acecreativity and novelty of short stories produced with LLM-generated ideas
Source article: Generative AI enhances individual creativity but reduces the collective diversity of novel content
Creativity is core to being human. Generative artificial intelligence (AI)-including powerful large language models (LLMs)-holds promise for humans to be more creative by offering new ideas, or less creative by anchoring on generative AI ideas. We study the causal impact of generative AI ideas on the production of short stories in an online experiment where some writers obtained story ideas from an LLM. We find that access to generative AI ideas causes stories to be evaluated as more creative, better written, an…
Contested: both sides are scored from claims and sources, not community votes.
Today's short stories analyzed; an informal encyclopedia of short story art as exemplified in contemporary magazine fiction for writers and students by Neal, Robert Wilson, 1871-1939. Public domain
In an online experiment reported July 12 2024, researchers gave some writers LLM-generated story ideas and had independent evaluators rate the resulting short stories. By that date they observed that access to AI ideas caused higher ratings for creativity, writing quality, and enjoyment, especially for less creative writers.
The same experiment also found a collective cost: AI-enabled stories were more similar to each other than human-only stories, pointing to a narrower scope of novel content. This matters because it suggests a tradeoff between boosting individual creative output and preserving diversity of creative work, but it remains uncertain whether the pattern holds beyond short stories or outside the experimental setting.
- Online experiment tested causal impact of LLM-provided story ideas on short story writing.
- Stories with AI ideas were rated higher on creativity, writing quality, and enjoyment.
- Benefit was concentrated among less creative writers.
- AI-enabled stories were more similar to each other than human-only stories.
Writers who received LLM-generated story ideas produced short stories that were rated as more creative, better written, and more enjoyable, with the largest improvement for less creative writers.
Writers who used LLM-generated story ideas produced short stories that were more similar to each other, reducing collective diversity and producing a narrower scope of novel content.
The rundown
The study randomly gave some writers access to LLM-generated story ideas and compared their output to writers without AI help, measuring human evaluations of creativity, writing quality, and enjoyment.
While individual ratings improved, similarity analysis showed AI-assisted stories clustered together more than human-only stories, framing the tradeoff as a social dilemma where writers are individually better off but collective novelty shrinks.
Results are from an online experiment on short story production using LLM story ideas, so may not generalize to other creative domains or real-world writing workflows.
Sources
- Peer-reviewedScience Advances2024-07-12
How should this claim be treated?
The debate