While the concept of ownership comes so naturally to humans that even preschoolers can apply it, robots frequently struggle to understand the abstract notion.
Fortunately, researchers have developed a robotic system that allows bots to grasp ownership-related etiquette. This means robots can be programmed to determine who owns what and what it’s allowed to do with people’s possessions.
According to Matthias Scheutz, a human-robot interaction researcher at Tufts University in Massachusetts, instructing robots on ownership-related etiquette “is really, really important.”
One of these well-mannered robots will be presented at the AAAI Conference on Artificial Intelligence on February 1. The robot can learn who owns what from specific statements and from its own observations. It also learns not to mess with other people’s belongings from direct orders. For example, if the bot is told not to touch something by the object’s owner, it will conclude that this is a general rule to be followed.
Artificial intelligence researcher Xuan Tan and his colleagues at Yale University examined the robot’s conduct in a trial with blocks on a table. In one phase, Tan played with the red blocks, thus teaching the machine that they were his. When he asked the robot to pick up everything on the table and it reached for a red block, Tan told the robot, “that’s mine.”
After learning that it should not touch Tan’s belongings, and inferring that the rest of the red blocks must be his too, the robot cleared the table of everything except the red blocks. When Tan’s co-author instructed the robot to throw away a red block, it responded: “Sorry, I am forbidden to throw it away if it is owned by Xuan.”
Scheutz explained that this research is an important advancement in the development of robotics; one that goes far beyond mere blocks on a table.