Summary: By developing a novel automated evaluation framework called CASPER, researchers analyzed thousands of human-authored and machine-generated stories across eight distinct axes of literary theory. The findings conclusively demonstrate that AI models systematically strip away one of the defining traits of memorable fiction: mystery.
While human authors routinely embrace narrative ambiguity, leave profound questions unanswered, and allow characters to remain beautifully contradictory, AI models uniformly “play it safe.” They rely heavily on flat, predictable archetypes and force storylines into artificial, perfectly tidy resolutions.
Key Facts
- The safe Resolution Bias: Lead author Anneliese Brei notes that AI systems possess an inherent mathematical bias to wrap up narratives cleanly. They aggressively resolve internal conflicts, answer every mystery, and ensure characters fit perfectly into their designated story arcs by the final page.
- The Illusion of Scale: A critical revelation of the study is that scaling up parameter size does not solve the problem. Massive, state-of-the-art flagship LLMs generated characters that were just as flat and archetypal as those produced by significantly smaller, less complex models. The deficit is rooted in how models understand storytelling, not processing power.
- Embracing the Unresolved: When analyzing human writers, the CASPER framework revealed a high comfort level with chaos. Human-authored fiction regularly leaves characters unresolved, morally gray, or fundamentally open to interpretation, the exact structural ambiguity that makes a story stick with a reader.
- Evaluating Character Evolution: The study systematically mapped character behavior against eight core dimensions of literary theory, analyzing the precise transition from hyper-exaggerated caricatures to realistic individuals, alongside tracking whether characters genuinely evolve or simply follow a script.
- The CASPER Benchmark: Beyond exposing creative limits, CASPER functions as a vital, standardized benchmarking framework. It enables AI developers and creative studios to evaluate whether upcoming, next-generation models are genuinely advancing narrative depth and character complexity rather than simply becoming more grammatically fluent.
- A Takeaway for the Writing Community: For authors leveraging AI as an interactive brainstorming assistant or co-writer, the UNC study offers a definitive warning: letting a machine dictate character development risk homogenizing the narrative, making a human touch essential to reintroduce contradiction, subvert expectations, and deliberately inject uncertainty.
Source: UNC Chapel Hill
Researchers at the University of North Carolina at Chapel Hill have found that while artificial intelligence can spin increasingly convincing stories, its characters may still lack one of the qualities that make human-written fiction memorable: mystery.
As AI writing tools become more common in publishing and entertainment, Carolina researchers wanted to understand whether the characters created by these systems are as varied and nuanced as those crafted by human authors. Their findings suggest that, despite advances in technology, AI still tends to rely on familiar patterns.
The study examined how characters in stories generated by AI compare with those written by people. Drawing on ideas from literary theory, the researchers analyzed eight different aspects of character portrayal, including whether characters seem realistic or exaggerated, whether they evolve over time and whether they remain mysterious or fully understood by the end of a story.
To do this, the team developed CASPER, an automated framework that evaluated thousands of stories and measured character traits in ways that had never before been systematically applied to AI-generated fiction.
“We found that AI models tend to ‘play it safe’ with their characters, in the sense that they wrap up storylines neatly,” said Anneliese Brei, a graduate student in computer science at UNC-Chapel Hill and lead author of the study.
“Human writers, on the other hand, are sometimes more willing to leave questions unanswered and let characters remain mysterious. That difference matters because ambiguity is often what makes a story linger with a reader.”
The research comes at a time when AI tools designed specifically for creative writing are gaining traction. Platforms such as Sudowrite and Squibler can help draft novels, while AI is increasingly being used in film and television to generate script outlines and dialogue. Surveys have also shown that many fiction writers now incorporate AI into some part of their creative process.
Their analysis revealed that AI-generated characters often lean more heavily on recognizable archetypes and tend to arrive at tidy resolutions by the end of a story. Human writers, by contrast, appeared more comfortable allowing characters to remain unresolved, contradictory or open to interpretation.
“One of our most surprising findings was that bigger and more powerful AI models don’t necessarily create more varied characters than smaller ones,” said Nicholas Sanaie, an undergraduate student in computer science at Carolina and co-author of the study.
“That tells us the challenge isn’t just about scale. It’s about how these models understand storytelling itself.”
CASPER gives researchers, developers and creative professionals a way to benchmark whether newer AI systems are actually improving portraying complex characters rather than simply becoming more fluent writers. It could also guide the development of future storytelling tools that better support creativity and narrative depth.
“As more people collaborate with AI to write novels, screenplays and other creative works, we need ways to understand both what these systems do well and where they fall short,” said Snigdha Chaturvedi, associate professor of computer science at UNC-Chapel Hill and senior author of the study.
“CASPER gives us a lens for evaluating character depth and diversity, which can ultimately help developers build storytelling systems that better reflect the complexity of human experience.”
For writers experimenting with AI, the findings offer a practical takeaway: AI may be an increasingly capable creative partner, but the most compelling stories may still require the distinctly human willingness to embrace uncertainty, contradiction and characters who don’t fit neatly into familiar molds.
Key Questions Answered:
A: This “neatness” bias is built directly into how large language models are trained. AI models are trained on mathematical probabilities to predict the most satisfying, logical next word based on massive mountains of internet data. When building a story, the model naturally optimizes for high-probability paths, which means it leans toward predictable structures and tidy resolutions. It is programmed to provide answers, not sit with discomfort. Human life, however, is full of loose ends and contradictions, qualities that human writers deliberately capture, but AI views as statistical anomalies to be smoothed over.
A: CASPER is an automated computational linguistics framework designed by the UNC team to turn abstract literary theory into measurable data. It analyzes thousands of text blocks, tracking how characters are described and how their actions evolve from the beginning to the end of a narrative across eight specific dimensions. To measure “mystery” or ambiguity, CASPER scans whether a character’s internal motives are entirely explained by the narrator by the climax, or if their behaviors remain beautifully unmapped, contradictory, and open to multiple interpretations, quantifying the exact invisible traits that separate flat caricatures from unforgettable literary icons.
A: Not at all. The researchers view AI as an incredibly capable creative partner, but one that requires a firm, human hand at the wheel. AI is fantastic at helping writers brainstorm plot outlines, cure blank-page syndrome, or quickly flesh out simple background descriptions. The real lesson here is that you cannot delegate the soul of character development to a machine. If a novelist lets an AI write their main characters unchecked, those characters will inevitably turn into flat, safe clichés. The true magic of storytelling still requires a human writer who is willing to step in and intentionally mess up the code by adding unresolvable flaws, messy contradictions, and a healthy dose of genuine mystery.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- Journal paper reviewed in full.
- Additional context added by our staff.
About this AI and creativity research news
Author: Gabriella Neyman
Source: University of North Carolina at Chapel Hill
Contact: Gabriella Neyman – University of North Carolina at Chapel Hill
Image: The image is credited to Neuroscience News
