I think the uncanny valley is one of those revolutionary ideas that has been so thoroughly adopted that people wonder why you even have to describe it.
The point of the uncanny valley is that when simulating humans, increased fidelity does not inevitably result in a smooth improvement in human response to the simulation. And when I say that, you may respond “well, duh”, but see sentence 1.
Hmm. Can we formalize this comment such that it follows the following logic:
10 If A = ( [looks] Dead ) then 50
20 If A = ( [looks] Ill ) then 50
30 If A = ( [looks] Hostile ) then 50
40 If A = ( [looks] Artificial) then 50
50 print “Uncanny”
60 If A < > (“Uncanny”) then 10
Think subsections and nuances can be useful sometimes 
There was a Popular Mechanics article a few years ago that tried to convey this idea too. Most of the commenters, iIrc, said that the research misunderstood what “uncanny valley” meant–which just underscored the fact that nobody can even agree on a real definition of the term, much less prove it’s real.
Here’s my writeup of that from 2010: http://www.kith.org/journals/jed/2010/03/11/12737.html
You say uncanny valley . . . I say untapped dating pool!
See, this is an example of the uncanny valley – they look real, but not quite, right?
The people, I mean.
I don’t think it’s limited to faces. Put lifelike human feet on C3-PO and he becomes 100 times more creepy right off the bat.
Yeah. I’m also reminded of the beginning of Monsters, Inc. Once it’s revealed that we’re in a simulation lab, the kid in the bed is shown to be just a realistic-looking (in the context of a Pixar movie where the heroes look like stuffed toys) human torso on a spring-loaded lever. That fake kid looked much creepier to me than he looked when we thought he was an actual boy in bed. And I think that may have resulted in the simulation dummies in the sequel Monsters University looking more like actual dummies than resembling actual humans. This was explained in the movie as the result of the University acquiring older, secondhand, less-state-of-the-art testing equipment, but I honestly think it quite possible that more than a few kids in the audience were freaked out by Torso Boy in the first movie and Pixar deliberately made the dummies far less realistic in the sequel, since they’d be getting so much screen time to serve the needs of the sequel’s story.
Since when did anybody (Mori? The Valley itself?) predict that a cartoon image that was specifically designed to be human-like (and also beautiful!) could serve as a suitable continuum endpoint in an uncanny-valley study? This should be silly to anyone who knows how to read and who is capable of critical thinking.
When did “uncanny valley” become a single discrete monolithic authoritative unambiguous scientific position thing that goes around making predictions? (A trademark of bad science journalism is the portrayal of “science” as a uniform collective of people who agree about everything to the extent that one sensational unexpected finding *Changes What We Thought We Knew About The World!*™®.)
The uncanny “valley” refers to a valley of data points, on a graph. An interpretation of that valley is what makes predictions, not the valley itself. Aside from the cartoon stimulus being way off the mark here.
Science journalism trope: “New experiment forces scientists to think in a new way about X. We have (accidentally or on purpose) misconstrued X in an overly narrow and strong fashion, in order to depict it as having been subverted and/or thrown into doubt.” What I say to this is: ahhhh, no.
Showing someone image 9 of a ten image morph isn’t quite the same as just springing image 9 on them cold and acting like it’s image 10.
That particular sequence is available in this pdf. The end result is Jennifer Hewitt’s head pasted between Jasmine’s body and hair, so there isn’t really much making the next-to-last one more uncanny; both just come off as strange photoshop pics rather than true images. I think I’d feel off if they were moving.
A loosely-defined term, with built-in moveable goalposts is a SPLENDID tool for arguments though.
As a couple of commenters have touched on, it seems like the experiment may be trying to test the wrong thing. The key to the uncanny valley doesn’t lie in the image, or the robot, or the animation – it’s the subversion of the viewer’s expectations. Our discomfort isn’t that the robot’s eyes are dead-looking, it’s that we expect them to look alive, and they don’t. The phenomenon occurs when the item exists in a context where those expectations are raised and then dashed. This is why it’s more commonly associated with near- (but un-)realistic renderings like The Polar Express, but it can still occur with non-realistic renderings like the test dummy in Monsters, Inc…
It’s such a commonly-understood response that in more extreme renderings, it forms the basis for a lot of horror movie cliches – the dog that’s really an alien, anything with creepy kids, etc.
This topic was automatically closed after 5 days. New replies are no longer allowed.
