Corpses, Androids, and The Polar Express: A Social Neuroscience Perspective on the Uncanny Valley

In 1921, K. Capek introduced the world to the term “robot” in his fictional play in which android laborers eventually turn against their makers.  Mary Shelley’s Frankenstein and other literary works have addressed similar themes about tampering with nature and have exploited a common fear of humanoid creations (1).  By 1971, Masahiro Mori had coined the term “uncanny valley” to explain why humans have an innate fear of exceedingly humanlike robots (2).  The uncanny valley is a category for anything that appears very human, but isn’t exactly human enough to pass as a healthy human being.  This type of human imitation leaves viewers with a sense of eeriness, the defining feature of the uncanny valley (2).  Mori’s theory resonates well with issues that western literature had addressed years before.  But the question arises: Does his claim hold any scientific weight, and if so, why does the uncanny valley exist?

Mori depicts his theory with a graph of “familiarity” vs. “human likeness” (Figure 1) (2).  His graph reveals two extremes; one is a machine (at the origin), and another is a healthy human (at the right-most point).  An increase in the human-likeness leads to a corresponding increase in familiarity, or pleasantness (2).  For instance, Disney’s Wall-E, the robot with large, humanlike eyes, is more endearing than a machine.  As we continue moving to the right, the familiarity of the machine keeps increasing up to a point, where it drops abruptly.  This is the uncanny valley.  Objects in this valley resemble the human form to a high degree, but they don’t quite master all of a living being’s nuances, such as those of skin color and texture. They look quite real, and yet we can tell they are fake, which creates an unsettling response. Imagine a corpse or a zombie; both have features of real, animate humans, yet they are prime examples of the uncanny valley (2). Movement only accentuates the valley, since the jerky, uncomfortable movements of a classic robot are nowhere near the fluid nature of biological motion.  That is why the graph shows two different curves, one for still objects and another for moving objects (2).

Figure 1: Curves showing human emotion in response to human likeness. The uncanny valley is the dip where the likeness is too human for comfort.

Due to developments in modern technology, realistic imitations of human form are being created more than ever before.  Characters in video games and 3-D film animation are leading viewers closer to the realm of the uncanny.  Robert Zemeckis’ The Polar Express is an often-cited example of the uncanny valley, since his computer-generated characters are lifelike, but possess inanimate qualities that add a peculiarity to the screen (3,4). The film Final Fantasy possesses realistic, 3-D animated computer characters and was criticized for the same reason.  The animation director for Final Fantasy, Andy Jones, even said, “It can get eerie.  As you push further and further, it begins to get grotesque. You start to feel like you’re puppeteering a corpse” (5).  On the contrary, companies like Pixar have produced highly successful films that rely on stylized, non-realistic characters, as seen in A Bug’s Life, Toy Story, and The Incredibles (4,6). 

Although Mori’s theory about the uncanny valley has existed since the 1970s, few people have attempted to study it scientifically.  In recent years, the theory has sparked the interest of researchers in psychology and neuroscience, who want to know how this phenomenon manifests in the brain and why it occurs.  Thalia Wheatley, a psychology professor at Dartmouth College, is exploring the uncanny valley as an extension of her research on neural processing streams for animate and inanimate objects.  Wheatley reasons that neural mechanisms for detecting animate objects have been selected for, and honed, during human evolution because such capacities are crucial for recognizing potential predators, family members, or mates (7).  Through fMRI research, Wheatley and colleagues have implicated a group of brain regions, known as the social network, in animacy detection.  This system is activated instantly when a human confronts other animate beings (7). The uncanny valley exists on the border of the animate and inanimate; Wheatley predicts that the brain’s inability to concretely classify something as animate or inanimate creates conflicting messages in the frontal cortex, the decision center that allows us to make conscious and informed responses.  This conflict may result in a fear response, which could be the basis for the unsettling feeling that uncanny images elicit (8).

Why would we need to fear inanimate beings that fall in the uncanny valley?  For one, Mori defined corpses as the epitome of the valley. Fear of corpses could be advantageous to help us avoid disease, or other causes of death (2).  Karl MacDorman, who wrote one of the only review papers on the uncanny valley, explains that objects in the valley elicit fear because they can provide a conscious or subconscious reminder of death (3). Corpses are the quintessential example of this fear, but robots, and other objects that have corpse-like skin color, texture, and eye gaze speak to our fear of death as well.  Aside from fear of death, MacDorman cites other fear-based explanations of the uncanny valley.  For instance, many robots look quite human, but possess certain levels of disfigurement (3). This could serve to remind us about war and battle, another threat to our personal safety.  MacDorman also explains that the uncanny, inflexible movements of robots may elicit our underlying fears of losing control over our own bodies (3).In addition to fear, MacDorman addresses another plausible emotional cause for our response to the uncanny valley: disgust.  In 1987, Rozin et al. posited that humans developed a disgust response in order to avoid what could be an infectious disease (3,9).  A human with leprosy, or some other disfiguring and contractible illness, would fall into the uncanny valley, showing that the valley may represent an underlying reaction to disease (3).  This is not to say that the disgust theory is independent from the fear theory; both may serve to explain  our reaction to the uncanny valley.  

