Growing up, I was surrounded by remarkable women — my mother, my teachers, and later my colleagues. I have always been nurtured and inspired by these “wonder women”, whose intelligence and resilience became the guiding lights along my path. Over time, that experience led me to a broader question: can we meaningfully capture and measure the contributions made by women in academia?

The Gender Diversity Benchmark: Where HKUST Stands

Gender diversity in research is frequently discussed, yet measuring it in a meaningful manner remains a challenge. One widely used benchmark is the CWTS Leiden Ranking (2025), which evaluates gender diversity using authorship data from 2020–2023 and reports the proportion of female authors at each institution.

Among local peer universities, HKUST has the lowest proportion of female authorship (17.6%).

This figure serves as a good reference, but it also has limitations. Total authorship is fundamentally a lagging indicator. It reflects years (even decades) of historical hiring, promotion, and tenure decisions, capturing everyone from senior faculty appointed many years ago to short-term visitors. In other words, it tells us where we have been—but not necessarily where we are going.

To understand where HKUST might be heading, we need a forward-looking lens.

Why “First Authors” Matter

To gain a more forward-looking perspective, we shifted our focus to the first authors. In many disciplines, the first authors are PhD students and early-career researchers who lead experiments, analysis, and manuscript preparation. They represent the next generation of scholars currently being trained at HKUST.
 

Data and Methodology

Web of Science does not provide a direct gender indicator for authorship. To address this gap, we first identified first-author publications and then inferred gender based on full names using large language models (LLMs), drawing reference from the Nature Index methodology (Nature Index, n.d.). This approach aligns with the data source used in the CWTS Leiden Ranking, ensuring consistency while allowing a more granular analysis.
 

Dataset overview
 

• 9,520 publications indexed in Web of Science
• Coverage covered: 2021–2025
• Gender inference: Custom Python pipeline using LLMs GPT4o and Claude 3.5 Sonnet
• Confidence threshold: Greater than 80%
• Single-author papers excluded
• Final validated sample: 5,889 first-author records
   

Limitations

Name-based gender inference is inherently complex. Cultural naming conventions, particularly for East Asian names, where transliteration systems such as Pinyin can obscure gender cues and increase misclassification risk. Recent research has highlighted this challenge and proposes improved approaches for Chinese names, though limitations remain.

Although LLM-based methods demonstrate improved performance over traditional rule-based tools, the results should be interpreted as probabilistic estimates, not definitive classifications. Continued development of culturally robust inference methods remains essential for future bibliometric research.
 

The Trend: Breaking the 20% Ceiling

Despite year-to-year fluctuations in total research output, the proportion of female first authors at HKUST has shown a steady increase. In recent years, the figure has surpassed the 20% threshold.

This shift is modest but meaningful. It suggests that the pipeline of young female researchers is gradually widening — a critical signal for long-term institutional change.

See the Pen Gender by lee chung (@lee-chung) on CodePen.

(Interactive Chart: Yearly Trend) 

Representation Across Disciplines: Where is the Talent?

A breakdown by Web of Science Subject Categories highlights where research activity is concentrated—and where gender gaps remain most pronounced.
 

• The Volume Paradox: High-output STEM fields such as Engineering, Electrical & Electronic Engineering, and Computer Science dominate HKUST’s publication landscape. Female representation in these areas remains relatively low, typically around 14–15%.
• Scale Meets Diversity: More encouraging signals emerge in mid-sized fields (over 100 papers) such as Biochemistry and Environmental Sciences. These disciplines combine substantial research volume with 28–37% female representation, indicating that a healthy and diverse talent pipeline is already in place.

See the Pen Gender2 by lee chung (@lee-chung) on CodePen.

(Interactive Chart: Research Landscape Representation)

Research Impact: Quality vs. Quantity

Representation alone does not tell the full story. To assess research influence, we examined the Category Normalised Citation Impact (CNCI), where 1.0 represents the world average.

The findings are striking.
 

Key Highlights
 

• Humanities: Female-led papers achieved an average CNCI of 8.42, compared with 1.72 for male-authored papers.
• Governance and Public Affairs: In Law and Political Science, female First Authors recorded a CNCI of 4.66, indicating a strong concentration of high-impact scholarship.
• Social Sciences: Female authors achieved an average impact score of 3.38, significantly higher than the male average of 1.43.

Across disciplines—from Art to Information Science—female First Authors consistently “punch above their weight,” demonstrating that research excellence is not constrained by representation alone.

See the Pen Gender3 by lee chung (@lee-chung) on CodePen.

(Interactive Chart: View the Gender Impact Landscape) 

Key Takeaways

This analysis highlights two important messages:

• Representation among early-career female researchers is improving, even in a STEM-focused institution.
• Impact is already strong. Across multiple fields, female First Authors at HKUST are producing some of the most influential research on campus.

This analysis is not intended to create a “gender war” or to position research achievement as a competition between men and women. Rather, it aims to illuminate contributions that may otherwise go unnoticed. Academic excellence is a shared endeavour. It grows from the collective wisdom and contributions of diverse minds, and recognizing diversity of talent ultimately benefits the entire academic community.

Looking Ahead

This pilot study is only a starting point. Looking forward, we hope to extend this into a longitudinal study, tracking these First Authors over the next decade. Such an analysis would allow us to answer the critical question: Are we successfully helping these brilliant First Authors transition into the Principal Investigators of tomorrow?

As International Women’s Day approaches, we hope this study serves as both a data-informed recognition of the outstanding contributions made by female researchers at HKUST and a reminder of the importance of creating an inclusive and supportive research environment for the future.

References

1. Centre for Science and Technology Studies. (n.d.). Indicators. CWTS Leiden Ranking. Retrieved February 10, 2026, from https://traditional.leidenranking.com/information/indicators#gender-indicators
2. Clarivate. (2025, October 17). Working with author positions. InCites Help Center. https://incites.zendesk.com/hc/en-gb/articles/25087884023441-Working-with-Author-Positions
3. Konkiel, S. (2018, June 19). A guide to identifying author gender for bibliometric analyses. The Bibliomagician. https://thebibliomagician.wordpress.com/2018/06/19/a-guide-to-identifying-author-gender-for-bibliometric-analyses-by-stacy-konkiel/
4. Nature Index. (n.d.). A guide to author gender ratio data on Nature Index. Nature. Retrieved February 23, 2026, from https://www.nature.com/nature-index/author-gender-ratio/methodology
5. Springer Nature. (2024, October 12). First author and corresponding author defined. Springer Nature Support. https://support.springernature.com/en/support/solutions/articles/6000214118-first-author-and-corresponding-author-defined