# Structural gender inequities in dental specialty training in Türkiye: evidence from a national cross-sectional study

## TL;DR

This study finds gender imbalances in dental specialty training in Türkiye, with women underrepresented in surgical fields, and shows that AI tools for gender classification have limited accuracy.

## Contribution

The study reveals significant gender disparities in dental specialty choices and evaluates the reliability of AI-based gender classification in an educational context.

## Key findings

- Female students dominate in Pediatric and Restorative Dentistry, while males are overrepresented in Oral and Maxillofacial Surgery.
- AI-based gender classification (NamSor) had moderate accuracy (67.9%) but misclassified nearly half of male names and a quarter of female names.
- Gender disparities in dental specialties correlate with university type and highlight the need for equity-focused policies.

## Abstract

This study investigated gender-based disparities in specialty preferences among dental students in Türkiye and evaluated the predictive performance of an artificial intelligence–based gender classification tool (NamSor). This study aimed to determine how these disparities shape specialty distribution and to assess whether AI-based gender identification aligns with verified data.

A cross-sectional descriptive design was employed using publicly available data from Turkish university websites and the Council of Higher Education. Gender was manually verified for all specialty students and compared with NamSor’s predictions. Statistical analyses included Chi-square tests, Bonferroni adjustments, Kappa coefficient, accuracy metrics, and ROC curve analysis to determine the agreement between actual and predicted gender classifications.

Significant gender disparities were observed across specialties (χ²=443.55; p < 0.001). Female students were predominantly represented in Pediatric Dentistry (17.9%) and Restorative Dentistry (12.4%), whereas males were concentrated in Oral and Maxillofacial Surgery (26.5%). Gender distribution also differed according to university type. NamSor demonstrated moderate accuracy (67.9%), with higher sensitivity in females (75.1%) than in males (53.6%). The Kappa coefficient was 0.283, indicating a fair agreement between the actual and predicted gender. Misclassification was notable, as 24.9% of women and 46.4% of men were incorrectly classified. The AUC was 0.64, suggesting a modest discrimination ability.

Gender-based imbalances persist in dental specialty education in Türkiye, particularly in surgical fields, where women remain underrepresented. The modest performance of the NamSor algorithm underscores that AI-based name classification should be regarded as an exploratory method and applied only with caution, while verified administrative data remain the reference standard for gender identification. These findings highlight the importance of equity-focused policies, mentorship initiatives, and gender-sensitive strategies to support balanced participation and career development in dental education.