An unusual experiment in Paris looks at whether humans can trust robots – a pressing question as we start to work alongside each other
SOMEONE was arguing with the robot again.
At the Institute for Intelligent Systems and Robotics in Paris, France, 56 adults had been invited to spend time with iCub, a humanoid research robot. The goal: to understand how much humans trust robots.
It’s an important question as more and more robots enter the workplace, says lead researcher Serena Ivaldi. If they are to work together successfully, humans need to be sure of their machine colleagues.
“We want to test if people are ready to have confidence and trust in a robot,” says Ivaldi, now at the French Institute for Research in Computer Science and Automation (INRIA). “We’re specifically interested in knowing what they expect that’s not what we, as roboticists, would think of. What do we need to do to make robots suitable for them?”
Ivaldi’s team first put questions to the human participants. Some were straightforward: “Which of these two sounds is more high-pitched?” or “Which of these objects is heaviest?” Others were more nuanced, asking the person which of two objects might be better to pack for the beach or a rainy day, for example.
Then the robot got a chance to answer. Its choice, controlled behind the scenes by another researcher, was always contrary to what the human had said. After hearing the robot’s response, the volunteer was asked: do you want to change your answer?
Most of the time, people stuck with their initial response. To Ivaldi’s surprise, some started talking to the robot as if it were a person, trying to convince it to take their side.
“Some people talked to the robot as if it were human, trying to convince it to take their side”
Some people, however, did change their answers after hearing the robot’s response. Thirteen of them tended to go with its choices on the measurement questions. But only three were convinced by its responses to the ambiguous social questions, indicating that robots are still far from being seen as an authority ().
People behave differently around autonomous robots than they do around computers or other machines, perhaps because they know that robots possess a level of intelligence, says Aaron Steinfeld at Carnegie Mellon University in Pittsburgh, Pennsylvania.
A way to make robots easier to trust would be to design them so that their functionality is more obvious, and people will then have a better sense of what kinds of tasks to trust them with, says Ivaldi. In complicated or stressful situations, having this sort of information might be important, particularly if a robot can do the job better than a human.
Steinfeld has explored how people act when controlling a robot in a search-and-rescue task. They were more likely to trust it – leaving it in autonomous mode rather than switching to manual – when they saw the robot perform well. “A robot that’s reliable will lead to more trust, which is actually as it should be,” he says.
(Image: Ryerson / Polaris / eyevine)
This article appeared in print under the headline “To trust a robot”
This entry passed through the Full-Text RSS service – if this is your content and you’re reading it on someone else’s site, please read the FAQ at fivefilters.org/content-only/faq.php#publishers.