• Introduction
• Robots take the field
• Electric lady land

If you are used to technology being on a screen, the way that humans react to robots is a little surprising. People flip out, to put it mildly—especially when small blue and red robot dogs play soccer, as they did at the recent RoboCup robot soccer championship in Atlanta. Do not attempt to explain to two ten-year-old girls from Alabama who are screaming, “C’mon Red 7, you can do it! C’mon, pass the ball! Pass it!” that the dogs can’t hear and no amount of yelling will influence the outcome of the match—or that, because both teams use the same hardware, the Sony AIBO, what we’re really watching is a battle between algorithms and the dogs themselves are incidental. When you take technology off the screen and embody it in a form that is familiar, when you give it a face and legs for instance, people will project a personality upon it, and suddenly the tool takes on a life of its own.

Figure 1: Spectators watch the RoboCup 2007 “Four-Legged League” final.

The idea that technology can be more than just a collection of functions—that a tool can take on attributes previously reserved for animate objects—is not new of course. Adjectives such as “fun,” “friendly,” “intelligent,” and “sexy” are now routinely applied to tools like the iPhone in a way they never would have been to your grandpa’s tractor. In the age of robotics, however, the personification of technology is taking on new and sometimes unsettling dimensions. Creating these personalities, and crafting the interactions these tools have with humans, is one of the next great challenges for design.

According to Stuart Shepherd, president of the leading robotics firm Kuka, the pattern of adoption for robots is similar to that of computers, only faster. Like computers, robots started out being used by industry and the government. Now, however, they are starting to seep into the consumer market. The first applications are innocuous enough: robots shaped like Frisbees that vacuum the floor or mow the lawn and then put themselves away again. But more complex systems are beginning to make an appearance. The Mitsubishi Wakamaru robot has just started work as a receptionist at the People Staff agency in Nagoya, Japan. Wakamaru can recognize faces, carry on simple conversations with a vocabulary of 10,000 words, and perform manual tasks. Wakamaru can also guide visitors to their destination within the building and sing a special “thank you” song.

In terms of interfacing with humans, robots are just getting started—a parallel might be the Apple II. But consider this: the Japanese government is investing $35 million a year in developing robots. By 2015, they expect the industry to have grown to $10 billion, which means a robot in almost every Japanese home. Shepherd believes that the American market for robots won’t be far behind. He expects that in the next 10 to 15 years domestic robots will comprise 85% of their U.S. sales.

Figure 2: A humanoid reaches for a Coke in a simulated home environment.

The reason for this push towards automation has less to do with robot dogs and screaming girls than it does with grandmothers who need help with the chores. The Japanese population is rapidly aging and there are fewer young people to care for the elderly. In robots, the Japanese government sees a multifunctional tool that can do the yard-work, make tea, tidy up, and alert the authorities if there is an accident. But robots can only provide this assistance if they are able to communicate with grandma without scaring her to death.

In the terminology of the robotics industry, the ability of humans to interact with robots is often called “acceptance.” It’s an interesting choice of words because it assumes what the rest of us are only beginning to suspect—the robots have arrived.

Robots take the field

There seem to be two basic approaches to robotics, both of which were on display at the RoboCup soccer competition held in Atlanta in July. The humanoid robots play at about a four-year-old’s level. They miss the ball, they fall over, they get distracted by the crowd. One reason they play like four-year-olds is that they face many of the same challenges. They aren’t great with balance—leaning left to lift the right foot; kicking forward without falling backwards. They also have a hard time judging distances and aiming accurately. Because the camera they use for visual tracking is up on top of their head, it is constantly rocking back and forth as they move so they have a hard time keeping a bead on the ball.

Figure 3: Members of the champion NIMBRO humanoid kid-size soccer team.

The robots shaped like overgrown hockey pucks, on the other hand, are soccer playing machines: they sweep down the field in perfect V-formation, scatter, then stop on a dime. They have rotating flippers like the beater-bar on a vacuum cleaner that snap the ball down the field at foosball speeds. They cover their opponents, avoid offsides, and are penalized for pushing. They have special attachments that allow them to punt the ball up, bouncing it off the lid of their opponent, a move which they execute with such skill that this year there’s a new rule—no bouncing on the lid of the goalie.

Figure 4: On screen interface, and small-size “hockey puck” robots playing soccer.

The secret to the hockey pucks success is that they are essentially soccer-playing appliances built to make the best use of available technology. They’re wirelessly linked to a mainframe, which knows where each puck is and rearranges the pucks, calculating shots and analyzing the odds. Each team is also allowed a single overhead camera that helps them keep track of their opponents. The humanoids, on the other hand, are restricted to the computing power they can carry with them and the sensors that fit in their tiny heads.

If the technology exists to make hockey pucks that can bend it like Beckham, what’s the point of having this league of hapless humanoids tripping over their two left feet?

The answer is that to interact effectively with humans, robots need to understand how humans experience the world, and that includes understanding our limitations. “Before robots can really be integrated into our daily lives, they need to share our mental model. They need to understand the world the way we understand it,” says Dr. Henrik Christensen, director of the robotics center at Georgia Tech. They also need to be able to get up the stairs.

“Let’s take a simple scenario,” says Dr. Christensen. “You get a domestic robot, you open the box, and it says ‘Hi I’m Timmy, take me on a tour of your house.’ So you show Timmy around—this is the kitchen, this is the living room, this is a closet—now all these locations are connected by a hallway. To you, the hall is unimportant; it is a space for passing through. To Timmy it’s another room. Timmy needs to be able to form a hypothesis and then verify that hypothesis. He needs to ask, ‘This is a hallway, right?’ Right now in robotics we are very good at mapping things out and moving pieces around but we are much less good at building dialogues.”

