A Walk Through The Uncanny Valley
edited: Tuesday, December 09, 2003
By Bret M Funk
Posted: Tuesday, December 09, 2003
Become a Fan
An exploration of 'The Uncanny Valley', a theory proposed by cyberneticist Mashahiro Mori, and its effects on both modern science and science fiction.
First published in The Illuminata in September 2003.
A Stroll Through the Uncanny Valley
By Bret Funk
While the building of machines to mimic human action is far from novel, the creation of machines to mimic humans is a more recent concept, one only just beginning to escape the realm of science fiction. The first few years of the 21st century have seen robots walking bipedal, navigating unaided through obstacle-filled environments, recognizing individuals and carrying on limited forms of communication with them. These robots—ungainly, lumbering monsters; bug-eyed metal infants; and the occasional talking box—while human only in the most liberal sense of the word, inspire not only wonder, but a certain amount of empathy among the ‘real’ people who see them.
For robots to make the leap to human, though, synthetic polymers (preferably ones that feel, look, and sweat like skin) need to replace metal shells, and that synthetic skin (pulled by servos instead of muscles) must react and emote the way ours does; eyes need to replace sensors (or sensors need to look like eyes); movements need to become more fluid, less cautious; and responses to stimuli must appear natural and not preprogrammed. In essence, to become human, robots must stop being robots. They must become androids.
Though considered an unrealistic dream by many, the roots of the technologies necessary to create androids exist today. A number of scientists have made leaps in robotics, creating human-like motion, faces capable of displaying a phenomenal array of emotion, and most importantly, machines that can evoke feelings. There can be no doubt; somewhere down the long road of scientific discovery the certainty of androids exists.
The question then becomes not ‘Can we build an android?’ but ‘Should we build one?’ for it seems the greatest threat to artificial humans is not the destination, but the journey. In the mid-70s, an enterprising scientist named Mashahiro Mori developed a rather disturbing theory. Mori is, for want of a better term, a socio-cyberneticist, and his speculation into human-android relations has resulted in a number of intriguing experiments and at least one book (The Buddha in the Robot).
In one study, Mori plotted robotic designs against human emotional response. As one might expect, robots that bear only the faintest resemblance to humans receive almost no emotional reaction, and that as robots become more similar to humans, the positive emotional response and empathy they receive increases. However, somewhere after 90% human, the curve makes a sudden and dramatic shift downward, far past the level of neutrality, earning those near-humans a strongly negative response. As androids continue toward 100%, the curve rebounds, signifying the creation of a machine so life-like it is indistinguishable from a real person. This trend holds true on more than the macro level; it is repeated for a number of more specific variables as well (appearance and movement, for example).
Mori dubbed the unexpected dip ‘The Uncanny Valley’: the area where an observer would see something almost human, yet at the same time—and perhaps even on an unconscious level—fundamentally wrong. These creatures are monsters, shunned by society because they are simultaneously too human and not human enough. They will be (and are) ostracized, ridiculed and disdained, considered less than human (even when they are human); and in some scenarios, hunted to extinction.
Since technological progression occurs in steps and not leaps, it will not be possible to create a perfect android without walking through the Uncanny Valley. At 93-98% human, these creatures will be no less deserving of love, no less capable of feeling, and no less entitled to rights and privileges than their next-generation cousins. Knowing that they will be repulsive to most of mankind, how can we in good conscience consider creating them?
Some critics have claimed that Mori’s hypothesis is flawed, that, in a time which had few robots (and even fewer that appeared at all human) he could not possibly have accurately assessed human emotional response to near-humans. I, on the other hand, contend that evidence exists today, in both the real world and in literature, to support Mori’s theory.
In recent years, a number of semi-human robots have hit the mainstream. Honda’s Asimo (http://world.honda.com/ASIMO), the robot-in-a-spacesuit, entered our world to a round of applause. Though bipedal, human-shaped, and capable of not only recognizing human positions and reacting to them but also of navigating highly-difficult terrain (like stairs), Asimo remains far to the left of the Uncanny Valley. His younger cousin, MIT’s Kismet (http://www.ai.mit.edu), has rivaled Asimo is praise but far exceeded him in cooing. This disembodied robotic head, with its big bug-eyes, bright red lips, and approximations of human expressions, is one part infant, one part Gizmo (of the mowgli/gremlin variety). Still to the left of the Uncanny Valley, one nevertheless sees an increased emotional response toward Kismet, evidence that it is easier to anthropomorphize robots that appear more human.
The Uncanny Valley concerns more than just appearance (much to the relief of mannequins everywhere) and GRACE (http://www.palantir.swarthmore.edu/GRACE), spear-headed by Carnegie Melon University, demonstrates that how a robot acts is just as important as how she looks. In 2002, Grace was left alone in the Shaw Conference Center with a daunting task: navigate through the maze-like building, register to speak on the subject of robotics, and then give a fifteen minute presentation on her hardware and capabilities. After asking for directions, Grace went to the registration area where she waited patiently in line (after pushing a much-surprised Leslie Kaelbling out of her way), registered for her speech (and demanded one of the conference’s fancy give-away bags), and then made her way through the crowd to deliver her presentation to an impressed audience. Though hardly human in appearance, Grace was almost treated as a person by those fortunate enough to encounter her.
