Exploring Racial Bias in Rating Facial Attractiveness Among Artificial Intelligence Platforms
Facial Plastic Surgery · 21.10.2026
DOI: 10.1055/a-2945-2437 ↗Kurz gefasst
Fachartikel in der Zeitschrift Facial Plastic Surgery, erschienen am 21.10.2026. Verfasst von Nitika V Tripathi und 6 weiteren Autor:innen. Beteiligte Einrichtung: MedStar Georgetown University Hospital, Department of Otolaryngology - Head and Neck Surgery, District of Columbia, United States, Washington. Die Arbeit nennt 9 Literaturangaben.
Abstract (Originalsprache)
As artificial intelligence (AI) becomes increasingly integrated into aesthetic medicine, its influence on beauty standards and clinical practice warrants closer evaluation, especially among a diverse population. This study aimed to evaluate whether AI-based facial attractiveness platforms assign differing attractiveness scores across ethnic groups despite standardized image characteristics, with the hypothesis that significant differences would exist between ethnic groups and AI platforms, reflecting potential preferential scoring patterns favoring Eurocentric facial features. Cross-sectional comparative study evaluating AI-generated facial attractiveness scores across ethnic groups using multiple AI-based rating platforms. A database of 120 stock AI-generated images of women from diverse ethnic backgrounds (Black, White, Latino, and Asian) was developed from an online platform, with 30 photos per ethnicity. Images were of young adult women in frontal view with neutral expressions. These photographs were then evaluated using five different AI-based facial rating websites: Prettyscale (PS), Attractivenesstest (AT), Beautyscoretest (BST), Hotchat3000 (HC), and Attractivenessscore. White females received the highest average score (8.18), followed by Asian (7.79), Latino (7.72), and Black (6.91). Among AI-based platforms, PS yielded the highest scores (8.50), while BST yielded the lowest (6.65). Post hoc pairwise comparisons indicated significant differences in attractiveness scores between certain ethnicities, such as White versus Black (+ 1.27, p < 0.001), White versus Asian (+ 0.39, p = 0.05), and White versus Latino (+ 0.46, p < 0.001). Pairwise differences in attractiveness scores between AI-based platforms were also significant, such as PS versus BST (+ 1.85, p < 0.001), PS versus AT (+ 0.38, p = 0.007), and PS versus HC (+ 1.54, p < 0.001). These findings suggest that some AI platforms may exhibit bias toward White females. Understanding these trends is essential when consulting with patients who seek cosmetic enhancement based on societal standards.
Autor:innen und Institutionen
- MedStar Georgetown University Hospital, Department of Otolaryngology - Head and Neck Surgery, District of Columbia, United States, Washington.
- Georgetown University School of Medicine, Department of Otolaryngology - Head and Neck Surgery, District of Columbia, United States, Washington.
ORCID-Profile: Nitika V Tripathi, Amir A Hakimi
Zitieren
APA
Tripathi, N. V., Clementi, E., Aribindi, S., Singh, L., Hakimi, A. A., Chu, E., & Reilly, M. J. (2026). Exploring Racial Bias in Rating Facial Attractiveness Among Artificial Intelligence Platforms. Facial Plastic Surgery. https://doi.org/10.1055/a-2945-2437
Vancouver
Tripathi NV, Clementi E, Aribindi S, Singh L, Hakimi AA, Chu E, et al. Exploring Racial Bias in Rating Facial Attractiveness Among Artificial Intelligence Platforms. Facial Plastic Surgery. 2026. doi:10.1055/a-2945-2437
BibTeX
@article{Tripathi2026,
author = {Tripathi, Nitika V and Clementi, Emily and Aribindi, Seetha and Singh, Lovedeep and Hakimi, Amir A and Chu, Eugenia and Reilly, Michael J},
title = {Exploring Racial Bias in Rating Facial Attractiveness Among Artificial Intelligence Platforms},
journal = {Facial Plastic Surgery},
year = {2026},
doi = {10.1055/a-2945-2437}
}
EUVIMED