Microsoft Researcher: Effective Visualizations Require User Trials to Test Design Choices

by   |   October 18, 2012 10:19 pm   |   0 Comments

Microsoft research data visualization

Above, images from a Microsoft research project into the best way to visualize data collected from sensors about a doctor’s empathic state dealing with a patient. The first design showed this relationship using a seesaw. Tests showed a more effective visualization used flower petals. Image courtesy of Microsoft.

SEATTLE—A picture is still worth a thousand words—but it has to be the right picture.

And the most effective way to ascertain which visualization is best is to test it with live subjects—using real people performing real world activities in, of course, the real world.

Related Stories

The OECD better life index empowers users to sort and analyze data.

       Read more»

Telling stories with visualizations: lessons from data journalists.

       Read more»

That’s the key message longtime Microsoft researcher Mary Czerwinski told VisWeek listeners in Seattle on Oct. 16, in an entertaining presentation keynoting the annual week of sessions devoted to translating complex data into visual representations.

“You never get your design right the first time,” said Czerwinski, who works in Microsoft’s Visualization and Interaction Research Group. “Only field research will tell you how real world users perceive your work. It’s just impossible to replicate real world conditions in the lab.”

Though Czerwinski described principles that derive from numerous broadly-focused research studies Microsoft has conducted over the past two decades, visual data representations—such as feedback reporting illustrations—have been key elements in all such studies. This dates back to efforts in the late 1990s to create new on-screen organizational schemes for website usage within Internet Explorer, during which Czerwinski and colleagues discovered that 3D-type cues, such as shadowing, were effective aids to help computer users organize material, even in two-dimensional frameworks. The result of that project, called Data Mountain, never reached commercial application—although cues such as shadowing are now commonplace online.

Simple, Clear Visuals Trump Fancy Images and Sounds
Another interesting discovery such early research yielded is that auditory cues are far inferior to visual aids. “We are very visual creatures,” Czerwinski declared, acknowledging a fact that underlies the entire field, nicknamed “infovis,” which drew more than 1,000 participants to the Seattle conference.

Fast-forwarding to recent projects, Czerwinski described lessons Microsoft learned in an effort to help physicians boost what’s called “clinical empathy,” and an ongoing study in the field of affective computing in which various sensors measure a user’s emotional state. Again, in each, visual representation has proven a key element—if effectively trialed and adjusted.

The clinical empathy project, called Entendre, is based on the fact that cancer patients show much better recovery rates when they are in the care of empathetic health care professionals. Could treatment improve under a training program that utilizes real-time feedback to clinicians about their empathic levels? At first, the study provided feedback to physicians using a visual scheme—a colored “sun” icon poised above a seesaw. This sun-and-seesaw visualization proved too complicated and it was replaced by a single image, a blooming lotus flower whose varying colors (warm or cold) and rate of unfurling depict overall empathy. Early tests showed promise and the program is soon to undergo real-world trial in a Seattle cancer care center.

“What we’ve found is that the best visuals must be simple—you cannot have ‘too much stuff.’ And they’d better be subtle and beautiful,” Czerwinski told her audience. “To be useful, people have to want to look at it.”

AffectAura, the project in which sensors monitor users’ emotional states throughout the day, is designed to help individuals identify and manage stress triggers. “Turns out that stress is a killer in today’s world,” Czerwinski reported. Again, the graphic devices (balloons) used to represent stress rely on warm and cold colors, and size, to reflect emotional states such as engagement and arousal (positive reaction). Now researchers are considering real-time signals, called “actuated external objects,” to provide instant messaging—a butterfly flapping its wings, say, to warn of heightened stress. This project elicited a half-dozen follow-up questions from attendees intrigued by the idea of having their computers, in effect, measure their bliss—or stress.

“If you’re looking for images that are instantly communicative to humans, nature has thousands—lightning, flowers, mountains,” Czerwinski explained of the butterfly and lotus graphics. “But please remember that we always start with this question: Is it even going to work?”

Visweek is an annual gathering that convenes data professionals from around the world for three corollary conferences—IEEE Scientific Visualization; Information Visualization; and Visual Analytics Science and Technology; as well as symposia on Biological Data Visualization and Large-Scale Data Analysis and Visualization. The founding sponsor of the conference is IEEE, the world’s largest technology professional association.

Eric Lucas is a Seattle-based travel, business and natural history writer. Visit him at


Tags: ,

Post a Comment

Your email is never published nor shared. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>