And then we come to Dr. David Hanson of the University of Texas, Dallas (http://iiae.utdallas.edu/projects), a man who modeled a robotic head after his girlfriend’s (actually, she only became his girlfriend after he asked permission to make her into a robot). After extensive study of the human head, Hanson created a series of robotic ones, some with over thirty motors to mimic muscle action. His latest, the one that looks like his girlfriend, is almost uncanny in its appearance, and in a recent Popular Science article, the expressions made by both girlfriend and robofriend are virtually indistinguishable. Yet Hanson’s research, though garnering much kudos from fellow roboticists and fans of Popular Science, has been met with a degree of uncertainty. Some feel that his faces are too lifelike, that his robotic heads have about them an eerie, disquieting quality. This reaction leads one to wonder if Hanson’s creations are skirting the edge of the Uncanny Valley.
Support for Mori’s theory is also apparent in science fiction, and since SF is created by humans, it should be indicative of their beliefs. Robots that are vaguely human in form and/or attitude are often loved. R2-D2 and C-3PO, Robby the Robot, and Twiggy of Buck Rogers stand out among a number of semi-human robots who have captured the hearts of humanity. When one hears R2’s sad whine or sees Robby frantically running across some planetscape, waving his arms and shouting ‘Danger, Will Robinson, Danger!’, one cannot help but feel that they are a character—a being!—as deserving of love, sympathy and respect as any human (and more deserving than some).
This trend continues, culminating perhaps in the character of Lt. Commander Data, an android so close to human that he stands on the cusp of the Uncanny Valley. Slightly odd in appearance, mannerism and action, he is not quite perfect, but close enough that most in the Star Trek universe treat him no differently than anyone else. Still, there are a few who fear him, eye him askance, or treat him differently because, in their minds, he is a thing pretending to be human. This deception makes him unlovable, and his unwitting assertion that he is just as good as a real human being jabs at both his opponents’ pride and their understanding of humanity.
Androids closer to human than Data are firmly entrenched in the Uncanny Valley. The Terminators, Blade Runner’s replicants, and the androids in the Alien movies are prime examples of machines so close to man they are all but indistinguishable on the surface, yet treated fundamentally differently – feared, hunted and exterminated, or treated as a sub-class of humanity – because they are artificial. One might argue that in the first two cases, humanity was merely responding to a threat, but it is important to note that in both cases, the threat only came about because humanity feared what it had created. Past these are the robots from movies like Android and Cherry 2000, machines human in appearance and action, but disliked (or liked) more because of their human qualities than because they are machines. They stand to the right of the Uncanny Valley, on the slope toward perfection.
Yet no film more clearly shows Mori’s Uncanny Valley than A.I. In it, the young android David, the first android with real emotions, makes a suitable replacement for a grieving family until their real son returns. Suddenly, David is an outcast, forced to enter the world of the mechas, machines close but not quite human, who serve humanity as slaves. These machines stand at the base of the Valley, and the depressing tones that permeate the movie only emphasize the truth of Mori’s idea.
Perhaps most interesting is the fact that Mori’s theory can be applied outside of robotics, to other branches of speculative fiction and to the real world as well. The Uncanny Valley may well explain humanity’s fascination with the undead – zombies in particular. Those shambling, almost human and often-carnivorous automatons strike a nerve in us; we see in them creatures both fundamentally human and fundamentally wrong. It explains why we fear ghosts and vampires but are repulsed by zombies and their ilk.
Creatures like Frankenstein’s monster, Dr. Moreau’s half human, half animal creations, and the alluring alien from the Species movies are no more evil (and often less so) than the humans who made them, but they are the ones feared, driven from society, and exterminated, all because they stand too deep in the Valley. Mori’s theory also explains the derision and fear associated (within the world of the comics) with heroes like Spider-man, the Hulk, and the Uncanny X-men (emphasis mine).
However, the most convincing proof of Mori’s concept may, if fact, reside in the real world. Humanity’s near-paranoia regarding cloning and genetic manipulation (and other, completely innocuous but often misunderstood things like irradiation) are likely due to the Uncanny Valley. In fearing these things, humanity shows that it fears the creation of something quasi-human. With every taunted cripple, with every homophobic outburst, with every man, woman and child who cringes from the scarred visages of the afflicted or eyes an autistic child with a mixture of pity and revulsion, society shows its disgust for anything that stands just outside of what it considers ‘normal’. For confirmation of the Uncanny Valley, one need look no further than John Merrick, the Elephant Man. Ostracized from society, Merrick was treated like a sideshow freak. He was man – a real, living human! – treated like a monster; and yet this monster was a being of remarkable intelligence, compassion, and profound understanding. Near the time of his death, he wrote:
"Tis true my form is something odd,
But blaming me is blaming God;
Could I create myself anew
I would not fail in pleasing you.
"If I could reach from pole to pole
Or grasp the ocean with a span,
I would be measured by the soul;
The mind's the standard of the man.”
(The True History of the Elephant Man, 1980)
In the end, humanity fears the Uncanny Valley because the creatures who stand in it wear on their surface what the rest of us hide inside. We see in them something not quite right, not quite human; and at the same time, we see in them aspects of ourselves, and the realization that we are not so different intensifies our fears and focuses our anger. This fact is as true for androids as it is for mutants, but androids face a far more profound problem, one that may leave all artificial life stranded in the farthest depths of the Uncanny Valley:
By creating an android that is also human, one with intelligence and emotion, we will have proven ourselves neither special nor unique. One can only speculate how humanity will take such a revelation, but the odds are against us taking it well.