This 1960s android built by NASA to test spacesuits may fall into the “uncanny valley” of human likeness.

Autism is a disorder marked by an inability to relate to or understand people, and accompanied by impairments in language (11).  Research in the past few years has shown that social robots may help children overcome debilitations of the disorder.  Robots can be very predicable, a trait that is comforting to autistic children.  Additionally, these robots may be designed to possess social behaviors. Autistic children can imitate and react to these behaviors, thus acquiring social interaction skills that they need (11).  It is interesting to consider whether or not the uncanny valley would become an issue in designing robots for the treatment of autistic patients.  The neural basis of the uncanny valley most likely relates to our social and emotional processing (8), making it is possible that autistic people, who lack many healthy social processing skills, are less susceptible to the effects of the valley.  This remains to be explored scientifically, but it may strengthen evidence that autistic patients are particularly receptive to social robotic therapy. 

In the future, researchers may look towards patients with a rare disorder called Capgras syndrome to learn more about the uncanny response.  Patients with this syndrome possess a belief that their friends and family members are imposters (4).  For example, a specific case involves a man who experienced brain damage from a car accident.  Since the accident, he became unable to accept the fact that his wife truly was his wife (4,12).  Patients will admit that their friends and family members look exactly how they remember, but there is nevertheless something not quite right about them — something that appears unauthentic.  As expected, these feelings can be extremely unsettling and isolating to sufferers of Capgras syndrome.  A few patients have even explained that former loved ones seem like robots, a fact that has linked the syndrome well to the uncanny valley (4).

We can begin to understand Capgras syndrome by first considering the fact that the brain reacts differently to an image of a familiar face than to a novel one (13).  When healthy individuals observe images of familiar faces, their brains establish a facial recognition response (within the temporal lobe), and also an emotional response (mediated by the limbic system) (13).  This emotional response creates familiar associations with the individual in the image.  Capgras syndrome may be explained by a disconnect between these two brain responses (13).  Patients recognize the familiar people around them, but this recognition is accompanied with the emotional response equivalent to viewing a stranger.  Thus, patients are lead to believe that their friends and family members are imposters (13).  Research on Capgras syndrome emphasizes an important link between the emotional and recognition areas of the brain.  Perhaps uncanny images produce a Capgras-like phenomenon in healthy individuals; the familiarity in human form that we witness may be undermined by our subliminal emotional response, creating disconnection between recognition and emotion (4). 

Although there is much to be researched on the topic of the uncanny valley, the evidence to date supports the existence of Mori’s theory.  By discovering more about the phenomenon, researchers can learn how to approach human realism without entering the uncanny valley, and thus expand the fields of robotics, film studies, and video game design.  Further, improvement in these technologies may increase the popularity of new treatments in social disorders.  The uncanny valley should not limit new technology, but instead pose a challenge to developers, inspiring increased determination towards surmounting obstacles in human perception.

References

1. T. Shibata. Proceedings of the IEEE 92, 1749-1758 (2004).
2. M.B. Mori. Energy 7, 33-35 (1970).
3. K.F. MacDorman, H. Ishiguro. Interaction Studies 7, 297-337 (2006).
4. F.E. Pollick. To Appear In: Grammer K, Juette A, editors. The Vienna series in theoretical biology; Analog communication: Evolution, brain, mechanisms, dynamics, simulation. Cambridge: MIT Press.
5. L. Weschler. Why is this man smiling? Digital animators are closing in on the complex system that makes a face come alive. Wired. (October 6, 2002). Available at http://www.wired.com/wired/archive/10.06/face.html-[quote: pg. 4].
6. C. Bartneck, T. Kanda, H. Ishiguro, N. Hagita. 16th IEEE International Conference on Robot & Human Interaction Communication. Jeju, Korea. 368-373 (2007).
7. T. Wheatley, S.C. Milleville, A. Martin. Psych Sci. 18, 469-474 (2007).
8. T. Wheatley. Personal Communication. 5 December 2008.
9. P. Rozin, A.E. Fallon. Psych Rev. 94, 23-41 (1987).
10. E. Klinger, S. Bouchard, P. Legeron, S. Roy, F. Lauer. I. Chemin, P. Nugues. Cyber Psych & Behav. 8, 76-88 (2005).
11. K. Dautenhahn. Robotica 21, 443-452 (2003).
12. H. Ellis, M. Lewis. Trends in Cog. Sci. 5, 149-156 (2001)
13. W. Hirstein, V.S. Ramachandran. Proc. R. Soc. Lond. 264, 437-444 (1997).