In a study on AI and creativity, researchers from the University of Exeter Business School and the UCL School of Management looked at the two ways that AI affects creativity—individual and group. While generative AI models, such as GPT-4, have made it easier and faster to produce creative content, they also tend to homogenize and flatten the overall diversity of creative works. This research, published in Science Advances, explores how AI influences human creativity by studying the use of GPT-4 in writing short stories.

Methodology

The researchers conducted a study involving 293 participants recruited through Prolific. The participants were instructed to provide ten varied words to test their inventiveness. They were then asked to write an eight-sentence story for young adults on one of three topics: an adventure in the jungle, on the open seas, or a different planet.

Group Division

Participants were randomly divided into three groups:

• Group 1: Relied solely on their ideas.
• Group 2: Had the option to receive a single-story idea from GPT-4.
• Group 3: Could elect to receive up to five-story ideas from GPT-4.

AI's Influence on Creativity

The study found that 88.4% of participants who could use AI assistance took advantage of it. The creativity of the stories was then evaluated by a separate group of 600 reviewers based on stylistic characteristics, novelty, and usefulness. Novelty refers to a story’s originality, while usefulness reflects the potential for the story to be developed into a book or other publishable work.

Impact on Less Creative Writers

The results showed that AI significantly improved the output of less creative writers, who benefited the most from the AI-generated ideas. These writers' stories were evaluated as more creative than those of writers who did not use AI assistance. However, for writers who were already considered creative, AI made little difference in the quality of their stories.

Homogenization of Content

Despite the individual creative boost provided by AI, the study highlighted a critical downside: the homogenization of creative output. Stories influenced by AI were more similar to each other than those created independently by humans. This is because AI models can only generate content based on the data they have been trained on, leading to less distinctive and more uniform outputs.

Implications for the Publishing Industry

The findings suggest that if the publishing industry widely embraced generative AI, the books and other media produced could become more homogeneous. Since AI models are trained on the same corpus of data, their outputs tend to be less diverse, potentially reducing the variety and richness of creative works available to the public.

Conclusion

The study provides valuable insights into the role of AI in creativity. It demonstrates that while AI can enhance the creativity of individuals, particularly those who are less creative, it simultaneously reduces the diversity of collective creative output. This dual impact raises important considerations for the use of generative AI in creative industries. It highlights the need for further research into methods for preserving and promoting originality in AI-assisted creative processes.

Source: MIT Technology review

Article: Science advances

Image source: Unsplash