The research was led by Professor Karim Jerbi from the Department of Psychology at the Université de Montréal, with participation from renowned AI expert Yoshua Bengio, postdoctoral researcher Antoine Bellemare‑Pépin (Université de Montréal), and PhD candidate François Lespinasse (Université Concordia).
The study was published in Scientific Reports and represents one of the largest direct comparisons ever made between human and AI creative performance.
To assess creativity fairly across humans and machines, the team employed the Divergent Association Task (DAT), a widely‑used psychological test that measures divergent thinking, or the ability to generate diverse and original ideas from a single prompt. Participants, both human and AI, were asked to produce ten words that were as unrelated as possible, a task designed to reveal creative thought patterns.
The results were striking: some AI models, including versions of GPT‑4, outperformed the average human on certain creativity tasks.
“Our study shows that some AI systems based on large language models can now outperform average human creativity on well‑defined tasks,” explained Professor Jerbi. However, the team also found that the top creative humans still had a clear advantage, especially among the most creative 10 percent of participants, whose scores surpassed every AI model tested.
In addition to simple word association, the researchers evaluated performance on more complex creative tasks, including poetry (haiku), plot summaries, and short stories. Here, the AI systems sometimes matched or exceeded average human performance, but the gap between AI and the most imaginative humans widened further, reinforcing the idea that AI excels at structured creativity but does not yet replicate the depth and originality of human expression.
The study also explored how creativity in AI can vary based on technological settings. For example, adjusting parameters like temperature (which affects how predictable or adventurous the AI’s responses are) influenced the diversity of its output. This finding suggests that AI creativity is not fixed and can be guided by human input and design choices.
Researchers emphasise that while AI systems are now powerful tools for idea generation and exploration, they are more likely to augment human creativity rather than replace it. “Generative AI has, above all, become an extremely powerful tool in the service of human creativity,” stated Professor Jerbi.
Leave a Reply