Skip to Navigation
University of Pittsburgh
Print This Page Print this pages

May 13, 2004

Visualizing the Numbers

What do travel times to the airport, the relative impact of HIV on the countries of the world, medieval manuscripts, Prozac patent activity and FBI homicide statistics have in common?
They’re all among the wide variety of analytical projects tackled by the Visualization Information Systems Center (VISC), part of Pitt’s School of Information Sciences.
Directed by Ken Sochats, assistant professor of information science and telecommunications, and boasting a multi-skilled team of six staffers, VISC enables its clients, such as the FBI, the Pittsburgh Public Schools, the Greater Pittsburgh Airport Authority, auditing firms and other private sector organizations, as well as faculty researchers, to manage and analyze organization-specific (and often mammoth) data sets and to present information in visually informative and malleable ways.
Replacing old-school statistical, tabular and graphical techniques with dynamic visual presentations, VISC designs systems catering to a user’s needs.
“Basically, we built the Visualization Information Systems Center because we were already doing a lot of things that had a visual base, like [computer] graphics, which is what I teach,” Sochats said. “I defy anybody to think of any activity that doesn’t include visual information or assistance, so in a sense we fit into everything.”
While the projects run from soup to nuts, work at VISC falls into three basic domains: spatial analysis, scientific visualization and action systems, Sochats said.
“Spatial analysis includes using geographic information systems to analyze various phenomena, and that may range from aerial photography to mapping and routing,” he said.
Scientific visualization is the presentation of information – such as what goes on inside a blast furnace or how air flows over an airplane wing – in easily comprehended visual form.
“Action systems include created animations of real systems, and using virtual reality and modeling to portray various phenomena,” Sochats said, ranging from a virtual tour of the Pittsburgh campus for Homecoming activities to virtual presentations of key areas inside buildings for use by emergency first responders.
VISC does not develop visualizations per se, he noted. “We essentially develop systems. A visualization is like a map; it’s a snapshot. However, a mapping system is what we would develop for navigation, routing, presentation – all of which is much more valuable. Someone once said, ‘A picture is worth 10,000 words.’ Well, a visualization system is worth 10 million words, because I can continue to create new views of the data.”
As an example, Sochats cited Project VIPER (Visualizing Information for Pittsburgh Public Schools Education and Research), a collaboration between VISC and the Pittsburgh Public Schools (PPS).
“PPS is now in the midst of complying with the No Child Left Behind Act,” he said. “Most schools are now reporting average yearly progress, just lumping everybody together.”
VISC has taken the student data sets from the PPS real-time information database and created a district-wide atlas that can be accessed for a multitude of demographic variables. “You can map a point for every student, colored-coded by the school they go to. You see where a student lives in relation to where the school is.”
If a dot representing a student appears outside the school district boundaries, a systems user can determine why by accessing the student’s records via “drilling” – the term for returning to the original data. “When you drill into the data set by clicking on the dot, you’ll find all the student information: gender, race, participation in school lunch, what grade they’re in, etc. One of the basic principles in visualization is that records never lose their integrity,” that is, whatever comes up on the screen is generated from the original data. “That principle is violated in statistics and other aggregated techniques. But visualization presents the fundamental data without change.”
If the original data are inaccurate, the dots’ positioning often flags the mistake for an expert, Sochats said.
The task for VISC is how to take extremely large data sets and allow the user to extract the needed information. “If I’m a teacher, how do I tell that this student is in trouble or is doing well? What kind of interventions are needed, what kind are working, etc.? How do we visualize and present information such that the user grasps it right way?
“Our ability to maintain the integrity of the individual’s data allows decision-making where you have to go down to the individual student level,” Sochats said. He added that the VIPER project puts PPS way ahead of the national curve for No Child Left Behind compliance.
Similarly, a user can drill into an individual school’s data, for example, to find a map of a school’s bus routes. “Rather than taking the average of where all students live in relation to a school, we can take a look at where the bus route should be in relation to all students,” he said. In addition, the area’s terrain and natural and human-made boundaries, such as the rivers and parks, or bridges and highways, can be factored in.
“If we consider Pittsburgh’s overall school districting problems, for safety purposes you don’t want a kid crossing a river or going through park areas,” he said. “So we’re able to look at all of the feeder patterns for the schools. Rather than using streets, which are not straight in Pittsburgh, we’re looking at zoning alternatives.
One thing Project VIPER has discovered is that school feeder patterns were historically done by zip codes, “which are not particularly useful for allocating students to schools. Now we’re adding some logic to this, taking into account our terrain.”
These kinds of factors can inform decisions on school closings and allocations of resources, he said.
In this process, generally defined as data mining, VISC uses three basic approaches, Sochats said. “First, there’s statistical or algorithmic analysis. Second is artificial intelligence, where we try to do just the opposite: In statistics, we first compile some data and try to interpret the results; in AI we try to encode for experts what they’re looking for and then layer that onto the data,” he said.
The third approach is visualization with the human user in mind. “We assume that the human is part of the system. People bring their own expertise to the keyboard, and they often have expertise in the domain. We want to present the data, not only to provide the most leeway, but such that it reaches out from the screen and slaps the user along side the head and says, ‘This is the effect, notice me. Wake up and notice me.’ So the challenge is: How do we present data such that it conveys the correct message and so a user can recognize data, can manipulate it and can navigate through it?”
Sochats said the visualization systems are analogous to the steam shovel as an extension of a person’s arms. “This is an extension of our sensing and decision-making capacity.”
While the VISC systems have different visual renderings depending on a user’s needs, they share the same underlying architecture. “We built to a common architecture that allows us to swap in different kinds of visualizations so we can adapt the visualization to the best portrayal of the data,” Sochats said.
In designing these systems, Sochats keeps in mind the famous Rorschach drawings. “There are people who will say, ‘I don’t see a darn thing’; and others who say, ‘This is intuitively obvious to me: I can see immediately what is happening here.’ So we often give people options as to how to portray the data.”
After only five years in existence, VISC rarely turns clients away, Sochats said. Clients include corporations, government agencies and individual faculty members. “We’re almost like a Rand Corp., and now that we have a set a skills, we can go after different projects,” he said.
VISC also has about a dozen active funding sources, including the commonwealth, the Provost’s office, the Ben Franklin Trust, the National Endowment for the Humanities and the Mellon Foundation.
Sochats added that while he is the public face of the center, pitching projects to various clients and making presentations to organizations in all sectors, the center could not survive without a team approach.
“Teamwork is essential for a center like ours that involves lots of different skills to come together, and the fact that we’re dabbling in everything,” he said. “We have an extremely powerful team which allows us to go out and attack anything.” The team includes Matt Kelley, project supervisor, who coordinates daily operations; Matt Nicol, systems administrator, who takes care of the servers and networks; Sean Ellis, project supervisor, who is the artist-in-residence and Robert Regan, visiting research professor, who is an expert in geographic information systems.
Two additional staff on loan from the PPS during Project VIPER are Glenn Ponas, program manager for strategic data initiatives and project manager for the VIPER project, and Chris Temple, GIS analyst who generates maps.
-Peter Hart

Leave a Reply