Teaching robots to laugh is humorously difficult

A robot that smiles can be seen as ominous, but would it seem less creepy if it laughs at the right time during a conversation? That’s a theory that scientists have recently tested. A group of researchers at Kyoto University in Japan have made fun of a smiling robot named Erica, powered by an AI system focused on conversations.

Since laughter is a normal part of human dialogue, they reasoned, it could be helpful to see how humans respond to chatty robots that also allow them to laugh. Their findings were published last week in the journal Frontiers in robotics and AI.

Artificial intelligence is good at logic, but laugh? Not so much. For starters, researchers recognized that there are several reasons why people laugh — and that complicates matters. To make the AI ​​system easier, they generalized laughter into two categories: shared social laughter, when the AI ​​laughs in response to a human’s laughter, and solo laughs happily, when the robot laughs in response to a subject or laughs. while talking.

The researchers trained the AI ​​model on how and when to laugh by allowing it to participate in a form of speed dating with male college students. Erica was operated remotely by a female actress who spoke into a microphone and controlled physical movements such as head nods and other gestures.

Chats lasted 10 to 15 minutes and data was collected from 82 conversations. Researchers recorded the conversations using microphones and cameras and annotated them based on when social laughter and people’s solo laughter occurred, and how these laughter differed. This data was then used to train the AI ​​system to teach it when to smile and what type of smile to use. They then applied their shared smile algorithm to existing conversational software and asked 130 volunteers to listen in and rate how well the robot simulated empathy, understanding and human likeness.

Overall, the researchers noted that in situations where shared laughter was appropriate, Erica and his algorithm did a good job of convincing people that it was paying attention to what was being said. But it had some shortcomings and limitations. Erica reacted well to laughing, but had no idea when to laugh on her own. The researchers wrote in their discussion that it may be because learning to respond to a prompt is easier than actually understanding why the content of the conversation can be funny.

Whether Erica really gets human humor is just one part of a big project that roboticists and engineers are undertaking: giving robots social skills. Since 2017, scientists have been working on how to make a robot laugh convincingly (big tech companies such as Microsoft, IBM and Meta are also interested in this). A month ago, Italian engineers debuted a bartender robot capable of chatting (unfortunately, it will be shelved in the near future due to privacy concerns). The idea is that giving facial expressions, body language, speech and the ability to understand and respond to people’s demands makes robots more attractive and better at everyday interactions.

But eventually there can be a slippery slope from a social exchange that feels natural to an eerie valley scenario. There are also ethical concerns with robots that are too credible. Nevertheless, there are practical reasons for continuing to work in this field: making talking robots less creepy and more approachable by giving them the right human-like functions, experts think, will be especially helpful if one day integrates them into healthcare, hospitality or other service industries. .