AI and Narrative: Can an Algorithm Tell a Story?

Language Models as Narrative Machines

ChatGPT, GPT-4, Claude — large language models (LLMs) — are, essentially, machines for generating text, trained on vast arrays of human narrative. They can write a novel, a screenplay, a poem, a narrative report. But is this truly “narrative” in the full sense?

Peter Brooks defined narrative through “narrative desire”: a text driven by the desire to reach an end, to find resolution. This assumes intention, purpose, meaning. An LLM has no desires — it predicts the next token based on statistics. Does this create meaning — that is an open philosophical question.

From a pragmatic point of view: AI creates texts that function as narratives — eliciting emotions, holding attention, providing a sense of meaning. Readers respond to them as to narratives. Whether this constitutes narrative is a question about the minimal conditions for narrativity.

Algorithmic Recommendations as Narrative

YouTube, Netflix, TikTok — algorithms recommend the next video, the next episode, the next post. This creates a personalized narrative: each user experiences a unique “story” of content assembled by the algorithm.

But this narrative is not optimized for meaning — but for engagement. The algorithm does not think “what will help this person grow?” — it thinks “what will keep them on the platform longer?” This is a structural contradiction: the algorithmic narrative serves the interests of the platform, not the user.

The “filter bubble” effect (Pariser): the algorithm creates a narrative that confirms already existing beliefs. This is a narrative without conflict, without challenge, without growth — a narrative as a drug, not as a tool.

Question for reflection: Your information flow in social media is a narrative created by the algorithm. Who is the “author” of this narrative? Whose interests does it serve? How might you become the “author” of your own informational narrative?