Stories are a fundamental part of human experience, conveying knowledge about navigating the world, according to researchers meeting at the Santa Fe Institute (SFI) from December 10–12. The gathering brings together experts from fields such as computer science, folklore, physics, marketing, cognitive neuroscience, economics, mathematics, and psychology to develop new ways of understanding stories.
“Stories are everywhere,” said SFI External Professor Peter Dodds of the University of Vermont. “Even sports, or the differential equations that describe fluid dynamics, are kinds of storytellers. Stories that people tell and retell almost always involve characters and events that are connected and unfold over time, wrapped around essences of power, danger, and survival.”
Dodds is co-organizing the working group titled “Towards a Data-Driven Science of Stories.” The aim is to connect different scientific approaches for analyzing stories across cultures and eras.
“We want to illuminate the spectrum of stories across time and cultures, just like other fields have found spectrums of stars, species, or words,” Dodds said.
With advancements in artificial intelligence—such as large language models capable of processing vast archives—it might seem straightforward to identify patterns in stories. However, researchers argue that a true science of storytelling requires more nuanced tools drawn from complex systems research.
“Explaining a joke kills the humor, and we don't want to make stories dull by studying them. Our working group will explore new explanatory models and computational tools, drawing from complex systems science, to develop a science of stories that honors what a story is, and doesn't reduce it to a bag of words,” said Sam Zhang, statistician at the University of Vermont and recent SFI Applied Complexity Postdoctoral Fellow.
Zhang co-organized the initiative with Dodds; Samsun Knight—a novelist and marketing professor at the University of Toronto—and Juniper Lovato from the University of Vermont also contributed.
The participants plan to devise methods for mapping out story plots using visual representations. They aim to show how characters interact within networks over time and create datasets that allow researchers to identify core features common among different narratives.
“How do we take 100,000 stories, computationally turn them into temporal networks of characters, and then have a whole bunch of species to look at?” Dodds asked.
Understanding stories in this way has broad applications beyond entertainment—including analyzing propaganda’s role in shaping national narratives or viral marketing strategies.
