The Uncanny Valley


2024 Update: This feature image was created with Adobe's AI Generative Fill Tool using the prompt "Uncanny valley doll next to young boy". When this article was originally published in 2017, computer generated images were not widely available to the public.

If you've never heard of it before, the uncanny valley is a phrase used to describe the discomfort-inducing zone that an image or likeness of a human passes through on its way from abstract to representational. The term was first coined in 1970 by Japanese roboticist Masahiro Mori in an essay that received little attention at the time, but has since exploded in popularity and relevance. He hypothesized that "a person's response to a humanlike robot would abruptly shift from empathy to revulsion as it approached, but failed to attain, a lifelike appearance." (E.N. on orig. translation, MacDorman & Kageki 2012).

Graphing the Uncanny Valley

Graphically describing the uncanny valley. Our perception varies greatly in the spectrum between an industrial robot and a healthy person. (Unsurprisingly, zombies are the worst.) The effect is amplified by movement.

You're probably familiar with a few examples already. It's a phenomenon seen in fields ranging from robotics to animation, and that is both unfortunate and unavoidable.

Aside from the implications on the future of robotics, the idea of being able to trigger a base human instinct like fear or horror—unintentionally—is fascinating. It got me thinking about how much humans rely on visual feedback and about what can be considered to be unintentional communication in general. At Fat Pencil Studio we create visuals and presentations intended to tell a story, so we are very purposeful with what elements we choose to include and exclude. Our clients often ask us to add clothes or other identifying characteristics like body build or even skin color to our 3D models, but we argue that almost always, that’s not a good idea.

Our experience tells us that as a human stand-in gets closer to a lifelike appearance, every detail then becomes more important and more likely to push it into the uncanny valley. Because our brains are so well conditioned to respond to differences from a normal, healthy human, small details stand out too much, and omissions become obvious and ultimately distracting. The bigger picture is lost, and the workload grows.

It goes without saying that visual communication is all about sending the right messages. From birth we are taught to associate certain feelings and emotions with different colors and shapes, and we all bring with us varied backgrounds and experiences that determine how much we are influenced by those conventions. The key to good visual communication is knowing your audience, knowing exactly what you need to say, and keeping everything else out of the picture.

Our top argument for why we keep our models simple and generic is that we never want to suggest anything about peoples' personal lives, or paint a picture that opens the door to personal judgement. Often we're working on cases where there are victims and where people's freedom may be at stake. To do the best job we can, we must represent only what has direct bearing on the circumstances and avoid provoking biases we know exist. When making models I always assume that my audience will bring with them some kind of internal bias that may sway them, and that I can't ever know what it may be. Since we never know what's truly in other peoples' minds, it becomes critical that any visuals come stripped of details that make suggestions we don't have control over.

Jannine Hanczarek was a Designer at Fat Pencil Studio from 2017-2020.