Current technology may help us bridge the uncanny valley

Why does Frankenstein scare us but C-3PO doesn't?

The answer lies in what’s known as "the uncanny valley," a theoretical dip between points of human compassion and empathy. Valley researchers contend that as robots approach a certain level of human recognition our empathy turns to repulsion—that is, until they reach a range of motion and likeness that’s recognizable as fully human.

Seeing the valley is easier than seeing how to get around it. And in a world full of computer-generated animations—in everything from movies to video games—avoiding the valley is crucial to the future of entertainment technologies. To learn which way to go, Microsoft Research recently gave $100,000 to Carnegie Mellon researchers Jessica Hodgins, professor of computer science and robotics, and Kevin Pelphrey, associate professor of psychology.

Their project, "Exploring the Uncanny Valley," will use eye tracking and functional magnetic resonance imaging (fMRI) to record user response to animated sequences, modified video footage, and sequences from commercially released movies. Eye tracking will help recreate normal eye function by measuring focal points and recording eye movement relative to the head; fMRI will map brain activity to help design an accurately-responsive model of human behavior. The end result will be to provide guidelines for the production of animated human characters for use in video games, movies, and as avatars.

In addition to improving entertainment technology, the project should also offer a greater scientific understanding of the perception of human motion. Ultimately, the research could impact future generations of artificial intelligence designs, which could lead to genuine human interaction with robots or avatars.

The uncanny valley was originally proposed by Japanese roboticist Masahiro Mori in 1970, yet it wasn't until recently that the theory was tested experimentally. The theory holds that as the robotic appearance, motion, and behavior tips closer to human resemblance, we tend to focus increasingly on the non-human components to the point of distraction or fear. Exactly where that tipping point is remains unknown. Contemporary advancements in computer animation, and the popularity of movies such as The Polar Express, have made the uncanny valley a high-stakes concern. Carnegie Mellon researchers hope to locate the tipping point, and learn how to move past it.

The A. Richard Newton Breakthrough Research RFP Award recognizes projects "which clearly leverage innovative computational techniques…[and] demonstrate potentially high impact by solving problems of great importance to science or society." In addition to Carnegie Mellon, teams from nine other universities were awarded similar research grants.

Microsoft Research has been a long-time supporter of Carnegie Mellon, awarding numerous research grants and fellowships to faculty and students, as well as providing $1.5 million to establish the Center for Computational Thinking. Over 300 Carnegie Mellon alumni are employed by Microsoft, including Roy Levin (MSC '77), research director, Silicon Valley; David Drach (Tepper '94), V.P. and chief technology officer, Microsoft FRx; and Anoop Gupta (MCS '82), corporate vice president, education products & solutions, technology policy and strategy.


--Douglas Phillips