Why do ultra-realistic robots often make us feel uneasy, while teddy bears don’t? The answer lies in the concept of the “uncanny valley”.
Have you ever experienced a sort of unease that you can’t quite put your finger on in the presence of a doll or a mannequin in a store window, or after being shown a particularly realistic robot? Have you ever had the feeling that “something is wrong” without really knowing why? If you have, which is highly likely, particularly if you’re familiar with science-fiction movies and computer generated images, you have experienced what’s known as the “uncanny valley”.
In 1970, Japanese roboticist Masahiro Mori published an article in Energy magazine entitled “Uncanny Valley”. The name refers to the notion of uncanniness (a familiar and unsettling strangeness), which was first mentioned at the start of the 20th century by psychiatrist Ernst Jentsch, then by Freud, in reference to a story by Hoffmann about Olympia, a lifelike clockwork doll.
In his article, Mori presented a study he had conducted on the negative effect that the appearance of a robot can have on a person.
The study consisted of presenting a sample of people with different types of robots, from the articulated arms that weld together the bodywork of a car, to Bunraku dolls — gigantic puppets used in a type of Japanese theater, to ultra-realistic androids.
Quite logically, the empathy felt toward the industrial robots was equivalent to how you might feel about a toaster, in other words nothing at all. Still quite logically, a teddy bear brought out a bit more emotion, a cute robot even more so, etc. To put it simply, the more the android resembled a human, the more people empathized with it. On paper, this would assume the form of a nice ascending curve… up to the moment where participants were shown overly human robots. This almost human appearance, which had previously evoked empathy and acceptance, suddenly transformed into repulsion on reaching this higher level of realism.
Although it is still a theory at present, champions of the concept believe it is linked to facial recognition, an important ability in humans (and primates). It only takes a single glance (170 milliseconds) to know whether or not a person in your presence is known to you, whether or not that person is hostile, and even whether or not they are in good health.
When faced with a teddy bear, facial recognition isn’t taken into account as it’s an object. However, in the case of a hyper-realistic robot, it seems that the brain has difficulty handling the contradictory information of this strange face (friend or enemy, agitated or not, etc.). It is also interesting to note that at the bottom of this “valley” curve, something that is “too realistic” evokes the same negative feelings as a corpse.
Whether or not it is true, the theory of the “uncanny valley” has plenty more room for exploration; researchers are considering whether in the future people would experience the same level of unease in the presence of very advanced artificial intelligence.
Erica (Hiroshi Ishiguro Lab, Osaka University)