AI is changing content creation, but at a cost. In an innovative study, researchers from the University of Exeter found that although AI can really improve certain aspects of humans’ creativity, this could eventually contribute to an outright decline in the total diversity of content.
This runaway success was published in the journal Science Advances and shows how AI can go both ways in creative fields. A new study reveals that AI overshadows humans when creating more novel story ideas and an engaging factor to please audiences and publishers alike. However, this effect does not hold in general. On the other hand, it proves less effective at enhancing the work of already creative writers and seems to homogenize output, reducing overall diversity.
Image Source: Envato
Study Overview
The study, conducted by a research team from the University of Exeter Business School and its Institute for Data Science and Artificial Intelligence, in collaboration with the UCL School of Management, divided 300 participants into three groups.
The exercise was to create a short ‘micro story’ targeting young adults, using different degrees of help from AI. All in all, there were three different groups: one working without the guidance of AI, another that used ChatGPT for only one prompt consisting of three sentences, and a third that could choose from up to five AI-generated ideas.
The results were striking. Participants who received the most AI assistance improved the most. Their stories were rated for 8.1% more novel and 9% higher in overall quality, which was written without AI as compared to those. Also, stories created with AI assistance are found to be better written, more enjoyable, and less boring.
Creativity for the Less Creative
One of the most striking elements about this particular line of logics has to do with AI’s effect on less creative writers. Results regarding the yardsticks of intrinsic creativity were insignificant in the Divergent Association Task but found that those with lower DAT scores did benefit from AI assistance; that is, their stories improved in novelty by 10.7 percent and in usefulness by 11.5 percent—effectively evening out the playing field with more creative writers.
It has also meant that AI serves as a really powerful democratizing tool for those folks wrestling with creative tasks.
One downside would be the loss of collective novelty.
A note of caution, however, was also sounded in this study. This study checked the similarity between stories with OpenAI’s embeddings API. According to this measurement criterion, they witnessed an increase of 10.7% in similarity among AI-aided stories when compared with those written without AI. This homogenization could indeed be considered as even while individual creativity may flower, plural originality of content suffers.
The problem was highlighted by Oliver Hauser, a professor of economics at the University of Exeter Business School. “Our results show how generative AI can enhance individual creativity and remove disparities in creative ability,” he said. “However, there is a risk that, should AI at large be used for creative writing, we will see a reduction in the uniqueness of stories.” In fact, Anil Doshi, an assistant professor at University College London’s School of Management, had similar concerns. “The more the publishing industry relies on AI, the more alike these stories could become, reducing the diversity and novelty overall published.”
The Future of Artificial Intelligence in Creativity
These results give a more subtle view of AI in creative industries: while it does have the potential to improve individual creativity, especially for the worse off, this is at the cost of lower collective novelty. In a time when AI further penetrates a decent number of creative fields, stakeholders need to consider these implications to balance benefits and drawbacks effectively. This makes the study like a beacon for both the potential of AI and a precautionary tale on its limitations. To this end, Professor Hauser put it more appropriately: “Despite the enhancement effect of generative AI on individual creativity, there may be a cautionary note if generative AI were adopted more widely for creative tasks.”
Leave A Comment