Anyone who hears the robot complaining will cry

Trash talking to distract opponents has a long tradition, and now researchers at Carnegie Mellon University have shown that frustrating words can be distracting even when they are spoken by robots.

The trash talk in the study was decidedly mild, with phrases like "I have to say you're a terrible player" and "You're getting worse and worse." Even so, the players who played against the robot (a commercially available humanoid robot called Pepper) performed worse when the robot abused them and better when it encouraged them.

Aaron M. Roth, the lead researcher, said that many of the 40 participants in the study were skilled gamers and fully understood that the machine was the source of their discomfort. One participant said: "I don't like what the robot said, but that's the way it was programmed, so I can't blame it."

The researchers found that overall, human performance was weakened regardless of the skill level.

The research, presented last month at the IEEE International Conference on Robotics and Human-Computer Interactive Communications (RO-MAN) in New Delhi, India, deviates from typical human-robot interaction research, which focuses on how humans and robots can work most effectively together.

"This is one of the first studies of human-robot interaction in an uncooperative setting," said Fei Fang, an assistant professor in the Institute for Software. The research is important in an era when the number of robots and IoT devices with artificial intelligence capabilities is expected to grow exponentially. "We can expect home assistants to cooperate, but in situations like online shopping, their goals may be different from ours," she said.

The study was the result of a research project in a course taught by Fei Fang on methods for social good in AI. The students wanted to explore the use of game theory and bounded rationality in robotics, so they designed a study in which humans would compete against robots in a game called "Guards and Treasures," a game that is typically used to study interactions between defenders and attackers in security game research.

Each participant played 35 games against the bot, hearing both encouragement and trash talk from the bot. Although the human players' reasonableness improved as they played more games, those who were criticized by the bot scored worse than those who were praised.

People's behavior is greatly influenced by what other people say, but research shows that people also respond to what machines say, said Afsaneh Doryab, a systems scientist at CMU's Human-Computer Interaction Institute (HCII) and an assistant professor in the University of Virginia's Department of Engineering Systems and Environment. This ability for machines to respond quickly could have implications for automated learning, mental health treatments and even using robots as companions, she said.