Getting AIs To Understand Humor May Improve Consumer Experience

While Siri and Alexa can tell jokes, they don’t know why humans find things funny. But a Purdue University scientist hopes solving that problem will open new doors for device creators.

Phones can already parrot back things they’ve been told are funny, like this joke from Siri:

“What’s the best method to find out how heavy your red hot chili pepper is? Give it a weigh, give it a weigh, give it a weigh now.”

But Siri doesn’t know why people might find that funny. Purdue University assistant professor Julia Rayz says examining the complexities of humor is one of the hardest – and potentially most rewarding – problems machine learning experts have yet to solve.

“So it’s almost an ultimate test in understanding of text or speech,” she says.

Rayz says extending the conversation between humans and AIs is a competition – and the prize is added attention from consumers.

“Interestingly enough, so far humor is one of the ways of keeping a human interest,” says Rayz. “Not if a computer is telling a knock-knock joke, but if something that’s a little but more involved, you kinda of go, ‘oh wow, you can do it alright. Let me poke around a little bit more and see what I can do with it.’”

And if computers understood humor, their manufacturers could make money.  That’s because a more natural conversation pattern makes users engage longer with their devices and helps companies better target customers using information the devices collect.

Purdue University professor Muhammad Rahman studies technology usage and says if a device can understand the nuances of humor, its developer could better tailor services and goods to a consumer.

“So if it doesn’t understand those emotions, it cannot learn over time of what my preferences and how I make judgement based on the information I am getting from this tool,” says Rahman.

Rayz says humor goes beyond the simple logic that technology is presently capable of comprehending, and dives into complexities including understanding of background information, tone and sarcasm.

“Until we incorporate all of it together as a single piece, I don’t think we’ll be able to say oh we made it or no we didn’t. So it very much has to play together all of it,” says Rayz.

These challenges researchers, including Rayz, still have a ways to go before your personal assistant Siri or Alexa, might be able to fully understand humor.