Humans may prefer to work with robots that can communicate and express emotions, even if that means they're less efficient, according to a new study from University College London and the University of Bristol.
Researchers tested how people reacted when robots messed up a given task. They had participants work with three different versions (A, B and C) of the same robot, BERT2. Each would bring the humans ingredients to make an omelet: BERT A never erred, but BERT B and C both dropped an egg at some point.
Only BERT C could communicate with the humans and say "I'm sorry." It would also be visibly dismayed at the mistake, with an exaggerated look of sadness displayed on its face. BERT C would then show that it was going to try a different approach to the task, thereby rectifying the mistake.
![Mashable Image](https://cdn.statically.io/img/helios-i.mashable.com/imagery/articles/07B4tlV0pZwYzvuxr2MhEbH/images-1.fill.size_2000x1292.v1611680323.png)
The participants appreciated the transparency and semblance of control. One was quoted in the paper as saying that the vocal interaction with BERT C kept him from wondering what would happen next.
"It also let me know when he realized that he had dropped the egg," the participant said. "With the non-vocal machines there is a nervousness about when I should be holding out my hand, etc."
People also reported that C took the least amount of time to complete the omelet-making task, even though A was actually the fastest. But the participants said that A "seemed slower" or "quite slow."
Again, trust played a big role in evaluating the robots. Overall, more people said they'd use BERT C for a work-related task over the other robots.
The researchers noted, however, that we lack large-scale, long-term data on these concepts, and the paper itself has not yet been peer-reviewed. Still, the researchers were optimistic about their findings.
"I think that BERT's cartoon-like visage was something people responded very well to, the fact that it/he was just human-like enough without seeming creepy or threatening," Researcher Adriana Hamacher wrote to Mashable in an email. "[The study] drives home the fact that people respond well to robots that react and behave as human beings do."