A Drive Through Uncanny Valley

In the early 1970s, Japanese researcher Masahiro Mori noticed something interesting. As someone who had spent most of his professional life working with robots, Mori noticed that the more lifelike a robot was, the more people liked it… but only to a certain point. Once the robot became too lifelike, people were repulsed by it. Some people felt physically sick when looking at extremely lifelike robots, while others ran away in fear. And while those were somewhat extreme responses, nearly everyone reported some measure of uneasiness or distress when viewing an ultra-lifelike robot. Dr. Mori, it seems, had unwittingly discovered “Uncanny Valley”.

Back when robots looked more like a pile of car parts than a person – think of the robot from Lost In Space – people looked at them as simple machines, no different, really, than a drill or garbage disposal. Which wasn’t a bad thing, mind you, but it didn’t engender any feelings of warmth for the robots, either. People felt the same way about video games and displays of “virtual people” (like, say a “virtual teller” at an ATM); as long as the representations of people were crude, people had no problem with it.

But technology marches on, and soon video games had people that looked incredibly lifelike. Japanese researchers developed robots that were almost indistinguishable from real people. Hollywood found that movies could be made solely with computer-generated people. And Internet companies developed “virtual people” that could work the customer service desk or be “online ambassadors” for companies. And all of these creeped people out. The question was… why?

It all comes back to Dr. Mori’s work. He hypothesized that once you cross a certain threshold, people stop seeing robots (or video game characters or “virtual people”) as artificial constructs and start seeing them as something else. Although the Japanese have built some incredibly lifelike robots, they all have the same lifeless eyes. Their “facial muscles” are quite limited and the lack of movement in the face when the robots “talk” appears as odd and disturbing to many people. Many are convinced that the human mind interprets a highly lifelike robot as a corpse, with all the fear and trepidation it brings with it.

There is another theory that guesses that it has something to do with evolution. Humans have a tendency to graft human qualities to non-human objects. We look at things like cartoons and recognize them as non-human, but once we add some human qualities to it, we feel empathy for the thing. This would explain the continuing popularity of animals in animated movies: the brain knows that Donkey from Shrek isn’t human, but the fact that he dances and jokes delights us, and we connect with him. The Uncanny Valley hypothesis flips this on its head: we see an object like a robot and perceive it as human, but then the lack of eye movement and facial gestures stirs something deep within our evolutionary beings, perhaps even the “fight or flight” response.

So: why does any of this even matter? Well, in the long term it certainly has an enormous impact on the future of robotic development. But even more pressing is the entertainment industry’s need to understand Uncanny Valley. Many in the video game and movie industry saw the example of the 2001 film Final Fantasy: The Spirits Within. The movie was completely computer generated. The characters in the film were astonishingly lifelike… and filmgoers noticed. The movie jumped straight off the cliff and into Uncanny Valley; reviews on the actual movie itself seem mixed, but many reviewers complained about the computer-generated characters and how disturbing and creepy they were. The film was such a flop that it put its makers into bankruptcy. Although there could have been a million reasons why the movie failed, the characters lifelessness was absolutely near the top.

Since then, most movie and video game makers have tried two approaches. The first is to keep characters away from a “lifelike” human form (the humans in Ratatouille, for instance, look like cartoons). The second is to try and “jump” Uncanny Valley by combining live action with computer graphics (Gollum in the Lord of the Rings films, the recent Beowulf movie, for instance). Both Gollum and Beowulf were well-received by audiences (at least as far as Uncanny Valley goes), so this approach might be the wave of the future.

I personally hope so. I own an Xbox game based on the Dark Angel TV show, and the first time I saw the computer-generated Jessica Alba on my screen the first thing I felt was a rumbling in my stomach and a tinge of fear… something just didn’t “look right”.

Read more about Uncanny Valley here or here.

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.