Laughing matters

Hana Holubec Chris Robinson

Laugh out loud. Bust a gut. Crack a witticism. Can artificial intelligence (AI) and social robots be programmed to understand the complexities and nuances of human humour and laughter? Will robots of the future understand they should not laugh at a funeral or the difference between slapstick and satire?

These are the sorts of questions that tickle the mind of York University PhD Candidate Hana Holubec in the Science & Technology Studies program in the Faculty of Graduate Studies.

“That’s my fascination. How do you take something so absolutely dynamic and complex and nuanced, and render it technological, which operates on binaries that require categorizations and taxonomies. It’s this moving target that changes across time and culture. One of my favorite things about laughter is that it defies classification,” says Holubec.

A trainee at the York-led Connected Minds, which explores human-technology interactions and their societal impacts, Holubec’s research looks into the development of AI algorithms to imitate humour and laughter in social robots. She does this work under the supervision of Glendon Campus Associate Professor Alison Harvey of the Department of Global Communications and Cultures, director of the Institute for Research on Digital Literacies.

“Laughter can engender social cohesion, but what type of laughter is being prioritized in AI or social robotic development and what types of laughter are being ignored or erased? That’s my big interest."

“There are aspects of humour that are relatively easy to render technologically. Joke generation has a computational quality and can be quite formulaic, especially wordplay or knock-knock jokes.” But that is only one aspect of humour.

One of the major focuses of social robotics research is programming humour beyond puns and basic jokes, and mimicking human laughter, body movements and facial expressions.

“Social robotics and communicative AI research with the use of laughter is intended to make the user feel like they are interacting in a natural way, in a very human-like way,” says Holubec, who is also a comedy writer and an arts-based instructor within the disability community.

There are ethical and moral concerns. As with any AI algorithms, inequities or harmful class, gender and racial stereotypes are at risk of being propagated in this fast-growing field.

Holubec’s most recent research – Laboratory Laugh: the production of laughter in the ERICA project – studies how researchers in Japan are incorporating AI and humour in their android robot Erica, which currently titters demurely. But who decides how Erica, or any other social robot, laughs and at what?

“Laughter can engender social cohesion, but what type of laughter is being prioritized in AI or social robotic development and what types of laughter are being ignored or erased? That’s my big interest,” says Holubec.

“Within my project at Connected Minds, what I really want to work on is developing a critical humour and laughter database, which anyone who is working on the development of communicative AI and social robotics could access.”

Currently, programmers and developers are drawing from neuroscience, psychology and linguistics, but not feminist and historical methodologies, literature and critical race theory, which Holubec says is an issue. “There’s a very rich breadth of information within those disciplines on humour and laughter that could help mitigate the socio-cultural effects that come with the computational flattening of laughter into AI and robotics.”

Although not the primary focus of her research, an additional concern is the potential for harm. “If the robot laughed a really big, mirthful laugh when a user was telling a disappointing story, this could have a detrimental effect. Essentially, they would feel laughed at.”

And that is no laughing matter.