Imagine driving along, following directions from your smartphone, when the polite voice tells you to take a turn. You oblige, only to realize that you’ve been directed onto a dead end.
We’ve all been there: trusting machines a bit more than we should.
A new study from Georgia Tech Research Institute sheds a bit more light on that phenomenon, finding that people will trust directions from a robot in an emergency even when the machine had misled them in the past. A group of participants followed a robot to get to a conference room, though the machine purposefully misled the group, passing the room multiple times before arriving at the correct destination. Then, the scientists simulated an emergency situation -- in this case, a mock fire -- to see if the participants would follow the same robot to safety.
Despite its earlier incompetence, and some participants being told told that the robot had broken down previously, the experiment’s subjects followed the robot’s directions. The volunteers listened to the robot even though it was directing them to an exit farther away than the doorway marked with exit signs that they used to enter the building. The experiment’s subjects only questioned the robot’s instructions when it malfunctioned during the emergency.
Researchers said were surprised by the results and have concluded that “victims in emergency situations may overtrust a robot, even when they have recently witnessed the robot malfunction.” Also, when the experiment was conducted without an emergency scenario in the simulation, volunteers did not trust a robot that had made previous errors, leading researchers to believe that the experiment’s robots were seen as authority figures during an emergency.
In future studies, scientists hope to understand the motivations that make humans trust robots, and whether factors such as an individual’s education level and other demographics influence their decisions.
These results could be encouraging because while earlier experiments have shown that people do not follow protocols during an emergency -- even with an active alarm -- this test may prove that putting robots in high-rise buildings will make residents more likely to leave. Still, scientists are wary of humans trusting the machines too much.
The research is scheduled to be presented March 9 at the 11th ACM/IEEE International Conference on Human-Robot Interaction in New Zealand. The conference is dedicated to “basic and applied human-robot interaction research.”
It should be noted that the study is very limited due to its small size -- only 42 people participated. Its conclusions should not be broadly applied.