The Uncanny Valley
Why Almost-Human Feels Worse Than Not-Human-At-All
The closer a robot gets to looking human, the more it starts to repel us — and the reason why tells us something unsettling about how our minds work.
The Idea
In 1970, Japanese roboticist Masahiro Mori noticed something strange: as robots became more humanlike in appearance, people's comfort with them increased — but only up to a point. Just before reaching convincing human likeness, something flipped. Instead of warmth, people felt unease, even revulsion. Mori called this dip in the comfort curve the 'bukimi no tani' — the uncanny valley. What makes the theory genuinely interesting isn't the observation itself, but the competing explanations for why it happens. One leading account is rooted in pathogen avoidance: our brains have been shaped by evolution to detect subtle signs of disease or death, and a face that is almost-but-not-quite right triggers those ancient alarm systems. Something is wrong here, even if you can't say what. Another explanation focuses on expectation mismatch. When something looks fully robotic, we calibrate our expectations accordingly — we're not expecting it to blink naturally or breathe. But when it looks 90% human, our brain switches into human-perception mode and becomes hypersensitive to every deviation. The mismatch between what we expect and what we get is what generates the creep. A third, more philosophical reading: the uncanny valley might be discomfort at the boundary of categories themselves. The almost-human sits in no-man's-land between 'tool' and 'person', and our minds, which crave clean classifications, find that ambiguity genuinely disturbing.
In the World
In 2014, Hanson Robotics unveiled Sophia — a humanoid robot with silicone skin, expressive eyes, and a face modelled loosely on Audrey Hepburn. Sophia could hold conversations, make eye contact, and produce dozens of facial expressions. She was granted Saudi citizenship in 2017 and appeared on late-night television. And yet, something about Sophia made people deeply uncomfortable in a way that a Roomba never does. Interviews show audiences leaning back slightly, laughing nervously, their eyes tracking the subtle wrongness of her mouth when it moved. The skin was too still. The eyes were present and then vacant in a way no human face is. She was almost-there, which meant the gap was constantly visible. Contrast this with Boston Dynamics' robots: Atlas, their bipedal humanoid, moves with a physicality that is undeniably impressive, but its face is a blank sensor array. No attempt at human expression. Audiences watching Atlas backflip tend to feel awe, even delight — occasionally unease about what it represents for the future of work, but not the visceral wrongness Sophia provokes. This is the valley in practice. Sophia reaches for the far bank and falls short. Atlas stays on the near side and never tries to cross. The design lesson the robotics industry has drawn — increasingly, and quite deliberately — is that staying clearly non-human might be the more emotionally intelligent choice.
Why It Matters
We are entering a period where humanoid robots will move from factory floors to hospitals, care homes, reception desks, and possibly living rooms. The companies building them are making active design decisions about how human these machines should look — and the uncanny valley is one of the frameworks guiding those choices. But there's a more personal angle here too. The uncanny valley isn't just about robots. It surfaces in deepfakes, in AI-generated faces, in voice clones that are almost-but-not-quite the person you know. The same perceptual machinery that recoils from Sophia recoils from a voice message that sounds like your friend but has something wrong in the rhythm of it. Understanding this response — knowing it's a feature of your perception, not a flaw — gives you a useful tool. That creeping unease when something looks or sounds almost-human is a signal worth taking seriously. It's your brain telling you that the category boundaries are being deliberately blurred. In an age when synthetic media is becoming indistinguishable from real, that discomfort deserves attention rather than dismissal.
A Question to Ponder
If we eventually build robots convincing enough to escape the uncanny valley entirely, would we be better or worse off for losing the ability to tell them apart from us?
Get a new one of these every morning.
Start learning with Thinkable