I’m not sure about you, but when I used to watch my brother play Grand Theft Auto, the character faces always looked creepy to me. I’m aware that they’re not meant to look beautiful like Final Fantasy characters, but in terms of realism the character faces lacked-something…was it the lifeless eyes or the strange movements of the mouth? Perhaps a more relatable example is looking at those advanced Japanese robots that are meant to look like a young lady, or the poorly done Tyra Banks wax model at Madame Tussaud’s. What is it that makes looking at these imitations of humans so uncomfortable?
Japanese researcher Masahiro Mori described this phenomenon in terms of the uncanny valley in 1970. The closer something looks to a human, the higher is our affinity towards it. However, there is a point where our affinity for the object drops dramatically in this relationship, and this ‘drop’ is what Mori labelled the ‘uncanny valley’. Masahiro Mori depicted this relationship in a graph which can be found in Figure 1. For example, adding a face to an apple can make it seem cute and more likeable. However, the closer the object’s likeliness to an actual human, such as the Tyra Banks wax model- the more uncomfortable it becomes to look at it.
Figure 1. Showing the relationship between the human likeness (appearance) of an object and our affinity for the object. Obtained from Mori and Kageki (2012).
Mori (2012) also proposed that movement can exaggerate the peaks on the human likeness-affinity model. Figure 2 shows this relationship, it is essentially the same graph as Figure 1 but the peaks are extended, meaning that with movement, an object showing less human likeness is liked much more than if it did not move, but an object with a higher human likeness is disliked even more than if it did not move. For example, Heider and Simmel (1994) conducted an experiment where participants were shown a cartoon of two triangles and a circle moving around. When asked what was happening in the cartoon, most participants described the events in terms of the objects interacting in human-like ways, stating that one was bullying another or they had intent to act, even though, objectively it was nothing more than 3 objects moving on a screen. This ‘tendency to imbue the real or imagined behaviour of nonhuman agents with humanlike characteristics, motivations, intentions, or emotions’ is known as anthropomorphism (Epley, Waytz & Cacioppo, 2007). This video is available online so have a look and ask yourself what you see! However, when we see the advanced Japanese robot sing or dance, the feeling is once again discomforting.
Figure 2. Showing the effect of movement on the relationship between the human likeness (appearance) of an object and our affinity for the object. Obtained from Mori and Kageki (2012).
On the other hand, Bartneck, Kanda, Ishiguro and Hagita (2007) criticise the uncanny valley and state that cultural differences will vary how human-like objects are perceived in terms of the uncanny valley. Hanson et al (2005) also argue that uncanny can arise at any point on the scale of human likeness, for example MIT’s robot Lazlo and even for humans who have had multiple cosmetic alterations on their faces. The uncanny valley can also be difficult to test experimentally. Hanson (2005) attempted this by morphing a photo of an android and a photo of a human man to different extents (on a scale of android to human) and asked participants to rate the images on an uncanny scale; Hanson (2005) failed to find the relationship predicted by the uncanny valley.
Whatever the research shows, this area has important implications in terms of what will become of human-robot interactions and in computer graphics animation. Mori (2012) proposed that to make the experience more pleasant with androids, it would be wise to design them to be as ‘unhuman’ as possible. However, according to the criticisms to the uncanny valley theory, it appears that the answer is not as simple.
Bartneck, C., Kanda, T., Ishiguro, H., & Hagita, N. (2007). Is the uncanny valley an uncanny cliff?. Robot and Human interactive Communication, 2007. RO-MAN 2007. The 16th IEEE International Symposium (pp. 368-373).
Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: a three-factor theory of anthropomorphism. Psychological review, 114(4), 864.
Hanson. (2005) Exploring the aesthetic range for humanoid robots. ICCS/CogSci-2006 long symposium: Toward social mechanisms of android science, p. 16–20.
Hanson, D., Olney, A., Prilliman, S., Mathews, E., Zielke, M., Hammons, D., & Stephanou, H. (2005). Upending the uncanny valley. Proceedings of the national conference on artificial intelligence Vol. 20, No. 4, p. 1728.
Heider, F., & Simmel, M. (1944). An experimental study of apparent behavior. The American Journal of Psychology, 243-259.
Mori, M., MacDorman, K. F., & Kageki, N. (2012). The uncanny valley [from the field]. Robotics & Automation Magazine, IEEE, 19(2), 98-100.
Image credit: http://www.buzzfeed.com/danmeth/welcome-to-uncanny-valley#.rbn4NBxQl