Generative AI tools, like ChatGPT, have been gaining traction in recent years, sparking debates about their influence on creativity and the generation of novel ideas. To shed light on this matter, researchers from the University College London School of Management and the University of Exeter conducted a study that explores the effects of generative models on creative writing.
The study focused specifically on short story writing and examined how access to story ideas generated by large language models (LLMs) affected the creativity of human writers. The results of the study were nuanced. While generative AI led to stories that were rated as more creative, engaging, and better written, it also resulted in stories that were more similar to each other.
To measure the impact of generative AI on creative writing, the researchers asked participants to write an eight-sentence short story about a randomly assigned subject. Creativity was measured based on novelty and usefulness. Novelty refers to the extent to which an idea departs from the status quo or expectations, while usefulness refers to the practicality and relevance of an idea.
The researchers proposed two possible ways in which generative AI can affect creative writing. On one hand, it can serve as a “springboard for the human mind,” providing potential starting points for different storylines or helping writers overcome writer’s block. On the other hand, it may restrict the variability of a writer’s own ideas from the start, inhibiting the extent of creative writing.
To investigate these aspects of creativity, the researchers designed a two-phase online experiment. In the first phase, participants were divided into three groups: one group received no assistance or input from generative AI, another group received a three-sentence story idea generated by OpenAI’s GPT-4, and the third group could request up to five ideas from GPT-4.
The study found that access to generative AI ideas improved both novelty and usefulness in the stories. Interestingly, the group with access to five AI-generated ideas showed the most significant improvement. Access to more ideas allowed writers to break away from their initial assumptions and explore a wider range of possibilities. Additionally, writers who scored lower on baseline creativity assessment benefited more from generative AI, as it acted as an equalizer that removed any disadvantage or advantage based on the writers’ inherent creativity.
Furthermore, evaluators found the AI-assisted stories to be more enjoyable, better written, and more likely to have plot twists. Access to generative AI “professionalized” the stories beyond what the writers might have accomplished alone.
However, the study also revealed a potential concern. Stories based on AI-generated ideas were found to be more similar to each other compared to those written by the control group. This raises concerns about the potential homogenization of creative content if generative AI becomes widely adopted. Writers may become anchored to specific generative AI ideas, leading to a loss of collective novelty.
The findings of this study hold significance as more companies offer AI-powered writing tools and organizations use LLMs to create mass content. If the publishing industry embraces generative AI-inspired stories, there is a risk that the produced stories would become less unique in aggregate and more similar to each other. While generative AI enhances individual creativity, caution should be exercised if it were to be adopted more widely for creative tasks.
In conclusion, generative AI tools have the potential to enhance creativity in writing by providing novel ideas and overcoming writer’s block. However, there is a need to strike a balance to ensure that creative content remains diverse and unique. As the web becomes filled with content that has similar distributions, it may influence the training of future language models. The impact of generative AI on creative writing is a complex topic that requires ongoing research and consideration.
