Visualizing Better Human-Computer Interaction
by David Pescovitz
 Printer-friendly
version
Prior to joining the UC Berkeley faculty, Maneesh Agrawala was a computer scientist with Microsoft Research's Document Processing and Understanding Group.
|
When computers and people communicate, something is often lost in the translation. Essentially, computers don't know how we think. UC Berkeley computer scientist Maneesh Agrawala is helping bridge the gap. From designing systems that generate clearer driving maps to software that produces simpler step-by-step assembly instructions, Agrawala's research is about leveraging our understanding of how humans think.
"My interests are in computer graphics and human computer interaction, specifically perception and cognition," says Agrawala, who joined the Electrical Engineering and Computer Sciences faculty this fall.
Agrawala's aim is to tease out cognitive design principles that can be applied to computer graphics software. The key, he says, is to determine how information should be presented to us so that we can most easily perceive and understand it. For example, Agrawala's PhD thesis at Stanford involved the development of a system to draw route maps for driving directions. As soon as he began the research, he realized that a map that's drawn to scale isn't much use when you're actually driving.
"You can't see the main detail that matters to you, the turning points along the route," he says.
Three route maps for the same route rendered by (left) a standard computer-mapping system, (middle) a person, and (right) LineDrive. Note that the handdrawn map was created without seeing either the standard computer-generated map or the LineDrive map. (courtesy the researchers) [view larger image]
|
To find out how humans prefer to read a route map, he asked a friend to draw him directions from his house to work. Agrawala immediately noticed that hand-drawn maps are never drawn to scale and that the turning points are very carefully identified. Meanwhile, the lengths of short roads are exaggerated so as not to be missed, and unnecessary information and streets aren't included at all.
After a number of in-depth user studies, Agrawala designed a software package that automatically designs and renders route maps with a distinctly hand-drawn feel to them. The resulting system, called LineDrive, is now implemented on the Microsoft Network's Maps & Directions site.
"LineDrive was designed to show how to get from one place to another, but people want other kinds of maps too," he says. "If I'm going to have a party, I want to show everyone I've invited how to get to my house from all over the city."
At left, a set of TV stand assembly instructions drawn by a participant in Agrawala's study. At right, the assembly instructions generated by the automated system. (courtesy the researchers)
|
Agrawala is now working on a system that shows these multiple routes or approaches to a location on a single LineDrive map. The difficulty is that each individual route on a single map can't be rescaled separately from the others.
"The program has to consider all of the routes together in the graphic representation it generates," he says. "That's a much harder problem."
In another project, Agrawala and his colleagues examined the difficulty of assembling consumer products like furniture and toys. The included instructions are often tough to follow. That's because designers lack a set of design principles for producing them, Agrawala says. The researchers' goal was to create algorithms that a manufacturer could use to automatically generate clear assembly instructions. As with LineDrive, the first step was to determine what kind of instructions that humans prefer.
"We gave people a TV stand assemble but did not provide instructions," Agrawala says. "Once they assembled it, we asked to draw a set of instructions that would show another person how to build it."
It turned out that rather than present someone with an "exploded view" showing how an entire TV stand comes together, most people prefer step-by-step instructions with a single part added in each step.
Armed with that information, the researchers implemented the design principles in an automated system. The manufacturer need only enter a limited amount of data such as the geometry of each part, assembly orientation, and the sequence of the steps. The instructions are then automatically rendered in structural or action diagrams. A user study showed that the computer-generated instructions reduced assembly time by an average of 35 percent and cut the number of errors by half.
"With LineDrive and the assembly instructions project, we ask people to draw what's in their heads," Agrawala says. "Then we analyze the drawings to see how people cognitively understand information."
Maneesh Agrawala's home page
MSN Maps & Directions with LineDrive option
Lab Notes is
published online by the Marketing and Communications Office of the UC Berkeley
College of Engineering. The Lab Notes mission is to illuminate groundbreaking
research underway today at the College of Engineering that will
dramatically change our lives tomorrow.
Media contact: Teresa
Moore, Lab Notes editor, Director of Marketing and Communications
Writer, Researcher: David
Pescovitz
Web Manager: Michele
Foley
Subscribe or send comments to the Engineering Marketing and Communications Office: lab-notes@coe.berkeley.edu.
© 2005 UC Regents.
Updated 11/1/05.
|
