Is the Uncanny Valley an Urban Legend?
In 1970 Masahiro Mori, a robotocist publishing in the journal Energy, put forward the theory of the Uncanny Valley – that the closer non-human things get to being like a human the more icked out we are by the experience. Its become an urban legend and until the Eric Sofge article at Popular Mechanics titled The Truth About Robots and the Uncanny Valley: Analysis I would have sworn this was a fact.
On a superficial level, of course, the Uncanny Valley makes sense to us – we’re pretty icked by icky robotic nymph-maids, a little too sexy for a robot but a lot too robotic for a human. But Masahiro never pushed this forward into an actual theory, it was and is only a conceptual idea. More interesting still, the Sofge article mentions Karl MacDorman, director of the Android Science Center at Indiana University, who performed research that proved differences between gender perception by humans of robot patients. It appears the depth of the relationship within robot-human interaction is more complex than the superficial truths that surround the Uncanny Valley.
The real cruncher is that all of the roboticists and computer scientists interviewed for the Sofge article asserted that even when the Uncanny Valley is apparent at a distance or on a computer monitor it simply disappears on human-to-robot contact. Proximity trumps the Uncanny Valley. In fact, the Uncanny Valley had never been proven and had become an urban legend to which my own measure of perpetuated inaccuracy has contributed momentum.
However, from the world of Masahiro Mori in 1970 to the present day digital reality, it is worth taking into consideration that exposure to robots, computers, software, science fiction and the exponential increase in the availability of knowledge to the average Joe has desensitised us to the WOW factor of what may have originally been a much more powerful and significant Uncanny Valley. We hardly even doubt the reality of sexy maid-like robots with questionable marketing possibilities.
In an article by Anil Ananthaswamy in New Scientist titled Empathy with Robots Depends on Exposure, Anil writes about Sotaro Shimada of Meiji University in Tokyo who tested the Mirror Neuron System (MNS), thought to be related to empathy, by exposing test subjects to robots with both jerky and smooth movements. Using near-infrared spectroscopy to assess oxygenation within critical parts of the human brain, the results showed that MNS, and hence empathy, was activated only when the robots moved robotically.
In fact, only when humans and robots meet the expectations of associated movement (humans smoothly and robots jerkily) does MNS, and hence empathy, respond. Shimada theorises that exposure is a major factor in this triggering of MNS in human beings.
While the world of Masahiro Mori in 1970 suggested the Uncanny Valley might exist its a little unfortunate nobody in that time and worldview stood his theory to the test. Did the Uncanny Valley exist in 1970 and fade back through the new millenium? Its plausible, at least. In 2010 the Uncanny Valley seems to be just another urban legend.
January 27th, 2010 at 7:20 pm
You seem to contradict yourself. On the one hand, you say the uncanny valley is just an urban legend and, on the other hand, you say that we have become desensitized to what may have been a much more powerful and significant phenomenon. To admit that the uncanny valley is a less powerful and significant phenomenon than it once was is far from denying it by labeling it an urban legend.
There is no doubt that the term “uncanny valley” is used rather loosely at times to describe virtually anything about a robot or animation that looks odd, even if the entity is not really eerie or uncanny. But that kind of attribution or shared report is far from being an urban legend. An urban legend is a well-known story that people accept as fact when it is in fact fiction. When people watch Polar Express and find something disturbing about the wooden movements of the characters, that is not a story they heard about. Rather that is their personal and undeniable part of their life experience.
The fact that some researchers have become habituated to their freaky robots and their participants are too polite to criticize them hardly invalidates the empirical evidence that there is in favor of the uncanny valley. Instead of being too impressed by these anecdotes and the musings of a Popular Mechanics journalist, you might use Google Scholar to find the empirical, peer-reviewed studies that have been published on the matter. You are correct to say that the phenomenon is far more complicated than Mori’s article would imply.
You also say that the uncanny valley may have existed in 1970 but not now. In other words, there is a cultural component. Well, that hypothesis has already been tested in studies that used Indonesian participants who had never been exposed to robots before. We can roll back time in a sense by doing studies with people who have no expose to robots or animation.
January 28th, 2010 at 7:56 am
Lee, I meant the article more as a question rather than an assertion and expect that’s where my contradictions have become apparent… and thanks for the heads up on the Indonesian study. The simple fact is that like most people in the world I don’t know… about most stuff. Unfortunately as a blogger, rather than a paid and published journalist, I am less an authority on anything and more of an opinion on everything so rely on exactly this kind of feedback to help me out with my pursuit of knowledge.
Had I been a little better as a writer in this article it probably should have read more that “it is asserted the uncanny valley does not exist… there is some research and speculation based around certain articles… however (and this is where the contradiction appears) in my opinion we’ve become desensitised… the second linked article supports the theory that as we’ve become desensitized to 21st century life the effect has diminished (notably Japanese were in one study linked to)… and that is the limit of my knowledge because I’m not a roboticist or a computer scientist”.
