In the Garden of Forking Paths: Mapping and Making Sense of Storytelling at Scale

Co-sponsored by the Berkeley Institute for Data Science and the School of Information

Humans tell stories for many different purposes — to drive social movements, build trust, make sense of the world — and increasingly, we tell stories with and about AI. 

In this talk, I argue that instead of building benchmarks for AI-generated stories, we need descriptive, analytical methods for exploring and mapping narrative space. I present a series of projects that chart where storytelling happens and how it is changing by locating stories across online communities, characterizing narrative content in web-scale LLM pretraining data, and tracing how people iteratively generate and edit fiction in real user-chatbot conversations. This work reveals narrative axes that organize pretraining data, “solipsistic reader-writers” who obsessively regenerate the same stories in loops with chatbots, and edit trees representing the “forking paths” that people trace when exploring narrative space. The future study of storytelling needs more mixed-methods and cross-disciplinary collaboration, open and auditable data and models, and room  to craft our own stories.

The Cultural Analytics Series is a series of lunchtime talks and workshops highlighting research that focuses on the data-driven analysis of cultural phenomena.