Researchers Report Teaching Robots to Lie

Robots can be trained to do a lot of things that humans do - including learning how to lie.

Researchers at the Georgia Institute of Technology say they've developed algorithms which endow robots with the capacity deceive humans or other intelligent machines they encounter. Their work, which was written up in the International Journal of Social Robotics, was funded by the Office of Naval Research.

Their software was successfully tried out on a couple of camera-equipped robots playing hide-and-seek. The hider robot was programmed to use deception, something that was unknown to the seeking robot.

I can already hear you asking yourselves: Why on earth would they want to teach a robot to be deceitful? According to the study's co-author, Alan Wagner, a research engineer at the Georgia Tech Research Institute, the occasions where "social robots" will need to use deception will be rare. Still, he said, it was an important tool for the robot's interactive arsenal.

"Robots that recognize the need for deception have advantages in terms of outcome compared to robots that do not recognize the need for deception," he said.

In practical terms, the researchers described situations where deception may be required, such as in search and rescue, where the robot may need to calm a panicking victim. Also, they said, robots used in battlefield situations would benefit from being able to mislead an enemy.

The Georgia Tech work builds upon the work of Swiss researchers who in 1997 found that in certain conditions robots can learn through experience to lie to each other.