The 'uncanny valley' is real, and science can prove it

We now know that it is possible for humans to feel empathy for robots -- but what about more negative emotions?

The long-standing concept of the 'Uncanny Valley' describes the feeling of discomfort humans feel when we see something that looks or moves almost -- but not exactly -- like a living being. And whether it's an animation, video game character or a robot, it turns out humans are very easy to creep out.

The theory was first described by robotics professor Masahiro Mori to explain emotional responses to robots. He found that responses to increasingly lifelike robots became increasingly positive until a specific point was reached -- the beginning of the dip into 'Uncanny Valley' -- at which point it decreased. And then, as robots appear less and less human, human empathy towards them increases again.

At least, that was the supposition. But now psychologists at the University of California at San Francisco have found evidence that the valley is real.

The robot faces used in the experiment, ranked by how lifelike they areUniversity of California, San Francisco

Researchers took pictures of 80 robot faces and asked participants to rate them on a scale of 1 to 100, depending on how mechanical or human they looked. They then had to rate the robots by how pleasurable or enjoyable it would be to interact with them every day. The faces were then ranked by 'perceived friendliness' -- and researchers found that as they shift from mechanical to lifelike, their friendliness rose, then fell, then rose again.

The aim was to study the reactions to everything from a mechanical arm to a fully realistic android, in order to clarify the arc of the uncanny valley and help to design more likeable robots in future.

The robot faces used in the experiment, ranked by how lifelike they areUniversity of California, San Francisco

"Robots are transitioning from something that’s part of a technological environment to something that’s a feature of our social environment," said Maya Mathur, co-author of the study. "They're always teetering on this boundary of being really creepy and really likeable. That’s something we need to understand."

In another experiment, researchers asked participants to play a game of 'Trust' with the robot faces. Participants were given $100 and asked to share this with a robot to 'invest'. After this investment, the robot would decide how much to give back to the participant. Again, the amount of money that participants were willing to invest in the robot followed a similar 'valley'-style pattern. Researchers found a recognisable sharp dip and rise for studies involving both tangible robot faces and digitally composed robot faces.

"The research indicates that Uncanny Valley is a real influence on humans' perceptions of robots," said Mathur. "It robustly influences not only our conscious assessment of our own reactions, but also our ability to modify our social behaviour with robots".