Fooled You! Robots Learn How To Deceive

Robots are becoming more human every day. Some robots can already sustain damage and reconfigure themselves, kind of like how our bones heal after we break them. Now others can deceive other intelligent machines and even humans.

Researchers at Georgia Tech have developed algorithms that let robots determine whether they are in a situation where they should deceive other robots or humans.

To test the team's algorithms, these robots participated in a series of hide-and-seek games. During the games, colored markers were lined up along three potential pathways to locations where the robot could hide. The hider selected a location from the three location choices, and moved toward that point, knocking down markers along the way. Once it reached a point past the markers, the robot changed course and hid in one of the other two locations, making sure not to hit any markers by its actual hiding spot.

Developing this algorithm required interdependence theory and game theory that tested the value of deception in a specific situation. The game satisfied two key conditions that the robots needed to warrant deception: There must be conflict between the deceiving robot and the seeker, and the deceiver must benefit from the deception.

And it worked! In 75 percent of tests, the hiding robot was able to evade the seeking one. Most of the failed tests could be attributed to not knocking enough markers down.

This research was funded by the Office of Naval Research, so this robot's deceptive capabilities are designed primarily to serve in military capacities. The study's co-author, Alan Wagner, says these robots won't have much use off a battlefield.

"Most social robots will probably rarely use deception, but it's still an important tool in the robot's interactive arsenal," Wagner says. "Robots that recognize the need for deception have advantages in terms of outcome compared to robots that do not recognize the need for deception."

The study was published online on Sept. 3 in the International Journal of Social Robotics.