Skip to content

Good Bot, Bad Bot – Robots and the Uncanny Valley

Do you remember the film character WALL-E from the 2008 Pixar movie? With his binocular-inspired eyes, his gestures and movements, it feels like the trash compactor robot has come to live, has feelings and emotions even though his design remains mechanic. Viewers around the world were excited and touched.

Link: https://www.youtube.com/watch?v=ZisWjdjs-gM

In contrast to adorable robots like WALL-E, Dr. Hiroshi Ishiguro constantly shocks the world with his humanoid robots, with one of them looking like the scientist himself leaving us behind with a creepy feeling.

Link: https://www.youtube.com/watch?v=SsYLOxYmFbk

But why are we scared or feel uncomfortable watching Ishiguro’s creations? This phenomenon might be explained by the so-called uncanny valley hypothesis.

“The Uncanny Valley was my intuition

Masahiro Mori in an 2012 interview by Norri Kageki

Almost 50 years ago in 1970, Japanese scientist Masahiro Mori coined the term uncanny valley. He described the hypothetical relationship between human acceptance of robots and the degree of perceived humanness of the robot. Find a 2012 authorized and reviewed English publication of his original article here.

To be more precise, Mori argued that there is no linear relationship between affinity and anthropomorphism. Instead, there is a dip in the curve diagram – the uncanny valley – when a robot resembles human but not quite good enough, the acceptance by humans goes down and eeriness occurs. To sum up, Mori states that human reactions to robots are more positive when they have either no to little (graph before the valley) or perfect human likeliness (graph beyond  the valley).

From the 2012 official translation of Mori’s original article. Find publication here.

According to Mori, the presence of movement intensifies the relationship – it’s peaks and dips. As a result, the uncanny valley gets bigger. To illustrate this, he compares a prosthetic hand with a moving myoelectric hand with the latter giving the viewer of its movement an even stronger feeling of unease.

In a 2012 interview Mori explains that when he published his ideas on the uncanny valley in the 70ies, there was no response at all. With an increasing number of robots that look like human, people started to dig deeper into the phenomenon loosely described by Mori.

“Pointing out the existence of the uncanny valley was more a piece of advice from me to people who design robots rather than a scientific statement.”

Masahiro Mori in an 2012 interview by Norri Kageki

Originally, the theory was developed to explain human reaction to robots but empirical research extended the concept of uncanny valley to digital animations such as digital human faces, voices and more.

In 2006, MacDorman and Ishiguro were the first to present empirical evidence for the existence of a relationship of (negative) affect and human (non-)likeliness. The scientists used morphing software to alter images of humans and robots to create image material representing 12 levels of human likeliness for being subjectively rated. Still, the empirical findings are inconsistent and other paradigms didn’t find evidence for the uncanny valley theory.

Next to that, it’s also unclear how the non-linear relationship between affect and humanness can be explained (if it exists). For Mori, the feeling of eeriness is part of self-preservation. Not perfectly human-like robots with mechanic movement or unnormal blinking could remember us of illness and therefore as a potential threat to our health. Other explanations suggest that the negative affect arises from ambiguity of the stimuli. Material that can’t be easily categorized into “human” or “non-human” is perceived as ambiguous and may therefore elicit negative feelings of unease or discomfort. Evidence for the categorization explanation comes from scientists of the University of Zurich using functional magnetic resonance imaging to investigate brain representation of the perception of humanness.

Even though there is no consistent empirical support for the uncanny valley theory, designers follow Mori’s advice on staying on the first peak of the graph (see WALL-E example) from time to time. In contrast, the makers of the movie “Polar Express” used motion capture technology trying to overcome the uncanny valley and ended-up with devastating critiques:

“But those human characters in the film come across as downright … well, creepy. So “The Polar Express” is at best disconcerting, and at worst, a wee bit horrifying.

– by Paul Clinton for CNN