What causes the ‘uncanny valley’? How does it work?


How can humans tell that something is ‘off’ when presented with a CGI image that is very close to looking like the real thing? More specifically, why does this effect occur in things like CGI dragons or monsters or [that Sonic](https://pixel.nymag.com/imgs/daily/vulture/2019/05/01/01-sonic-the-hedgehog.w600.h315.2x.jpg), which are obviously not real to start off with?

And why does the ‘uncanny valley’ effect not occur in inanimate objects (for instance, I have never heard of anyone being creeped out by a CGI donut)?

In: Other

There’s this [video](https://youtu.be/PEikGKDVsCc) by Vsauce that kind of describes it really well.

We humans generally read emotions and stuff from people’s faces. And we know some inanimate objects aren’t human, and they’re not a threat. But for a humanlike object, we by nature try to read whether it’s a threat or not by reading the expressions, and when the expressions are unreadable, we, just in case, are scared of it as if it were a threat. It may sound stupid, but apparently it helped us survive so well that it is entirely ingrained in us.

When we see something that is recognizablely not at first glance, like C-3PO, then when are pattern recognition abilities detect something which suggests that this does have some what human features, we tend to find it endearing.

But when we see something that looks human at first glance, and our pattern recognition abilities detect that there is something not human about it, it sends a warning signal to our brains and something is not right. It is very similar to the signals you get when you notice that someone is sick oh, but even more intense.

The uncanny valley effect arises primarily due to our emotional connection to who we’re looking at.

Our brains do a lot of subconscious work when we look at someone, evaluating dozens of different factors without us consciously being aware of the evaluation.

If something looks sufficiently close to human, but just far enough that those subconscious emotional checks fail or start throwing off bad data, the subject looks cold and eerie.

The [uncanny valley](https://en.wikipedia.org/wiki/Uncanny_valley#Theoretical_basis) article on Wikipedia, specifically the section on theoretical bases for the phenomenon, has a wealth of information on the topic.

To your last question, I challenge you to find someone with an emotional connection to a donut. 🙂

I read that the one thing CGI artists have the most problems with are the eyes. You can wear the suit with the blue balls a la Avatar, and get some incredible motion reality, but the eyes are almost always a little creepy. Human eyes have tiny movements that are too hard to mimic properly and they end up looking dead inside. They’re integrating real eyes into filming more and more so the characters don’t get that creepy stoned look like Pixar characters have.

It’s all in the eyes.

Your brain is fine when it can tells something is human or not but when it appears to be the in between it gets confused and scared.
I will give and example:
When looking to a mannequin your brain recognized that the size is about right for a human and the volume and shape is also ok but it has no facial features your brain thinks
“Ok, this is a bit human but I can’t find the face to know for sure if its 🙂 or 😐 or >:( so I am not sure if I should signal that am in danger or not, let’s try both”
So getting the creeps looking at something is your body trying to warn you to that it is ok and it’s not fucking ok at the same time.