|Current Issue||Past Issues||About YSM||Subscriptions||Advertisements||Contact Us|
|80.3 - Spring 2007|
|Search YSM Articles|
> Spring 2007 > Articles
Social Machines:Using robots to study ourselves
By Ben Tannen
Kismet, a robot at MIT used to study facial expressions.
Can robots think? More importantly, what are they thinking? According to associate professor of computer science Brian Scassellati, these questions are philosophical and don’t have simple answers: “We can watch robots’ behaviors and interpret them and get robots to trigger certain behaviors in observers. Is that thinking?”
Whether or not robots think like humans do, Scassellati believes they are useful tools for studying theory of mind. Theory of mind is a concept central to cognitive science that helps explain how people interpret the world around them.
According to Scassellati, theory of mind is “the way we represent the beliefs, goals, and desires of others. We assume certain things we can’t directly perceive, see, or touch—this is part of our understanding of others.”
Humans are very capable of making relatively accurate assumptions based on their observations of their surroundings and how other individuals behave. Every day, people act based on what they think others know or want.
Empathy, the concept of putting yourself in another person’s shoes and relating to his situation, is a good example of theory of mind at work: without such a skill, a teacher may not understand that a student is pointing at the ball because he wants to play with it, for example.
How are humans able to attribute hidden meanings to the actions of those around them? Theory of mind is based partly on experience and training throughout life. However, there are also certain innate components of the mind that allow people to interpret the actions of others. The goal of Scassellati’s research is to use technology and robotics to study some of the innate aspects of human social development.Imitating Children
What are the simple social skills that children possess that later lead to a welldeveloped theory of mind? Scassellati alludes to his personal life to illustrate how young children do not yet have the ability to process information as adults do.
“I have a two–year-old son who, when he knows something, thinks that everyone else knows it too. If he hides something of mine, he thinks it is perfectly obvious that I should know where it is. Children don’t have the ability to represent what is in others’ heads when it differs from what is in their own heads.”
In order to study this interesting stage of human development, Scassellati’s lab has created Nico, an upper-torso robot whose body proportions are designed to match those of a one-year-old child. Nico is anthropomorphic, which means it is shaped like a human, although it does not actually look like a person.
“We use Nico to implement some basic social abilities in order to study them,” explained Scassellati. Basic sensory motor tasks are an important aspect of Scassellati’s research; he tests, for example, Nico’s ability to point to specific objects.
He also investigates Nico’s cognitive abilities, such as whether Nico can recognize objects to which someone else is attending. Nico can see someone staring across a room and understand that location as being where the person’s attention lies. It also has the ability to discriminate itself from the environment, just as a child understands that he is separate from the world around him.
Another area of research involves the ability to learn simple words and proper grammar. For example, Nico has taught itself the difference between using the words “I” and “you” in sentences.
Finally, Scassellati’s lab is studying sensory processing, which enables people to make sense of their surroundings. It also involves comprehension of peoples’ tones of voice, which are crucial for understanding the mood of a conversation; whether someone is yelling at you or praising you is a good indicator of how you should act. Most people learn these simple skills within the first year of life.Computer and Cognitive Science
A robot would theoretically be able to distinguish itself from its surroundings if it were able to remember all of the possible positions of its own parts. Then, when the robot looked at the world around it, it would be able to differentiate itself from its environment.
Given the complicated mechanical structures that comprise a robot’s body, this approach is extremely difficult and impractical.
Scassellati therefore studies the underlying algorithms and processes that robots use to accomplish specific tasks without human intervention. These algorithms are related to the timing of the robots’ movements.
As Scassellati noted, “The robot knows when it issued a command to its motors. It doesn’t necessarily know what this command will produce, but some amount of time later, we will see a body part move.”
Scassellati and his colleagues look for a predictable time lag between sending a command to a robot and its subsequent response. The problem with this method is that it is difficult to know what this time lag will be. The work of graduate student Kevin Gold has developed a system that allows a robot to learn that actions that occur within a certain time frame after a command are its own.
In one experiment, Nico learned to drum in time with a human musician.
Scassellati hopes that researchers will eventually be able to engage with Nico and future robots naturally, and that the robot will be able to pick up and understand the natural social cues that humans take for granted. Such an interaction would allow humans to integrate robots into their everyday lives while helping people to better understand their own thought processes. Furthermore, Scassellati hopes to use robots to aid in both diagnosing and treating people with autism, a disorder often characterized by an impairment or lack of the normal theory of mind. Gradually assembling further insights through studies with robots like Nico, Scassellati explains that, “as we put all of these tiny pieces together, we get closer and closer to understanding theory of mind.”
ABOUT THE AUTHORBEN TANNEN is a sophomore history major in Saybrook College. Robotix, Capsela, and K’Nex took up a decent amount of his childhood.
ACKNOWLEDGEMENTSThe author thanks Brian Scassellati for taking his time to explain the intricacies of robotics in easily understandable terms.
FURTHER READINGBreazeal, C. & Scassellati, B. (2002) Robots that imitate humans. Trends in Cognitive Science, 6(11), 481-487.
Copyright 2013 Yale Scientific Publications, Inc. - Disclaimer