AI MAKES INDIVIDUALS MORE CREATIVE BUT DIMINISHES CREATIVITY OF GROUP, STUDY FINDS

Simultaneously offering great potential alongside a grave threat to creative work, AI is touted to change the creative industries profoundly – a new study from researchers at University College London and the University of Exeter based on LLMs to help short story writers finds that less naturally creative people within the group studied were able to write more original stories but, across the overall group, the output was less original.

Why AI and creativity matters

Creative work and the creative industries stand in the immediate firing line of generative AI’s text and image creation. Many creative workers, therefore, see their professions likely to be eroded in the name of savings. This new study, which suggests that collective originality is diminished by the use of generative AI, foreshadows a world of samey creative in the name of savings.

The trouble is that creative work isn’t the aspect of advertising work that has been getting more expensive – for that, look to media.

But there’s a bigger question at play here: the study finds that individuals benefit – and would therefore be more likely to use the technology – but that overall output will become less unique, if more professional looking.

What the study says

“We find that access to generative AI ideas causes stories to be evaluated as more creative, better written, and more enjoyable, especially among less creative writers,” the authors, Anil Doshi and Oliver Hauser explain in the paper, published in Science Advances. “However, generative AI–enabled stories are more similar to each other than stories by humans alone.”

“Less creative writers experience greater uplifts for their stories, seeing increases of 10 to 11% for creativity and of 22 to 26% for how enjoyable and well written the story is.”

But, interestingly, for already creative writers there is little effect from adding AI on the creativity scores on their stories. “Having access to generative AI does not affect high [-scoring creative] writers’ already good performance on these outcomes,” the paper states.



The work looked at creativity across two broad metrics: novelty and usefulness (for publication). Examining a dataset of hundreds of short stories written, a control group wrote unaided by AI; another group was able to consult GPT-4 for ideas and a couple of sentences; a final group was allowed to use as many as five story ideas.

Results

• Novelty increases: Writers in the Human with one GenAI idea saw an increase of 5.4% in novelty over writers without generative AI access. Writers in the Human with five GenAI ideas condition show an increase in novelty of 8.1%.


• Usefulness increases: Stories from writers with one GenAI idea were 3.7% more useful than the control, while stories from writers with five generative ideas were 9% more useful than the control. “Having access to generative AI “professionalizes” the stories beyond what writers might have otherwise accomplished alone,” the paper states.


• Sameyness: It appears that writers were anchored to the generative AI ideas given to them. Compared to human-only stories, those with one AI idea and those with five AI ideas are 5.2% and 5%, respectively, more similar to the generated ideas they were given.

“While there is huge potential (and, no doubt, huge hype) for this technology to have big impacts in media and creativity more generally, it will be important that AI is actually being evaluated rigorously — rather than just implemented widely, under the assumption that it will have positive outcomes,”

explained Oliver Hauser, a researcher at the University of Exeter in comments to TechCrunch.

Source: warc.com