AI meets citizen science to unlock the nature of storytelling

A new project led by McGill University researchers seeks to understand one of humanity's oldest practices and most powerful tools—storytelling. From ancient oral traditions to modern-day literature and digital narratives, storytelling is an essential part of the lived experience that is not yet fully understood.

"The Lives of Literary Characters" is a first-of-its-kind initiative, harnessing artificial intelligence (AI) and the collective wisdom of readers worldwide to explore the question: why do we tell stories?

Just as our real-life social networks can reveal who we are as individuals, analyzing how characters interact in a story can shed light into the nature of storytelling across cultures.

"Characters are the scaffolding of great storytelling," explains Andrew Piper, a Professor in McGill's Department of Languages. Right now, it is possible to use AI models to detect who a character is in a story, however, it is far more complicated to understand if and how those characters are interacting.

The effort to solve that piece of the puzzle poses an unprecedented challenge, as there were over 1.5 million characters invented in English alone during the nineteenth-century, and today that number is magnitudes higher.

"The proliferation of stories today is astounding. There is simply no way to account for the breadth of human creativity without relying on AI. But these systems are often trained in biased or unknown ways. By enlisting the help of readers, we can build better, more transparent AI systems to surface the intricate patterns upon which human stories are based," said Piper, the project's director.

The research team is calling on volunteers to contribute to the project as citizen scientists. Through the Zooniverse web portal, readers answer questions about character interactions in short passages of contemporary fiction. The crowdsourced data will be used to train AI models to better understand literary characters. While much of AI development happens behind closed doors, the results of this project will be open to all.

"All of the data and models we generate will be open to the public, setting an important standard for transparent and inclusive AI. Our goal is not to build robot storytellers. We want to gain a deeper understanding of human storytelling," said Piper.