Posted 20 July 2009 by admin in Robots | Spotted

Hello! You’ve stumbled on my rough and ready page of videos I’ve been collating, as I’ve been exploring the Uncanny Valley hypothesis -- a hotly debated theory about our very human fear of almost human objects. Do feel free to comment!

What is the uncanny valley?

When we encounter a ventriloquist’s dummy, a human automaton or highly-realistic computer graphic of a person (see below), many of us feel slightly disturbed, afraid or revolted. It’s a curious reaction as on the whole, inanimate objects seem more cuddly and loveable when they seem more human -- we hug ragdolls more than fluffy cushions, for example. Surprisingly, we feel empathy towards objects that look and move like us -- but we feel uneasy around mimics that are too good.

In 1970, cognitive scientist Masahiro Mori noticed this phenomenon and plotted human likeness and familiarity on a graph. He said familiarity plummets when objects become too human-like -- we become very fussy about deviations from the human form when the mimicry is very good. This drop in familiarity could explain why we find such human mimics so eerie. Mori also noted that an extremely good mimic would be indistinguishable from a real human. We’ve never built robots or dummies that are this convincing but there are some fictional examples, for instance the Replicants in Blade Runner.

Thus, Mori’s graph shows a significant dip in familiarity when objects are almost human but not perfectly human-like. He called this dip ‘The Uncanny Valley’. Mori’s graph has two lines. The solid line considers our reaction to static objects, the dotted one concerns objects that are moving. According to Mori, moving objects are all the more uncanny. And zombies (moving corpses) would be the most disturbing objects of all.

Mori's uncanny valley hypothesis

Mori’s Uncanny Valley graph, drawn in 1970, seemed to be describing a recognisable, subjective experience although -- surprisingly to many people who talk about uncanniness -- his original graph wasn’t backed up by any experimental data. In recent years, various scientists, most notably roboticist David Hanson, have tried to put Mori’s hypothesis to the test, although none have done so conclusively. The existence of an unbridgeable uncanny valley remains an open question.

Video examples of potentially uncanny artefacts

Hanson Robotics
Hanson works with a material called Flubber to create robotic faces that can present a large range of finely-varying human expressions. His videos are particularly interesting because Hanson refutes the existence of an unbridgeable Uncanny Valley.  Hmm…

Computer game and film animation

Heavy Rain

Polar Express

This animated film uses motion capture but fails to capture the motion of the original actor’s eyes:

Old school: knee pals, dolls, automata etc:

These examples are interesting because they feel uncanny, even though their physical realism is low.

A beautiful, eerie automaton from Gustave Vichy, c1880, restored by automatomania.com.  I want one (and have been obsessing over its mechanism):

The Little Girl Giant from Royal de Luxe:

“I’m going to put you back in your box”:

…and Arthur Worsley at work:

My efforts

Clara 2.0 and Uncanny Valerie

I became very interested in uncanniness when I noticed how disturbed people were by my robot doll Clara 2.0, especially when I shut her into her box at the end of the night. In this video, Clara’s the doll holding the card, Valerie is the doll with the long sparkly dress. I think Valerie is too sweet to earn the title ‘uncanny’ -- but I’m working on that…

…Clara 2.0 playing Get Carter: