Relatable Robots Make Mistakes

It's pretty clear from the way that technology is progressing that one day, robots will be our cashiers, our restaurant servers, and maybe even our therapists. Right now, however, there are still a few glitches that need fixing. But what if those glitches are exactly what humans need to feel comfortable around robots? That's what the research suggests: we like our robots better when they make mistakes.

Participant interacting with the robot during the LEGO building session.

Just The Way You Are

For a study presented in 2015 at the International Conference on Intelligent Robots and Systems, volunteers interacted with two types of robots: Emotional Robot With Intelligent Network, or ERWIN, a robot with the ability to express five basic emotions; and a small yellow robot designed to study child social development named Keepon. In the first phase of the study, the robots did their usual robot thing, interacting with humans flawlessly and carrying out tasks with precision.

In the second phase, they made mistakes. ERWIN forgot people's names and favorite colors — looking sad when it did so — and Keepon went through wild mood swings, dancing with joy or staring at the ground in sorrow. When the volunteers rated their experiences, they overwhelmingly reported more meaningful interactions with the error-prone robots.

But those robots were "emotional," and it could have certainly been the human-like emotions that made people prefer them. In 2017, a team of researchers from Austria and the UK proved that it really is the imperfections that make robots likable. They had volunteers work with NAO units — white humanoid robots 58 cm (about 2 feet) tall — that asked them questions about robots and then gave them instructions on how to build something out of LEGO. For some people, the robot spoke each question and command flawlessly; for others, not so much. They cut off the volunteers before they could finish answering a question, repeated single words in a glitchy loop, fumbled with LEGO pieces, and instructed the volunteers to throw blocks on the ground. Again, the human volunteers liked the faulty robots better. Importantly, there wasn't any difference in how intelligent or or humanlike they thought each robot was, showing that mistakes don't lower our trust in robots; they just help us relate to them.

Perfection Through Imperfection

This makes sense when you think about what makes us like other humans. There's a phenomenon known as the pratfall effect that makes us like talented people better when they screw up. Why? Because it brings them off of their pedestals and down to our level. It's why we love Jennifer Lawrence for tripping at the Academy Awards, and why Joe Biden's gaffes have inspired dozens of lighthearted internet memes. Robots are perfect almost by definition, so when they make harmless mistakes, we like them a little bit more. There's an important lesson there for engineers: the perfect robot may be one that makes mistakes.

And if you'd like to create your own robot (mistake-free or not), you can get started with something simple, like these Raspberry Pi robot kits. And if the thing malfunctions, just think about how lovable it is.

Even Robots Get Sad

Written by Ashley Hamer August 25, 2017

Curiosity uses cookies to improve site performance, for analytics and for advertising. By continuing to use our site, you accept our use of cookies, our Privacy Policy and Terms of Use.