Study of 370,000 Essays Finds AI Constricts Creative Thinking, Masking It With Polished Language
Updated
Updated · The New York Times · May 27
Study of 370,000 Essays Finds AI Constricts Creative Thinking, Masking It With Polished Language
2 articles · Updated · The New York Times · May 27
Adam Green’s team found college-application essays written after ChatGPT’s release looked more vivid and polished but contained less original thinking.
The Georgetown-led research tracked more than 370,000 student essays over eight years and concluded that AI-assisted writing pushes ideas into a few familiar, homogenized categories.
Human judges still rated those post-ChatGPT essays as more “creative,” suggesting fluent language and elegant phrasing can hide weaker underlying ideas.
A separate study found human-written essays generated up to eight times more new ideas than AI-produced ones, reinforcing concerns that brainstorming itself is being outsourced.
Green argues that loss of creative thinking could leave students, workers and society less able to adapt, solve complex problems and produce genuinely new ideas.
Why does AI-assisted writing seem more creative to judges even when it contains far fewer original ideas?
As experts disagree, is AI a genuine threat to creativity or an essential tool for future innovation?