Artificial storytellers

Bibliographic Data

ID19234615
AuthorsDaniel Munro (0000-0003-4511-7037, Boston University, corresponding author)
Year2026
Pages1-27
Publication date2026-06-05
Peer ReviewedYes
Open AccessNo
TypeARTICLE
VenuePhilosophical Psychology (JOURNAL)
Journal identifiersISSN: 0951-5089 • E-ISSN: 1465-394X
PublisherInforma UK Limited (PUBLISHER • GB)
DOI10.1080/09515089.2026.2681784
OpenAlexW7163695976
LanguageEN
References cited50

Recent large language models (LLMs), such as GPT or Claude, can produce texts that look remarkably like human-authored stories. Scholarly and popular discussions often uncritically assume that these texts count as genuine stories or narratives. However, this paper steps back from this assumption to ask more fundamental, conceptual questions: When current LLMs are prompted to write stories, do the texts they output count as genuine stories? If not, are LLMs capable, in principle, of producing genuine stories? I consider these questions in light of my preferred account of the nature of stories. On this account, stories are by definition the product of a particular type of speech act: namely, the speech act of storytelling. I argue that, due to how they’re trained, recent commercially available LLMs are incapable of storytelling, so are incapable of writing genuine stories. Nevertheless, I argue that LLMs may in principle be capable of storytelling if trained the right way

Ask price · Narrative · Story telling · Storytelling · Artificial Intelligence in Games · Digital Games and Media · Digital Storytelling and Education

Tools

Open DOI