Trusting robots may not be that easy during routine situations, but a new study presented at the 2016 ACM/IEEE International Conference on Human-Robot Interaction has found that humans tend to trust, and even follow, robots if they are stuck in an emergency situation.

Conducted by researchers at Georgia Institute of Technology, the study had 30 participants. A majority of them followed an “emergency robot”, which bailed out humans during a simulated fire without even exactly following the clearly marked exit signs.

The study in charge and a graduate student, Paul Robinette, himself was surprised with the findings. Robinette said most members of the research team had thought that not many participants would trust the emergency robot, and that they would have to devise another strategy to prove that the robot was worth being trusted.

Another researcher said one more thing that was obvious after the study was one should slap the words “emergency” on the robot’s side to make people give up their logic.

For the experiment, the participants were in a conference room and were told to fill out a survey. After some time, a smoke alarm went off and simulated smoke filled the hall. Considering itself stuck in an “emergency”, the robot took charge of the situation and led the participants through the smoke, down a new path. Study participants had not seen that path before and the path wasn’t as per clearly marked exit signs.

The participants followed the robot through a new path even though they could have easily exited through the clearly marked path they originally came from. 26 individuals among 30 test subjects followed the robot while two never left the room.

In a statement, co-author Alan Wagner said, “People seem to believe that these robotic systems know more about the world than they really do, and that they would never make mistakes or have any kind of fault. In our studies, test subjects followed the robot’s directions even to the point where it might have put them in danger had this been a real emergency”.