The case against anthropomorphizing robots

It’s a violation against one of Dieter Rams’s key principles of good design, according to the author of one new study.

Close your eyes, and picture a robot. Do you see a C-3PO-like humanoid, metallic and shiny and with a quirky personality? Do you see a Wall-E, with big, puppy-dog eyes that make you feel less alone? Or maybe you see a more sinister, skeleton-like Terminator, with red lasers instead of pupils?


Whatever you dream up, chances are the robot has a face and some kind of humanoid form. So many of the robots that are in public or private spaces, designed to interact with humans in some way, do–especially social robots with cutesy names such as Kuri and Jibo. Even this coffee-making robot barista, which consists of just a robotic arm but was designed to wave and inspect its environment with a humanlike personality. There’s a simple explanation for this. People tend to anthropomorphize all objects anyway, whether it’s a bunch of metal hard-wired together or a series of geometric shapes. Putting a human face on a bot serves as one way to make our sci-fi reality seem much less threatening. That’s why robots designed to welcome people to various spaces, like retail stores, hotels, transportation, and other places, are usually humanoid, using humanlike expressions, gestures, and language to communicate to guests that they’re welcome in an environment.

[Image: courtesy miLAB/IDC Herzliya]

But are their overtly humanoid forms even necessary, especially when they can be complex to design, manufacture, and maintain? Do robots really need faces?

A new study from researchers at the Media Innovation Lab (miLAB) at Israel’s IDC Herzliya and Cornell attempts to answer this question, and upend the current paradigm that the easiest way to help humans trust robots is make them seem as human as possible. Instead, the researchers show that you don’t necessarily need to slap a humanlike face on a robot for it to communicate effectively. That means that it’s possible to create robots that are aesthetically and mechanically simple that can use the barest of gestures to interact with people. What’s more, the lack of a humanlike face on a bot can help to mitigate our corresponding expectation that the robot has some measure of human intelligence–and our frustration when it does not.

To see how people would react to a totally abstracted robot, the team had to build one first. They decided to focus their efforts on creating a robot that could perform a greeting, like you might see in a retail store. That proved harder than it seemed, because the robot’s form couldn’t resemble any type of robot participants might recognize. The resulting “Greeting Machine,” as they call it, is a small white sphere that can freely move around a larger white sphere, almost like a moon orbiting a planet. The bot’s curved edges were inspired by research showing that people tend to view curvilinear shapes as warmer and friendlier than shapes with hard edges.

To create movements that people would perceive as greetings, the team called in some experts–an animator, a puppeteer, a choreographer, and a comic artist–who created four different gestures that the researchers could test out on participants to get their reactions. In two of the gestures, the small ball moved toward the participant. The team called these “approaches.” The other two were “avoid” gestures, where the ball moved away from the participant. Even though there was no explicit intention whatsoever to these movements, the participants overwhelmingly perceived the approach gestures as welcoming, positive, and warm, while regarding the avoid movements as a signal that the robot didn’t want to interact with them.

[Photo: courtesy miLAB/IDC Herzliya]

“An interaction that lasted a few seconds with an abstract robot performing minimal movement led to rich descriptions of opening encounters both negative and positive,” senior research scientist and paper co-author Hadas Erel told IEEE Spectrum. “People attributed intent and emotions to the robot’s gestures, and the social interpretations were extremely consistent between participants.”


In other words, the simple, two-sphere robot could give either a positive or negative first impression–and even elicit complex emotional reactions–just through abstract movement. One person responded to the avoid gesture by saying, “When I would walk in and it would face away from me, it was like, ‘I don’t want to talk to you.’ It’s weird, because it’s an object and it shouldn’t make me feel anything, but it did. It’s the same as if a person wouldn’t want to talk to you.”

The study suggests that we are doomed to see anthropomorphic intention in everything, even if it’s just code and metal. Designers don’t need to slap a face on a robot for people to view it in a social way.

As Cornell professor Guy Hoffman points out, humanoid design of robots doesn’t follow Dieter Rams’s principle that good design is honest design, because a robot that looks like a human overpromises what that robot can actually do. “With designs like the Greeting Machine, we are trying to exemplify ‘honest’ robot design, while also arguing that this minimal design can still enable a deep emotional, social, and psychological effect,” Hoffman tells IEEE Spectrum.

The research opens up the door to more creative ways to think about human-machine interaction beyond google-y eyes, cute names, and adoring faces. Even if the robot has no face, we’ll find humanity in it anyway.


About the author

Katharine Schwab is the deputy editor of Fast Company's technology section. Email her at and follow her on Twitter @kschwabable