I’ll have to hunt up the Indonesian study… you wouldn’t have a link handy? Its actually an interesting subject because I’ve never for a second questioned the validity of the valley in my own mind until reading that first article… thus the article I published here.
January 29th, 2010 at 8:27 am
The more I think about the Indonesian study to answer this question of the way-back machine scenario, the less it really answers the questions. Not only is the effect of exposure relevant to the culture – ie. testing to see how pre-exposed culture treats the Uncanny Valley… – but also that culture’s ideas about who the person is, what makes a soul, etcetera. In other words, this would not tell us what an Anglo perceived or a Japanese perceived in 1970. So I’m not really convinced the study you mentioned from Google Scholar can answer the question using Indonesian un-exposed people today about whether the 1970 Anglo-US-Japanese Uncanny Valley experience existed or not, to what extent or not. Its very much apples and oranges.
So I’d have to stand by my last paragraph in the article, Lee. There are just too many cultural variables to transpose that over as a way-back machine onto Europe, the US and Japan in 1970.
Gee, Browne and Kawamura put forward a brief comment about the possible influence of different cultures on page 5 of Uncanny Valley Revisited.
By the way, I have looked and cannot easily locate the study you mentioned of Indonesian participants without previous technology exposure. Could you possibly put a link into the comments so that I can have a read of it? Sounds interesting.
Gee, F. C., Browne, W.N & Kawamura, K. 2005. Uncanny Valley Revisited, 2005 IEEE International Workshop on Robots and Human Computer Interaction. Viewed 29 January, 2010.
February 11th, 2010 at 1:42 am
When I googled uncanny valley Indonesia, I just saw some news articles, but from them I could find the paper:
http://www.macdorman.com/kfm/writings/pubs/Ho2007EmotionUncanny.pdf
Ho, C.-C., MacDorman, K., & Pramono, Z. A. D. (2008). Human emotion and the uncanny valley: A GLM, MDS, and ISOMAP analysis of robot video ratings. Proceedings of the Third ACM/IEEE International Conference on Human-Robot Interaction (pp. 169-176). March 11-14. Amsterdam.
You’re right that it’s not a way back machine for Australian or other Western cultures, but probably Indonesians who have not had robot exposure would still experience the uncanny valley more strongly than those of us who are bombarded with computer generated human characters or android robots.
February 11th, 2010 at 8:02 am
Thanks Lee, another interesting article.
In Methods 2.1 Participants it reads:
So I’m not sure that their lack of exposure to robots in practice is not affected by their actual exposure to the ideas and concepts of robots from the Internet, movies and other media. Also, different cultures have inherent differences in the way they view self (ie. some cultures traditionally believe a photograph steals the soul).
Similarly in 2.2 Materials and Procedures the paper states that:
It would be interesting to see how the perception of the uncanny valley either vanishes on real contact with the robot, as some research suggests, or the valley remains in the Indonesian cultural context, too.
Mmm its an interesting paper. But as both a Bachelor of Computing graduate and a Master of Business Administration post-graduate student I would still probably say your original final comment statement was a little strong:
(a) the Indonesian participants did have exposure to the concept of robots – ie. university students, professionals, etc – as they were recruited from Internet cafes
(b) the Indonesian participants were affected by Indonesian (probably Muslim) concepts of what constitutes actual life, the soul, the self
(c) the cultural component hypothesis I mentioned wasn’t adequately tested in this paper, only for Indonesians in the educated sectors of their society.
It would however show a potential difference between Indonesians with univerisity degrees who worked with robots versus these Indonesians who did not work with robots, if that makes sense. I would also like to see Indonesians who have no exposure to the Internet, from poverty backgrounds, non-educated and test the valley concepts on them. Similarly, Australian desert aboriginals may have unique perceptions worth testing.
But you’re right, the article I wrote was really asking a question as to whether the valley really exists or not – it had faults in the way I put those concepts across to the reader.
Another avenue that would be interesting would be to bring a neurologist / primatologist like Robert Sapolsky into the investigation. What is the brain doing? Rather than subjective Likert Scale observations it would be nice to see the “mechanical differences” in the way our chemicals change.
Apologies for the long response in this comment but it is an interesting subject. Its one that through my computer degree I didn’t question – the uncanny valley exists. So this was the first time I’d actually questioned it in my own head.
Interestingly, we do have inmates in prisons who still aren’t exposed to a great degree, uneducated etc who we could possibly attempt a western culture way-back machine type of experiment with… if only it were ethical.
Thanks again for the paper, I really appreciate it, Lee.
February 11th, 2010 at 8:20 am
I like these two sentences in Conclusions on page 175: