What evidence is there that the uncanny valley exists and what results, if any, have been gleaned from psychological studies of the phenomenon? All the stuff I’ve come across about the uncanny valley have been written by robotics or computer science experts, none of it from psychologists (i.e., the people whose business it is to study human perception). They all seem to take for granted that it exists and yet I can’t find any evidence beyond some vague anecdotal “of course it exists” answers.
If it has been shown to exist, have there been any studies about how it’s expressed cross-culturally? For example, Arab or Jewish viewers might be disgusted by a computer-generated talking pig with a strongly human appearance, or perhaps illiterate Azeri shepherds might find the whole idea of CGI to be inescapably odd, let alone seeing near-realistic animation of the human form.
Sorry, I assumed the term was widespread enough that I could talk about it without explaining. To those not willing to click through the Wikipedia link: The uncanny valley is
I can’t help wondering if it’s related to the response I experience when I see some movie star/celeb and I can tell they have had some sort of plastic surgery, but can’t quite put my finger on what it is that’s putting me off.
When this happens to tv stars I cannot watch the program anymore. I’m not listening to the dialogue or paying attention to the plot anymore. I’m just staring at them, unbelieving. And my mind seems to be cycling through; Is it the eyes? Is it the mouth? The chin? What the hell is it?
It totally creeps me out and I end up never watching that show anymore, even shows I previously adored.
Hmmm, maybe now I know what it’s called.
Am I correct? Is my response related to the uncanny valley?
Even though it originated in robotics, I think the idea has had the most application in CG. Polar Express and Beowulf were victims of the uncanny. The CG people fall just short enough of total realism to be off-putting. The techniques used to capture faces in movies like Pirates of the Caribbean and Avatar seem to have overcome the effect.
I don’t know of any actual studies, but most people recognize the idea when it is explained to them. It puts a name to the distaste they feel for attempts at realistic animation of humans, which isn’t present for more cartoonish depictions.
Well, from personal experience, the existence of the uncanny valley seems pretty undeniable to me.
Consider this sex robot:
Pretty creepy, yeah? But compare that a less human robot like C3P0 from Star Wars – C3P0 isn’t creepy at all. If that’s not the uncanny valley, then what?
There’s also this one:
That still picture looks OK. But if you ever see a video of that robot, then there’s something really off about how it moves, how it blinks, etc. It feels alien and unnatural in a way that something less human like C3P0 doesn’t.
Apparently there really is only anecdotal evidence, as discussed in this pdf.
Of course even those authors then go ahead assuming it does and discussing possible mechanisms that may underlie it.
Cognition experts are unsurprised by the Uncanny Valley, would in fact likely be shocked if it did not exist, but they have not explicitly studied it, as such, but … We are amazingly finely tuned to recognize faces. Face recognition uses multiple brain areas including a special area caused the Fusiform Face Area (FFA). We have the ability to recognize a face as the same face, and not a different face, whether it is presented to us in from this angle or that, in one emotional reaction and expression or in a host of others. With that skill comes an equally great skill at mismatch detection, which allows not only to tell people apart but to recognize small very very small mismatches between a facial expression and the emotional content of what is being stated and thus to often detect when someone is telling us something false or hiding something. Huge evolutionary pressures on doing that well and on having the abiltiy to hide untruths even better - an arms race that made us very very good at having expectations of what a face should look like for a particular moment and being very uncomfortable if it does not. We are sensitive to facial incongruities.
In short the research has been done on its mechanisms, even if not on documenting how close you have to be for it to exist, but you won’t find it searching for “Uncanny Valley” - see here for a taste:
If something is clearly an abstract representation of a face those circuits will not be on. How close to real does it need to be to turn them on and does experience change that? Them’s interesting questions.
You know, I thought I was clear about wanting more than just vague anecdotal evidence but I suppose I should have been more explicit about it. Anyway, I get suspicious when people say “Everyone agrees about this” (whatever “this” is) without firm evidence because inevitably it will be revealed that everyone doesn’t agree about it.
I don’t know, I think it’s more that you’re comparing the celebrity to how they looked pre-op and wracking your brain playing “spot the difference”.
If you can dig up more details I would be very interested in knowing more about this study. According to wikipedia the uncanny valley is a hypothesis developed by a Japanese roboticist, hence the prevalence of robotics/CGI writers on it and why I’m asking for any psych research on it.
But C3P0 is a costume worn by an actor, there is an actual person inside the suit and thus his actions are more fluid and realistic. And perhaps you’re being creeped out as much by the giant sex toy aspect of the robot than simply just the poor imitation (can you honestly say you’re not to some extent?).
I suspected the evidence was only anecdotal, thanks for confirming. Sure, experts can reasonably expect that the uncanny valley exists, but that’s still no excuse for not examining it directly. You’d think the CGI/robotics folks would be pushing for research on the subject so they could more easily fool human perception, perhaps it’s latent hard scientist disdain for the social sciences. Even something as simple as a survey on emotional reaction to different CGI images would be nice.
