In recent years, with the rise of generative AI tools such as ChatGPT, discussions about their impact on creativity and new ideas have grown increasingly fervent. Recently, researchers from the University College London School of Management and the University of Exeter conducted a new study specifically exploring the effects of generative models on creative writing. The researchers experimented with human-authored short stories, analyzing the impact on writers' creative output after exposure to story ideas generated by large language models (LLMs).
Image source: Image generated by AI, authorized service provider Midjourney
The study results indicate that while generative AI has indeed helped writers enhance the creativity, interest, and literary quality of their stories, it has also led to an increase in the similarity between stories. The focus of the study was on short story creation, where participants were required to compose an eight-sentence story based on randomly assigned themes.
The researchers measured the "novelty" and "practicality" of the stories in the experiment. Novelty refers to the degree of difference between the ideas and existing thoughts, while practicality refers to the actual application and relevance of the ideas. The researchers proposed two hypotheses: on one hand, generative AI could serve as a "springboard" for human creativity, providing potential starting points to enhance writing outcomes; on the other hand, generative AI might anchor writers to a specific idea, limiting their creative expression.
To this end, the researchers designed a two-stage online experiment. In the first stage, 293 participants wrote short stories based on random themes and were divided into three groups: human-only, those receiving one AI-generated idea, and those receiving five AI-generated ideas. The second stage involved 600 judges evaluating these stories for their novelty and practicality.
The results showed that writers who received generative AI ideas saw improvements in both novelty and practicality. Notably, the group that could access five AI ideas performed the best. Additionally, writers with lower baseline creativity scores benefited more, showing significant improvement. This indicates that generative AI can balance creative advantages among writers to some extent.
However, despite the enhancement of individual creative levels by generative AI, the researchers also found a significant increase in the similarity between stories based on AI-generated ideas. This raises concerns about homogenization of creative content. The researchers point out that if the publishing industry widely adopts generative AI, it could lead to a lack of uniqueness in overall creative works.
As more companies launch AI writing tools, these study results could have profound implications for future content creation.
Key Points:
🌟 Generative AI enhances writers' creative abilities, making stories more novel and practical.
📝 Stories based on AI ideas show increased similarity, potentially leading to content homogenization.
📚 The study results warn of the potential risks of widespread use of generative AI, affecting the uniqueness of future creations.