Figure 5: A humanoid robot receiving instructions.

This idea of dialogue with the user will be familiar to any interaction designer. The race to build a domestic robot is as much about interface as it is about hardware. And robot designers are putting the “face” in interface design. Not only is dialogue important, but the robot needs to react appropriately to what is being said. Studies have shown that when spoken words are combined with the appropriate facial expressions, people are twice as likely to understand what is being said and much more likely to remember it. Dr. Cynthia Breazeal of MIT describes the facial hardware of her Kismet robot: “Each ear has two degrees of freedom that allows Kismet to perk its ears in an interested fashion, or fold them back in a manner reminiscent of an angry animal. Each eyebrow can lower and furrow in frustration, elevate upwards for surprise, or slant the inner corner of the brow upwards for sadness. Each eyelid can open and close independently, allowing the robot to wink an eye or blink both. The robot has four lip actuators, one at each corner of the mouth, that can be curled upwards for a smile or downwards for a frown.”

Electric lady land

But robot design is about more than just another pretty face. Leading robot designer Tomotaka Takahashi has spent the last year or so perfecting a particular strut. It involves keeping the arms straight, lifting the knees high, and a having a distinctive swing in the back panel. The FT, or female-type, robot stops and places her hand on her hip, cocking her head to one side. The key to the FT's movement, Tomotaka explains, is that it is designed along an “X” shaped frame that pivots at the waist, rather than an “H” shaped frame, which produces the standard robot stagger.

Tomotaka created the FT because he believes that gender stereotypes are preventing robots from getting ahead in the world. “Most robots are perceived as strong hero-types,” he told me. “Very bulky and menacing. Good for fighting but you don't want to live with them.” The FT, on the other hand, is slight and stylish—almost coy. He believes that the FT will be more readily accepted in the home, and points to studies that show that female personas are generally perceived as more welcoming and less intimidating—which is why most automated call services use female voices. By challenging gender stereotypes about robots, Takahashi believes that in the future FTs will co-exist comfortably with human, serving an essential function as translators between humans and machines. But first, he says, he had to figure out how to hide the wires. “It is very difficult to conceal all the mechanics inside a slender form while maintaining proper balance—especially on smaller feet.”

Figure 6: The FT (female-type) robot.

The FT is just one example of how robots are providing a third dimension to principles that, until now, have applied primarily to screen-based media. Takahashi’s method for creating realistic movement will be familiar to most designers and animators: “When designing robots,” he says, “I try to include excess motions. Robots that move too rationally look unnatural. Like when you wave your hand, your shoulders and your head move too.”

It turns out that our desire to have robots that are “more like us” does have its limits. One of the key concepts in designing humanoid robots is the “uncanny valley,” a phrase that was coined by Japanese roboticist Masahiro Mori to describe what happens when robots resemble humans too closely. Suddenly, the same characteristics that enabled us to relate to the robot cause us to react with horror and repulsion. An example might be the work of David Hanson, who made a realistic robotic version of his girlfriend’s head mounted on a plank.

An interesting characteristic of uncanny valley is that it is different for different cultures. According to Dr. Christensen, “In many Asian cultures there seems to be much more of an emphasis on companionship and they are more comfortable with replication—the valley is much narrower. As a European [Dr. Christensen is from Germany], humanoid robots creep me out. Americans and Europeans tend to be much more concerned about functionality than they are with replication. We are building robots to complete particular tasks. We want clean floors, for instance.” These cultural differences are having an impact on the way that technology is being developed. Dr. Christensen summarized this impact thusly: “So, the Japanese build AIBOs [robotic dogs] and Americans build Roombas [robotic vacuums].”

Figure 7: The Kokoro Actroid DER2 feminine guide robot in Japan.

Even if skilled designers were not desperately needed to save the world from dysfunctional robots insisting on “acceptance”, robots raise fascinating issues surrounding how we humans perceive the world. Listening to robotics engineers discuss their work in the soccer tournament is like sitting with the angels on the fourth day of creation. At a symposium following the soccer tournament, a session entitled, “A new protocol for motion pattern generation towards behavior abstraction” raised the following question: “When a ball passes behind an obstacle, how does the robot know that ball still exists? And when it come out on the other side, how does it know that’s the same ball that disappeared?”

A more mundane, though no less important issue, that continues to fluster robots is laundry. How does the robot know that the crumpled up shape lying on the bedroom floor is a pair of pants that should be picked up and put in the dirty laundry? The robot knows that something is on the floor that doesn’t necessarily belong there—that there is an anomaly. But the wad of cloth doesn’t look like pants anymore. The robot could decide to pick up the object to see if it will regain its characteristic shape, but then it might risk shaking out the cat by mistake.

In the face of such complexity, some believe that we will soon reach a point where designing the next generation of robots is a task that can only be handled by other robots. Roboticist and author Ray Kurzweil lays out this scenario through an imaginary dialogue in his 2004 book, The Singularity is Near.

“Charles Darwin: If I may interrupt, it occurred to me that once machine intelligence is greater than human intelligence, it should be in a position to design its own next generation.

Molly 2004: That doesn’t sound so unusual. Machines are used to design machines today.

Charles: Yes but in 2004 they’re still guided by human designers. Once machines are operating at human levels, well, then it kind of closes the loop.”

For the next generation of designers, the primary question may not be what technology can do for them, but rather what skills and attributes remain uniquely human.

Where to go from here

If you enjoyed this article, check out these other great articles on Design Center:

• Bespoke futures: Media design and the vision deficit, by Peter Lunenfeld
• Uncertain futures: A conversation with Professor Anthony Dunne, by David Womack
• The People will be heard: Interactive technology in public spaces, by Jennifer Kabat

About the author