It’s the weekend, and you decide to visit your grandmother who lives alone. But when she arrives, she realizes she has another visitor, and she hears through the door both of them laughing. You can’t make anything until you walk in and find that the visitor, seated across from your grandmother’s dining table, is a humanoid robot — laughing at your grandmother’s joke.
That won’t become a reality this year, or in the next 10 years, but it’s exactly the kind of scenario a team of scientists is working towards. Researchers at Kyoto University in Japan are teaching a robot how to laugh in response to human laughter. The robot, named Erica, can detect when a person is laughing, decide whether or not it is appropriate to laugh in return, and choose to respond with two different types of laughter: a small laugh and a loud laugh.
Research recently published in the journal Frontiers in robotics and artificial intelligence, performed on a humanoid robot that has a synthesized human voice and can blink and move its eyes while talking to humans. If the idea of a bot laughing crazy at your jokes sounds annoying to you, that’s because it is. . . But scientists hope it will help build more empathetic AI systems.
Think of a robot today and you’re likely to associate it with tedious tasks, such as stacking heavy boxes in a warehouse, harvesting vegetables on a vertical farm, or even unpacking pipes. But with the domestic robotics industry expected to reach $19 billion by 2027, more complex and empathetic robots have entered the picture. ElliQ, for example, is designed to tackle loneliness among the elderly, and Ollie’s creators claim it can stimulate dementia or Alzheimer’s disease patients.
With Erica, too, empathy was key. “One of the ways we show how we understand feelings or understand a situation is through laughter,” says Divesh Lala, one of the study’s authors. Studies have shown that when one person imitates what another person does, the act, known as reflexology, can build a strong relationship between the two people. In this case, Erica has been trained to reflect human laughter so that she can relate to people. The scientists collected data from more than 80 conversations between male college students and a robot, which were originally remotely operated by four actresses. Then the dialogues were analyzed and the different laughs were categorized as “social” (like the kind you laugh at just to be polite or because you’re awkward) and “fun” (like that real laugh when your friend cracks a good joke).
The scientists then trained the algorithm to discern basic characteristics of each type of laughter — such as a quieter laugh when you’re polite — in order to reflect them accordingly. “If you assume every laugh is equal, she will respond to everything, but if you don’t respond to anything, it’s embarrassing too,” says Lala. “If a robot is able to distinguish between the two, that’s a useful finding.”
The laughter algorithm on its own is pretty limited, but if you combine it with other features like natural language processing and back-routing (a gesture, for example, or periodic verbal acknowledgments to show the bot is listening), you might end up with a conversational bot that can To help the elderly fight social isolation or, as in Lala’s projects, to teach social skills to people of diverse nerves. “If they talk to a robot, maybe they can train to laugh at the right time, but you have to be careful with this, you don’t want to rely so much on the robot,” he warns.
It’s worth noting that none of this has anything to do with actual humor. Erica can’t distinguish your dad’s cliched joke from clever puns. At least not yet. The algorithm is not trained to process the meaning of the words, just laughs. “Erica doesn’t understand the kind of sense of humor, but if you interact with the user’s laughter, the user might feel [like] “She understands something,” says Koji Inoue, the study’s lead author.
Next, the team wants to add different types of laughter to Erica’s portfolio and link her language processing ability and laughter ability accordingly, so she can decide what’s funny and what isn’t based on the meaning of the words. “Our goal is human-like interaction,” says Inoue. A fully-talking robot with a silly sense of humor might not be ready in time for your grandma to give it a whirl, but then wait a few decades, and you might find yourself at that dining table, cracking jokes with a robot named Erica.