Anyway, to those of you who maintain that the uncanny valley is inevitable, I give you these mannequins:
I’m actually quite delighted at the artistic achievement of these mannequins and not creeped out even though there are several signs they’re not human.
What evidence would you accept as NOT anecdotal? Isn’t the plural of anecdote called “data”? Thousands of people (including several on this thread) have reported the effect in their own personal experience, so it clearly exists. Now that doesn’t prove it exists universally, of course – maybe some people like you are immune to the effect, but the effect clearly exists for many people (such as myself). Therefore it’s real. QED.
I say from personal experience that computer generated Yoda from Star Wars looks “normal” to me, whereas computer generated people from “The Polar Express” look “somehow wrong” to me. I’ve met many many people who say the same thing. If the uncanny valley doesn’t exist, then can you explain why lots of movies made photo realistic CGI of animals and aliens, but no movie has ever even attempted photo realistic CGI of people? (Other than just touch-ups of live footage).
In any case, we’re talking about psychology, not physics or chemistry. It’s extremely difficult (some would say impossible) to do truly scientific experiments in psychology. There is no conceivably experiment to even ask the question “is this almost human robot more or less creepy than that less human robot?” without introducing potential experimenter bias in the results. I don’t know how it’s possible to get evidence you’d find convincing.
To summarize my reading of the article–“uncanny valley” is useless as just about anything more than a pop culture term, there might be a bunch of smaller ones, there might be gender-based differences in the perception of it, there may be differences between seeing it on a screen and in real-life, and any problems seem to be overcome quickly by exposure to it.
Even the clipped out part is just incorrect - for the first 50 minutes of Benjamin Buttons, for example, we were treated to a completely CGI human of the title character who was voiced by and later acted by Brad Pitt. (It may have worked because the initial form of the character did creep you out some and was supposed to and then became less creepy as it gradually youthed into Pitt. We accepted the creepy feeling as “right” and probably therefore turned off our incongruity detectors.)
Zoinks, again, research has been done that bears on the question, just not calling it “The Uncanny Valley.” The ones cited are examples of research that documents that we are very good at noticing structural incongruities, documents ERP markers of that identification, and investigates the neural mechanisms involved. Here’s another study that shows how incongruity results in a perception of “wrongness”, this time having the voice be incongruent to the face.
:eek::eek::eek: Let’s nip this one in the bud. There is a common saying which you are trying to remember here, but the saying goes “The plural of anecdote is not data” (emphasis added). The point of the saying is to emphasize that you don’t get good, verifiable information by putting together a bunch of anecdotes, even if you have hundreds or thousands of them. Good verifiable data is gathered under controlled conditions. Anecdotes are not. This means that no matter how many anecdotes you have, you still don’t have data.
The problem is, we’re dealing with something that is a subjective experience. The whole concept is that people find things “creepy.” To build an experiment about this, we’re still going to have to rely on people’s subjective feelings. The type of study that needs to be done would be calibrating people’s response to more and more human looking stimuli.
So, basically what I’m saying is, while the statement is wrong, the idea is not that far off. The only way we are going to get data about this is to put together people’s subjective experiences.
Anecdotes can be a starting point for scientific investigation but they can’t be the only thing you’re basing a theory on. For one thing, people have a habit of noticing patterns that don’t exist, leading to stuff like only wearing a certain shirt when one’s team is losing. Additionally, people often fabricate memories without consciously being aware of it. These objections are just off the top of my head, I could probably name a lot more if I had to.
And what do you mean it’s impossible to conduct scientific experiments in psychology? More than one hundred years of psychological experiments would say otherwise. If you have any specific objections to an entire branch of science please list them.
Another good article. Interesting how people get used to the uncanny within minutes, although in retrospect it shouldn’t be surprising considering how humans anthropomorphize stuff as inhuman as cars and weather patterns.
Ha, sorry, I had a bunch of tabs open in Firefox. I forget why I was even looking at that.
Sorry, didn’t mean to ignore what you said. I suppose if you were to conduct a meta-analysis collecting research on anything remotely resembling the uncanny valley then you could come closer to constructing an actual scientific theory instead of just the hypothesis floating around. I guess it’s just waiting for some budding researcher to notice a gap in the literature.
Sure, but it’s still possible to test people’s subjective experience of something. That’s the point of a whole bunch of the social sciences. And data from subjective experience is not in itself unscientific when collected in a rigorous fashion. For example, ethnographic observation is built on the researchers’ subjective reporting and the participants’ subjective accounts. However, considering the amount of training ethnographers receive in observation and memory tricks, plus the copious amounts of field notes they take supplemented by videos, photos, and what have you, then calling the whole thing merely anecdotal evidence would be very wrong-headed.
“WTF is the uncanny valley?” 
