Lai-Tze Fan: AI Storytelling out of the Box

This talk will explore changes to how we understand storytelling by generative AI systems (genAI), arguing that recent and seemingly divergent examples of genAI storytelling in electronic literature (e.g., Sofian Audry's AI-generated work for the sleepers in that quiet earth) and in information and media industries (e.g., Meta's interactive AI bots) serve as cultural and political warnings that we should be more careful about the "stories" we consume.

This title suggests that this talk will discuss the mechanisms that enable AI to tell stories. However, instead of addressing technical mechanisms, this talk will explore some cultural and political mechanisms of AI storytelling.

"Storytelling" here can be treated in the same way as the term "world building": both terms can be applied to creative practices in literature and games, but they can also describe the work it takes authors and designers to represent information so that it becomes immersive, engaging, entertaining, and persuasive--stories worth telling, seeking out, and sharing.

Insofar as genAI tools such as large language models represent information in closed-off systems (i.e., the models only know what they've been taught), they can be said to have "hallucinations" about the world and to tell us stories through limited knowledge and even incorrect assumptions. But what are some dangers of storytelling in the age of AI? For instance, how can incorrect assumptions reinforce stereotypes about groups of people? How can AI's pattern recognition be abused to fuel the spread of misinformation? Whose stories are told and who does not get their story told?

About Lai-Tze Fan

Lai-Tze Fan is the Canada Research Chair of Technology and Social Change, and an Associate Professor of Sociology & Legal Studies and English Literature at the University of Waterloo, Canada. She is Associate Professor II at the Center for Digital Narrative in the University of Bergen, Norway. Fan is the Founder and Director of The U&AI Lab at Waterloo, which uses research-creation methods for enhanced EDI outcomes in AI design. She is also on the Board of Directors of U Waterloo’s TRuST scholarly network, targeting misinformation and trust in AI, and on the Board of Directors of the Electronic Literature Organization.