So robots are social animals after all

CYNTHIA BREAZEAL is passionate about the possibilities of robots. Although the robots themselves cannot be similarly enthusiastic - yet - the computer science professor from the Massachusetts Institute of Technology (MIT) believes that, before long, humans will be interacting with "social robots", robots that can read our emotions and express their own.

While we are growing increasingly used to robot labourers - in manufacturing, in surgery, even, thanks to the little Roomba commercial robot, in vacuuming our homes - Breazeal, in Dublin recently for a Science Week talk, takes "a more human-centric approach".

"Everyone has been so focused on robots as labour devices, but it turns out they're social devices," she says. "A social robot acts with a person more in a partnership."

Most focus is on how robots interact with people, she says. She's worked extensively at developing robots at MIT, most famously on the expressive robot heads, Kismet and Leonardo.

Kismet is a friendly-looking, gremlin-like head with a gentle voice, capable of showing an astonishing range of expression given that the mechanics of the head are clearly visible. Leonardo is a cuter Kismet, with hair. As videos of people interacting with the heads show, it is easy to forget you are talking to a machine.

"Kismet mirrors human-to-human interaction" by picking up on low level cues and incorporating goal-directed behaviour, notes Breazeal. In other words, Kismet is trained to read basic expression and gesture and to know how to move a conversation forward.

Studies show robots can be very useful as motivators, she says, noting that one experiment where they were used as weight-loss coaches helped humans stick to their slimming plans.

Advances in human brain science are helping researchers like Breazeal to construct more socially capable robots. This is because breaking down and having better understanding of human consciousness and emotion, and the small signals humans use to interpret what others are saying and feeling, help scientists to build robots better able to understand humans.

Research into brain functions and consciousness can sound more like computer science, with neuroscientists looking at how the human brain processes "data structures". Program a social robot to incorporate some of the basic structures, and "you don't interpret the robot's behaviour in terms of data structures, but in terms of having a mind".

Robotic science is of interest to many human brain specialists and sociologists because of the light that can be thrown on human behaviour and brains through understanding how robots might think and behave, and how people react to robots like Kismet.

"The process of building robots is a quest of self-understanding," says Breazeal. "But it is not something that is just a model, but something in its own right."

And therein perhaps lies the fascination with robots, the reason why we understand and feel fairly comfortable with social robots on film and television, the C3POs and R2D2s of Star Wars (which launched Breazeal's own lifelong fascination with robots), or Mr Data in Star Trek.

There's a "fuzzy boundary" that's very compelling for us, where we are willing to see robots as not human, but not exactly machine either, she says.

That third dimension is what fascinates her: "What does it mean to be human-like?"

The Japanese, open and enthusiastic about robots, she says, are very interested in this question. Much of the cutting-edge research in creating humanoid robots happens there, with some researchers creating eerily lifelike robots.

Too lifelike for some people - a phenomenon that Japanese researchers have dubbed the "uncanny valley". As robots become more human-like they become more "uncanny" and disturbing, and response falls into an uncanny valley of discomfort.

Many Japanese robots would be in the uncanny valley for westerners, but not for Japanese, who grow up with a positive association with robots and enjoy the possibility of future robot helpers, especially for their aging population. Japanese firms such as Honda have pumped huge research funding into friendly robot helper prototypes like the Asimo robot, which Honda predicts could be commercially available within a decade.

By contrast, in the West, "part of what the field battles against is people are afraid machines are going to replace people".

Yet, Breazeal notes that robots are already merging invisibly into day-to-day western life. There's the previously mentioned Roomba; cars with more and more built-in intelligence that is robotic; and robots that are working to deliver medications, or which enable a doctor to visit a patient through telepresence.

Breazeal again focuses on robots as social partners, and she points to fascinating work with autistic people that she and other researchers are involved in.

At MIT, one of her classes (along with human-robot interaction) is on autism theory and technology. People with autism often need to be taught to interpret facial expression, voice and gesture to understand human emotion, she says - just as they are trying to do with robots. Understanding how to teach emotion is very insightful for creating social robots. But in turn, social robots could prove patient, consistent teachers for autistic people, demonstrating a human expression and what it means.

"The cruelty of autism is not that people don't want social interaction. They crave it! But because of this disability, they have difficulty doing it. The question is, can you use robots as a scaffold to teach this?"

She also thinks they would be good for other therapeutic uses and for working with children.

A robot-infused future, she says, is simply a matter of time.

http://web.media.mit.edu/~cynthiab/index.html