We live in an interesting technological era with questions about the morality and legality of using AI for creative content. Researchers have attempted to quantify what we feel when reading AI-generated text. Essentially, they ask if something can be too neat, too clean, or too to the point.
To answer this, imagine a story about a woman returning to her childhood home after twenty years. She finds a letter hidden behind a loose floorboard revealing a family secret. She confronts her mother, finally understands why her childhood was complicated, and learns a lesson about forgiveness.
At first, the plot seems reasonable with logical motivations and a tidy ending. However, a second version opens at the mother's funeral, jumps backward twenty years, then forward three days, then sideways into an irrelevant memory. The letter contradicts earlier lessons, the daughter makes a decision open to interpretation, and the ending refuses to explain its meaning.
This raises the question: can we say for sure which version sounds more human? Many human writers love tidy endings, and many AI stories are strange. But what if these intuitions reflect measurable patterns? The paper "Story Scope: Investigating Idiosyncrasies in AI Fiction" moves beyond surface features like vocabulary or repetitive sentence structures. Instead, it investigates narrative decisions shaping a story.
The authors examine who drives the plot, event predictability, subplots, narrator explanation, time jumps, and moral straightforwardness of characters. These features describe a story's architecture, which is harder to change than word choice. The research quantifies creativity by turning stories into data using feature vectors representing word frequency, sentence length, and punctuation.
However, if a model changes its vocabulary, feature distributions shift and decision boundaries fail. Story Scope asks if we can look beyond surface features to the narrative structure itself.
Source: Towards Data Science · Summarized by HeadlinesBriefing