One-on-One

But when Veloso's soccer-playing robots win their first game against a team of humans, will they be good sports? Brian Scassellati, an assistant professor of Computer Science at Yale University, wants his robots to pick up on human emotions and respond accordingly. He takes a more biological approach to his research, building on theories and models usually applied to human learning and development.

"We have some ideas about how children learn to manipulate objects, to reach out and grab an object or learn that when an object passes out of sight that it still continues to exist," says Scassellati. "So, we try to take these models from developmental psychology and we try to build that model."

Robots modeled on natural processes, says Scassellati, would make human-machine interactions much more intuitive and comfortable for the human. Moreover, it would enable the robot to learn through social interactions.

	Alan with Kismet, an "emotional" robot at the MIT media labs who learns from social interactions with its programmers.

"We work quite a bit with the modeling of social skills," says Scassellati. "We want to build robots that realize what's happening when you make eye contact with them and we want them to be able to respond to social cues."

These socially savvy robots could in turn serve as good instructors. Scassellati's lab is working on a robot that can recognize and respond to social skills - or the lack thereof - to help diagnose autism in children, a spectrum of developmental disorders that robs children of their ability to relate to others.

A robot with social skills could provide a more objective baseline than even the most skilled human clinician, who may unwittingly bring slightly different emotions to each interaction with the child in question. The robot, by contrast, could be set to interact at precisely the same social level every time.

"One of the robots we're building right now is actually the same dimensions as the average two-year-old," says Scassellati. "We'll be able to see a particular child today, and in two months, and in three months, and compare performance over that time period with an objective observer."

Once a child has been diagnosed as autistic, the same robot could then be used as a therapeutic tool. Some research suggests repeated social interactions can help autistic children break out of the strange shell in which the disorder envelops them. But the therapy must be finely calibrated for each individual child and the sessions can require super-human patience. So why not use a non-human therapist?

"Initially, when we start off with a child," says Scassellati, "the robot would do only very simple forms of social interaction, maybe just eye contact. Then as the child becomes more proficient with that, we'd add complexity to each interaction."

Scassellati envisions these social robots breaking out of the roles we traditionally assign to robots - that is, performing tasks either too dangerous or boring for people.

"More and more, robotic systems will become part of our daily lives," he predicts. "Toys, or pets, or different devices are having personalities and trying to predict what it is that you want. Things are just getting smarter all the time."

- - - - - - - - - - - -
4 pages: | 1 | 2 | 3 | 4 |