IMPORTANT
Peiqi "Patrick" Sui's talk on "Confabulation: What Could LLM Hallucinations Do For Storytelling?"
Thursday, November 14, 2024 · 11:30 AM - 12:45 PM
Abstract:
Are hallucinations always bad? Most of NLP research presumes a normative
stance that they are, but it overlooks the cognitive and communicative
affordances of a type of particularly story-like hallucinations (which
we'll call confabulations). Consider two general categories of LLM
applications: using them as tools, or interacting with them as viable
cultural agents. The two have very different training objectives in
terms of the tradeoff between factuality and alignment with the human
behavior of storytelling, and when it comes to ensuring the latter, LLMs
that could effectively confabulate would be especially useful. For
instance, confabulations could enable LLMs to perform speculative
narration and address omissions in history resulting from social
injustice, in the hope of enacting what literary theorist Saidiya
Hartman calls "critical fabulation" at scale, and giving interactive
storytelling a wider social impact.
--
Patrick Sui is a second-year PhD student in English at McGill
University, advised by Richard Jean So. He mainly works in digital
humanities & cultural analytics, and spends most of his time
thinking about how literary studies could uniquely contribute to AI
research about language. His current research topics include benchmarks
for close reading & interpretive reasoning, modeling close reading
behaviors with information theory, knowledge-grounded style transfer for
co-creative systems, AI literacy & writing pedagogy, and all kinds
of computational literary theory.