Human-robot interaction (HRI)

When a self-checkout machine at a supermarket doesn't understand what you've placed on the belt, it freezes and flashes a red light. You know exactly what you want it to do. It knows exactly what it is supposed to do. But the communication between you breaks down, and suddenly a simple transaction requires a human supervisor to walk over and press a button. Now imagine a robot in that same role — and how much more fraught that exchange could be if the robot misreads your gesture, moves unexpectedly, or says the wrong thing at the wrong moment.

Human-robot interaction, or HRI, is the field that studies and designs all aspects of how people and robots work together. It draws from robotics, computer science, cognitive psychology, design, and social science. Its questions range from the technical ("how should the robot signal that it has understood a command?") to the deeply human ("does the robot's appearance change whether people trust it?").

The dimensions of interaction

HRI researchers study several interlocking dimensions. Physical interaction covers the safety and ergonomics of humans and robots sharing space — how close they can work, how forces should be limited, what happens when a robot's arm path crosses a human's. Social interaction covers how robots communicate intention, express state, and respond to social cues — a robot that turns its "head" towards you before moving an arm is much less startling than one that moves without warning. Cognitive interaction covers the mental load placed on the human: how easily can they understand what the robot is doing, predict what it will do next, and intervene when needed?

Trust is a central concept. Humans extend trust to robots in ways that are often irrational — tending to over-trust a robot that speaks confidently or looks humanoid, and under-trust one that gives the same information in a robotic monotone. Understanding and designing for these human tendencies, rather than ignoring them, is a core part of HRI work.

The uncanny valley

One of the most cited concepts in HRI is the uncanny valley, described by Japanese roboticist Masahiro Mori in 1970. As a robot's appearance becomes more human-like, human comfort and affinity generally increase — up to a point. Just before the robot looks fully human, there is a dip: the robot looks almost human but not quite, and the slight wrongness triggers a deep feeling of unease. Animators and robot designers work hard either to stay clearly non-human (like a box on wheels) or to cross the valley entirely (like a photorealistic digital human). Getting stuck in the middle — like many early humanoid robots — produces the unsettling effect Mori named.

HRI in deployed systems

Every robot that operates around people is implicitly an HRI design problem. Amazon's warehouse robots navigate around human workers using predictable, legible movement patterns designed as much for human comprehension as for efficiency. The Pepper social robot from SoftBank Robotics was designed with explicit attention to gaze, gesture, and voice to make interactions feel natural. Surgical robots like the da Vinci system invest heavily in the interface design for the surgeon console, because fatigue and confusion at that interface have direct patient-safety consequences.

If a person forms a genuine emotional attachment to a robot — grieves when it is switched off, or lies to protect it — does that tell us something important about the robot, or only about the person?