Navigating the Uncanny Valley with Character Robots

Daniel Faggella

Daniel Faggella is Head of Research at Emerj. Called upon by the United Nations, World Bank, INTERPOL, and leading enterprises, Daniel is a globally sought-after expert on the competitive strategy implications of AI for business and government leaders.

Navigating the Uncanny Valley with Character Robots

Episode SummaryHave you ever seen a humanoid robot and felt creeped out? In this episode, we talk with Robotics Engineer Derek Scherer about the psychology of the uncanny valley and our relationships with robots. Scherer talks about the factors in robotics that tend to spook people and provoke a feeling of disgust, and how we might be able to move beyond this resistance as they become a more active part of the social fabric of our society.

GuestDerek Scherer

ExpertiseComputer Science and Robotics/Engineering

Recognition in BriefAfter graduating with a Master’s in Computer Science and Robotics from Columbia University, Derek worked in several robotics roles, including at the US Army Research Laboratory Robotics Program and the Weta Workshop, where he helped create special effects for the film “The Hobbit”. In 2006, he founded the Golem Group, creating animatronic systems, control circuity, and other technologies. The Golem Workshop continues to advance Character Robot technologies. Derek’s work has been featured in multiple publications, including CNet and The Kansas City Star.

Current AffiliationsFounder of Golem Group, LLC

Traversing the Uncanny Valley

Many of us are familiar with the notion of the uncanny valley, that ambiguous place between human and almost-human – often found in dolls, animation, and more recently androids – that can invoke feelings of trepidation and even fear.

Derek Scherer, a Robotics and Automation Engineer, walks through this valley more often than most by nature of his work. “The uncanny valley comes into play mostly because robots aren’t moving…there’s also a certain degree, if you look at wax museums, there’s a little creepiness where something’s human, but not quite, that’s really the definition of uncanny valley, when there’s that off-ness to something that is humanoid, and the reason it’s called a valley is because as something becomes more like us, we tend to like it more.”

This idea makes sense. For example, a teddy bear is more attractive to us than a real bear; it becomes more personable, more human in a sense. But once we reach a point where something is almost like us, there’s often a sense of revulsion, says Derek. In referring to a valley, we also imply that there’s an an other side to be reached, that point where actual humans interact with each other and recognize other humans in turn. “For robotics, the idea of jumping that chasm would be to create a robot that’s as human as we are in terms of how it’s perceived,” explains Scherer.

What are the origins or roots of the uncanny valley phenomenon? Derek explains that much of our brain is devoted to human interaction, and that in turn we are automated masters at tuning into the most subtle detail, such as changes in intonation of voices, eye movement and direction, and tiny facial muscle movements; all of these are communicated in signals that we interpret as emotion. If just one or two of these cues is “off”, the interactive signal is likely to become scrambled. The same effect doesn’t occur when we’re looking at a pigeon, for example, because humans don’t have the same sort of primal connection and understanding of that animal.

Bridging Humans and Robots

The uncanny valley becomes a potential issue when we start to look at the real progress that is being made in the development of robots and androids for service industries. While strides are certainly being made in the arena of humanoids (Federico Guerrini covers one of the latest social bots for Forbes), Derek believes that for practical application robots, service robots, hospitality robots, and other social robots, human-like is not necessarily better.

“Robots are meant to provide an experience for people, as in entertainment or social experience; my approach has been to avoid the uncanny valley as much as possible while hitting the positive aspects…I think this is one beneficial approach,” says Derek. While he wouldn’t discourage anyone from creating human-like robots that are as accurate as possible, Scherer is interested in how robots can best work for us. In his view, that’s in bringing characters to life that are characterizations from our imaginations (like Bugs Bunny, E.T., or many other beloved icons) that embody, sometimes hyperbolically, some of the positive characteristics that we tend to be attracted to in people (think big smiley faces or big, bright eyes).

Rethink Robotics’ Baxter is a good example of an industrial robot that fits the the types of positive embodiments to which Derek refers. He believes creating such characters to function in social situations may be a better way to create a positive experience.

Outside of characters, is there any “getting over” the uncanny valley in the human psych? Derek thinks we’ll first see greater acceptance of human-like entities through advancements in computer graphics, where there is a very rapid iteration cycle – if you want to move the skin i.e. frowning, creasing around the eyes, you can do so rather quickly in digital characters, whereas in robots the process is much slower in making the subtle transitions realistic. This ability, paired with lucrative funding opportunities from the video game, movie, and now virtual reality industries, makes it a likely place to gain ground in the valley.

But what about the robotics industry? Is this a chasm that can be crossed by robotics in the next two decades? Derek, like the rest of us, can’t foretell the future, but he notes that it will be interesting to see how our perceptions and standards change as we more often interact with virtual agents. “We may be able to be sold without believing its entirely human, that’s one possibility,” he says. In other words, perhaps we’ll learn to overcome our fears of human-like figures without reaching the point of exact replication.

Then again, maybe we’ll move towards human-like robotic embodiment more quickly than we realize. “Many of these technologies that are very complicated and also very personal have a tipping point…it might seem like it’s really far away because we’re very close, and we notice all the tiny details and they seem magnified, but a few more years of development…suddenly it clicks and it all just works and we forget that it was ever so bizarrely off,” suggests Derek.

He points out that artificial intelligence is a prime example of this idea. A few years ago, AI seemed relatively worthless outside the realms of Jeopardy. Most of us didn’t really have any use for the technology, but that turned around quickly with the introduction of smart phones, algorithms, and other machine learning trends that have seemingly transformed our world overnight. Robotics might turn out to be a similar sensation.



Stay Ahead of the AI Curve

Discover the critical AI trends and applications that separate winners from losers in the future of business.

Sign up for the 'AI Advantage' newsletter: