Why generative AI is widening the gender gap in academic research

Why generative AI is widening the gender gap in academic research

A new study found that male researchers use AI tools more effectively and gain a productivity advantage over their female colleagues. Can targeted interventions close the gap?

Generative artificial intelligence (AI) is driving productivity gains in several fields, including science. However, the impact appears to be uneven, benefiting male researchers more than their female counterparts. A recently published study inPnas nexusHighlights this growing inequality.

introduction

Generative AI is increasingly being integrated into research workflows, supporting scientists with data collection, literature review and analysis. By automating routine tasks, AI allows researchers to focus on innovative studies. In some cases, AI has enabled rapid production of research papers within an hour, improving both speed and quality.

Given these advantages, generative AI is becoming a standard tool in academic research. In fact, 80% of Nature readers say they have used Chatgpt or similar tools at least once. However, adoption varies significantly, influenced by socio-demographic factors, job satisfaction and workplace culture.

This disparity means that while some researchers are seeing significant productivity gains, others are lagging behind, exacerbating existing inequalities in science.

Both anecdotal evidence and survey data suggest that men are more likely than women to adopt generative AI. As a result, male researchers are able to produce more publications and accelerate their career advancement while putting their female colleagues at a disadvantage.

About the study

The study examined how ChatGPT affects research productivity across genders through two separate analyses.

Study 1: Analysis of research performance

The first analysis focused on preprints uploaded to the Social Science Research Network (SSRN) between May 2022 and June 2023. SSRN, a large open access repository, provided a rich dataset for assessing productivity trends. The researchers used a difference-in-difference (DID) approach to measure gender differences in research performance.

Initially, there was no observable change in productivity, likely due to the time required for researchers to become familiar with AI tools. As AI adoption increased, male researchers showed a relative productivity increase of 6.4% compared to their female colleagues. Specifically, men were 0.0004 more likely than women to upload at least one preprint per month.

This gender gap increased by 57%, increasing from 0.007 to 0.011 probability difference in research performance. To ensure an increase in AI-related papers, researchers have explicitly discussed publications about ChatGPT. The gap persisted, confirming that AI adoption is indeed driving inequality.

Further analysis that took co-authorship and individual contributions into account strengthened these results. In particular, the quality of research – interspersed with abstract views – has been made consistent, indicating that AI use has strengthened output without compromising rigor.

The productivity gap was most pronounced in countries where ChatGPT is widely distributed and used, such as the US, Australia and Spain. This correlation highlights the role of AI in reinforcing existing gender gaps.

Study 2: Attitudes towards AI

The second part of the study examined researchers' attitudes and usage patterns regarding generative AI. The results showed that men used AI tools more frequently and for longer durations than women.

Male researchers also reported greater efficiency gains and were more likely to recommend AI tools.

Importantly, these productivity differences were associated with usage patterns rather than inherent gender characteristics. The more researchers engaged in AI, the greater the efficiency benefits they experienced.

Conclusions

Both male and female researchers have access to generative AI, but men use it more effectively to increase their research performance. This discrepancy appears to be due to differences in attitudes and behaviors towards technology adoption.

The adoption of generative AI may exacerbate existing inequalities related to funding, leadership roles, access to research facilities, and evaluation metrics.

To prevent this technology from widening the gender gap, it is crucial to actively promote and train all researchers – especially women – to integrate AI into their workflows.

Without proactive measures, women researchers risk falling behind in an increasingly AI-driven academic landscape.

Sources: