Study shows why human-like robots give people the creeps
By Ben Coxworth
July 18, 2011
People seem to enjoy watching robots and cartoon characters move about, and usually don't mind seeing other humans going through their daily motions, but when it comes to artificial creations that are made to look very human ... they're not always so popular. Although we tend to like animated objects or images that look kind of like real people, once they reach a certain level of realism, they just become spooky. This threshold is known as the "uncanny valley," and an international team of researchers recently set out to determine just what it is about our brains that causes it to occur.
The study was led by Ayse Pinar Saygin, an assistant professor of cognitive science at the University of California, San Diego. Twenty test subjects between the ages of 20 and 36 were used, none of whom had worked with robots, or spent time in Japan, where human-like robots are more common than in most countries.
These people were shown 12 videos of the Japanese Repliee Q2 android (pictured above), doing things such as waving, nodding, taking a drink of water and picking up a piece of paper. They also watched videos of the human that the robot was modeled after performing those actions, along with videos of the robot performing them without its skin covering, so it didn't look human.
When viewed through a functional magnetic resonance imaging (fMRI) scanner, the activity of the subjects' brains suggested that a conflict arose when they were viewing the footage of the lifelike version of the android. More specifically, activity was noted in their parietal cortex, which connects the part of the brain that processes body movements with a section of the motor cortex that is believed to help us relate to such movements. In short, Saygin believed that there was a disconcerting difference between the way in which the test subjects expected the android to move, and the way in which it did move.
"The brain doesn't seem tuned to care about either biological appearance or biological motion per se," she said. "What it seems to be doing is looking for its expectations to be met - for appearance and motion to be congruent." She added that perhaps androids and lifelike animated film characters should be run past human volunteers while still in development, to see what kind of response they elicit.
Although it would be impractical to routinely conduct such tests using volunteers in an expensive fMRI scanner, Saygin and her team are now looking for an electroencephalogram (EEG) signal that corresponds to the uncanny valley response, which could be detected using inexpensive EEG technology.