"The Uncanny Valley" is a 1970 essay by Masahiro Mori, then at the Tokyo Institute of Technology, that introduced the concept of the same name. Mori hypothesized that a person's response to a humanlike robot would shift abruptly from empathy to revulsion as the robot approached, but failed to attain, a lifelike appearance.
Author: Masahiro Mori (Tokyo Institute of Technology) Original publication: Energy (Japanese journal), 1970 English translation: Karl F. MacDorman and Norri Kageki English publication: IEEE Spectrum, June 2012
Summary of argument
The essay proposes that as a robot or artificial figure is made more humanlike, an observer's affinity for it rises until, near the point of close-but-imperfect resemblance, the response drops sharply into eeriness or revulsion before recovering as the resemblance becomes indistinguishable from a real human. The dip in that affinity curve is the uncanny valley.
The essay was nearly unknown outside Japanese robotics circles for decades. It became widely cited only after the 2012 IEEE Spectrum English translation, which was authorized and reviewed by Mori.
Relation to AI-policy discussions
The essay serves as a philosophical anchor for several anthropomorphic-design discussions, spanning robotics, animation, and AI. It is older than the AI policy era but has become load-bearing as anthropomorphic AI deployment surfaces in litigation, policy, and product design.
In the Seemingly Conscious AI (SCAI) debate, Suleyman and colleagues develop the SCAI framework as the AI-system analog of Mori's robotics framework (Seemingly Conscious AI Risks — Bariach, Schoenegger, Bhaskar, Suleyman (Microsoft AI, 2025), We Mustn't Let AI Hack Our Empathy Circuits — Mustafa Suleyman (Nature World view, 2025)). The uncanny valley operates at the perceptual-aesthetic level; SCAI operates at the cognitive-attribution level. In this reading the two bracket the anthropomorphic-design risk space: the uncanny valley concerns AI that feels almost-but-not-quite human, while SCAI concerns AI that feels too convincingly conscious.
Lerchner's anti-functionalist position is structurally compatible with Mori's, in that both argue that a substrate problem — aesthetic for Mori, constitutional for Lerchner — cannot be overcome by behavioral fidelity alone (The Abstraction Fallacy: Why AI Can Simulate But Not Instantiate Consciousness — Alexander Lerchner (Google DeepMind, March 2026)).
The essay also functions as a philosophical anchor for AI Welfare / Model Welfare / Moral Patienthood debates about anthropomorphic interfaces. In litigation, the complaints in Garcia v. Character Technologies, Inc. and Raine v. OpenAI, Inc. both invoke anthropomorphic-design framings, and Mori's essay is cited as a foundational reference for that line of argument.
Relationships
- supports: AI Welfare / Model Welfare / Moral Patienthood (foundational anchor), Seemingly Conscious AI Risks — Bariach, Schoenegger, Bhaskar, Suleyman (Microsoft AI, 2025) (SCAI as AI-system analog), Garcia v. Character Technologies, Inc. (anthropomorphic-design framing anchor), Raine v. OpenAI, Inc.
- related: The Abstraction Fallacy: Why AI Can Simulate But Not Instantiate Consciousness — Alexander Lerchner (Google DeepMind, March 2026) (compatible substrate-matters argument), We Mustn't Let AI Hack Our Empathy Circuits — Mustafa Suleyman (Nature World view, 2025), Masahiro Mori (planned)
- instance-of: AI Welfare / Model Welfare / Moral Patienthood (foundational philosophical contribution)