Science & Tech

Visualizing science focus of panel

5 min read

Researchers discuss how to see the pictures in data

The huge load of data now coming from modern computer systems is so overwhelming that new methods must be devised to allow people to visualize the world in more understandable ways, scientists involved in computer graphics areas said in a special symposium today.

“What can humans actually understand?” asked Harvard Professor Alyssa Goodman, who arranged thea symposium on “Seeing Science” at the annual meeting of the American Association for the Advancement of Science, held February 14 through 18 in Boston. “We just don’t have the bandwidth and the receptors” to handle all the data now coming in from various disciplines, including astronomy, medicine, and the social sciences.

One problem, explained Goodman, an astronomer and Director of the University’s Initiative for Innovative Computing (IIC), is that an increasing number of dimensions continue to be added to data files. Two-dimensional images are generally understandable by most viewers, but adding dimensions such as time, color, movement, and connections make analysis horribly complicated. “You can have serious problems when you get beyond four dimensions,” she said.

Still, said Felice Frankel, Senior Research Fellow in the IIC, humans are “still far, far better at pattern-recognition than any computers,” despite the loads of data accumulating and huge power of modern computers. 

York University Psychology Professor Michael Friendly, speaking of the history of graphic presentations, noted that some topics are much more understandable when presented in charts, graphs, and drawings. He pointed out that a famed diagram showing the decreasing size of the Emperor Napoleon’s army en route to Moscow and back is a stunning example of quick, easy, and important communication.

“Every single picture has a story,” he added, but the story depends, in part, on “how it’s (the picture) related to other things.” Indeed, some things might not have been discovered — such as the relationship between sunspot activity on the solar surface and cooling of the Earth’s atmosphere — if observers keeping track hadn’t made diagrams of the sun. By diagramming the surface of the sun over a period of years, researchers spotted a “butterfly” pattern showing an 11-year regularity in the eruption and migration of sunspots that’s apparently linked to a 70-year-long era of near-absence of sunspotsthe spots, now called the Maunder minimum. During that time, northern Europe recorded exceptionally low temperatures.

Thus graphics, Friendly said, can often have “immediate and high visual impact” that sometimes lead to discoveries that might not otherwise be recognized.

Computer scientist Christopher Johnson, from the University of Utah, said humanity has already experienced “the data Big Bang,” meaning that in the year 2003 humans — (via all methods, and abetted by computers) — produced more data than had accumulated in the entire 40,000-year span of human history. That pattern has continued, with that much and even more data accumulating every year since 2003. “We’re going to have to understand it [that data] and use it,” he said.
One way to get control of the problem, Johnson added, is to convert the data as much as possible into visual forms so that “we can understand some of these large-scale sets of data. This is how we’ll progress in science,” — with the aid of high-performance computing.

Johnson noted that “the challenge is” for computer scientists to work with bench scientists “in true collaboration, to go beyond what either of us would have been able to do alone.”

Frankel, who helped chair the session, added there are some areas of research that are still too difficult to visualize, even with ultra-powerful computing. “For a lot of science we don’t have the tools yet to capture what’s going on. Quantum mechanics,” for example, “simply can’t yet be visualized because we don’t have the tools.”

Alyssa Goodman also noted that when confronted with spectacular computer-generated scientific images, “people confuse it with art. People get overwhelmed by the colors and can lose track of what it is” they’re looking at.
But Friendly advised pointed out that “there have always been learning curves for the new ways to present data.”

According to psychologist Barbara Tversky, from of Columbia University, the ability to communicate with graphic images is “a uniquely human” talent, and that “only humans create tools for their own mental well-being. ”We diagram how to do things.”

Harvard Psychology Professor Stephen Kosslyn, who has carefully studied the nature of human visual imagery, gave guidelines for using graphics in communicating. First, he advised, a speaker should use the Goldilocks Rule — give just the right amount of information, not to much and not too little. Second, is the Rudolph Rule — make the central point stand out (like the reindeer’s red nose) from the rest of the date. Third is the Rule of Four — make four main points because the audience will forget anything more. And fourth, the Birds of a Feather Rule — lump similar images in the presentation together, to avoid confusion.

Also, in a discussion of the current uses of “Power Point” presentations, Frankel said the practices discussed “may seem obvious, but you’d be flabbergasted to see what some scientists do with Power Point.”