics plotting
package. Each picture was beautiful and lush with ROY G BIV color and
most had wonderful 3-D shading also. However at nearly every slide we
heard the words "Now it is kinda hard to see here that ____, but if
you look closely enough you might make it out." Each and every slide
was gorgeous. Each and every slide was flawed.

John Dubberley

-----------------------------