Lorem Ipsum

I regularly see people talking online about the uncanny valley. It's an idea proposed in 1970 by roboticist Masahiro Mori: he suggested that if robots look very humanlike but not entirely human, they'll be perceived as creepy.

But Mori didn't, as far as I can tell, do any research to back up this idea. So for the past couple decades (I've seen this especially in the past five years or so), people have been taking it for granted that the uncanny valley is real, and using it to describe a variety of phenomena, without having any actual evidence for it other than gut feeling that it sounds plausible.

A recent Popular Mechanics article, "The Truth About Robots and the Uncanny Valley," discusses some attempts to actually measure the u.v. instead of just theorizing about it.

One of the things the article makes clear is that it's hard to even define exactly what the u.v. is supposed to be. But there've been some experiments that attempt to pin things down, and they've had inconclusive results. For example:

Proposing a wide-reaching theory is one thing, but applying any sort of academic rigor to vague notions of familiarity, repulsion and even humanity has shattered the theory into countless smaller ones. "It turns out that there may be more than one uncanny valley," [Karl MacDorman, director of the Android Science Center at Indiana University] says. "It's not the overall degree of human likeness that makes [a robot or animated character] uncanny. It's more a matter of a mismatch."

MacDorman did a study in which the results seemed to suggest that men were more subject to a u.v.-like effect than women. But MacDorman's experiments seem to have involved images on a screen. Other people seem to suggest that the u.v. doesn't exist at all in in-person interaction:

"In my experience, people get used to the robots very quickly," [roboticist David] Hanson says. "As in, within minutes."

[...]

According to all of the roboticists and computer scientists we interviewed, the uncanny is in short supply during face-to-face contact with robots. [...Despite some comments about a couple of creepy robots on YouTube,] [i]n person, no one rejected the robots. No one screamed and threw chairs at them, or smiled politely and slipped out to report lingering feelings of abject horror. [...] A previously skeptical journalist wound up smiling and cuddling with the ominous little CB2.

There are a bunch of reader comments at the end of the article, several of which say, essentially, "But you're measuring the wrong thing! The u.v. is when a robot looks 100% human, but behaves oddly!" Which just underscores the point that there are a bunch of different ideas about what the u.v. is.

And few of those ideas have actually been scientifically tested. It's always seemed to me to be a kind of Aristotelian approach (only insofar as it involves creating theories without experimental data): Mori came up with the idea of the u.v., it became a popular idea, and now most people take it for granted, without anyone having really tested it.

So it seems to me that the burden of proof that it exists should be on the people who claim that it exists. If it's real and measurable, then let's define and measure it.

I'm not saying it's not real. But it seems to me that the evidence we have for it mostly consists of people saying "Wow, [robot x] sure is creepy." And there are a lot of things that we find creepy that aren't even remotely human-looking.

So until we have (a) a clear and agreed-upon definition for the u.v., and (b) studies confirming its existence, I'm inclined to believe that there are a lot of factors that can contribute to someone finding a given robot creepy, and a lot of individual variability in responses to robots.