Laughter takes many forms, from a polite chuckle to an infectious howl of mirth. Scientists are currently developing an AI system that aims to recreate these nuances of humor by laughing in the right way at the right time.
The team behind the laughing robot, which goes by the name Erica, says the system could improve natural conversations between people and AI systems.
“We believe that one of the important functions of conversational AI is empathy,” said Dr Koji Inoue, from Kyoto University, lead author of the research. “So we decided that one way for a robot to empathize with users was to share their laughter.”
Inoue and his colleagues set out to teach their AI system the art of conversational laughter. They collected training data from more than 80 speed-dating dialogues between male university students and the robot, which was initially teleoperated by four amateur actresses.
Dialogue data was annotated for solo laughs, social laughs (where humor is not involved, such as in polite or embarrassed laughs), and cheerful laughs. This data was then used to train a machine learning system to decide whether to laugh and choose the appropriate type.
It may seem socially awkward to mimic a chuckle, but empathetic to join in a hearty laugh. Based on the audio files, the algorithm learned the basic characteristics of social laughs, which tend to be more subdued, and happy laughs, with the aim of reflecting them in appropriate situations.
“Our biggest challenge in this work has been to identify the real cases of shared laughter, which is not easy because, as you know, most laughter is actually not shared at all,” said Inoue. “We had to carefully categorize the laughs we could use for our analysis and not just assume that any laugh can be answered.”
The team tested Erica’s “sense of humor” by creating four short dialogues to share with one person, integrating the new shared laugh algorithm into existing chat software. These were compared to scenarios where Erica either didn’t laugh at all or gave out a social laugh whenever she detected a laugh.
The clips were played to 130 volunteers who rated the shared laughter algorithm most favorably for empathy, naturalness, human-likeness and understanding.
The team said laughter could help create robots with their own distinct character. “We think they can show it through their conversational behaviors, such as laughter, gaze, gestures and speaking style,” Inoue said, although he added that it could take more than 20 years. before it was possible to have a “casual”. chat with a robot as you would with a friend.
Professor Sandra Wachter, from the Oxford Internet Institute at the University of Oxford, said: “One of the things I would bear in mind is that a robot or an algorithm will never be able to understand you. He doesn’t know you, he doesn’t understand you and he doesn’t understand the meaning of laughter.
“They’re not sensitive, but they could get really good at making you think they understand what’s going on,” she added.