Humans Beat AI? Study Finds Students Writing Without ChatGPT Produce Broader Ideas | Education and Career News
Last Updated:
Human-written essays consistently added new perspectives, experiences and combinations of ideas, expanding the collective pool as more essays were included.
According to a study, human essays increased overall diversity two to eight times more than GPT-4 outputs. (AI Image)
A recent study found that students who write essays without using AI tools such as ChatGPT generate a far wider range of ideas than those relying on large language models, raising concerns about the long-term impact of AI on creativity. According to a study published in Computers in Human Behavior: Artificial Humans, human-written essays consistently added new perspectives, experiences and combinations of ideas, expanding the collective pool as more essays were included.
The study added that individually, many GPT-4 essays appeared highly creative and sometimes matched or even exceeded human writing on certain creativity measures. However, the difference became clear when creativity was assessed collectively.
The research examined 2,200 college admissions essays across three separate studies. It compared essays written by real applicants between 2018 and 2022—before ChatGPT became widely available—with essays generated by GPT-4 using the same prompts, reported ET Education.
The researchers introduced a new metric called the “diversity growth rate,” which measures how much each additional essay contributes to the overall pool of ideas rather than evaluating pieces in isolation.
While AI-generated essays were often strong on their own, they tended to repeat similar themes, structures and expressions. As a result, they contributed far less novelty over time.
Human essays increased overall diversity two to eight times more than GPT-4 outputs, according to the study. This gap widened as the number of essays grew, highlighting what researchers described as a “homogenising effect” in AI-generated writing.
Attempts to reduce this effect—such as prompting GPT-4 to be more creative, adjusting model settings, and using chain-of-thought techniques—improved individual outputs but did not significantly increase collective diversity. Even newer versions of the model showed similar or stronger tendencies toward uniformity.
The researchers emphasised that AI can still be creative at an individual level. However, they warned that widespread reliance on the same tools could lead to an “algorithmic monoculture,” where ideas become increasingly similar.
About the Author
A team of reporters, writers and editors brings you news, analyses and information on college and school admissions, board and competitive exams, career options, topper interviews, job notifications, …Read More
Read More