Can Artificial Intelligence Detect Gender Bias?

Artificial Intelligence can be used to detect gender bias in scientific research.

The application of ChatGPT in identifying gender-based discrepancies among peer-reviewed scientific articles has become a significant subject of discussion. Peer-reviewing is an integral process of scientific research that ensures that the information to be shared is accurate and of high quality.

The subjective nature of peer review can be questioned, however, especially regarding gender bias against authors. Can artificial intelligence, such as ChatGPT, be used to identify gender bias through the language processing of peer reviewed scientific articles? In this 2023 study conducted by Verharen, artificial intelligence was used to analyze language and its tone, such as a measure of politeness, to determine the presence of gender bias amongst reviewers while they are reviewing an article.

In this study, 200 neuroscience papers from Nature Communications were surveyed. A score of the language was calculated using ChatGPT 3.5. The peer review was scored on both sentiment and politeness on a scale from -100–100, and a mixture of statistical methods were used to test for consistency and statistical significance.

According to the data, authors with traditionally women names received significantly more impolite reviews; however, this was solely based on an overall lower politeness score rather than on sentiment. The reviewer’s institutional affiliation appeared to have no notable impact on the politeness or sentiment of the review.

The implications of this research are profound, suggesting that gender biases, whether overt or subtle, may pervade even the most objective scientific domains. The use of AI in this context not only shines a spotlight on the issue but also offers a potential tool for mitigating such biases in the future.

Join our mailing list to receive new WinSights articles on research-backed resources for inclusive science.

Study Details

Sample size(s): 200 articles

Participants: N/A

Design: Correlational Study

Reference:

Verharen, J. P. H. (2023, October 4). CHATGPT identifies gender disparities in scientific peer review. eLife. https://elifesciences.org/reviewed-preprints/90230

Summarized by WinSights team member Alisha Damji

Edited by: Margie Christ, Bilal Rashid