Berger-Estilita Joana, Gisselbaek Mia, Devos Arnout, Chan Albert, Ingrassia Pier Luigi, Meco Basak Ceyda, Chang Odmara L Barreto, Savoldelli Georges L, Matos Francisco Maio, Dieckmann Peter, Østergaard Doris, Saxena Sarah
Institute for Medical Education, University of Bern, Bern, Switzerland.
RISE-Health, Centre for Health Technology and Services Research, Faculty of Medicine, University of Porto, Porto, Portugal.
Adv Simul (Lond). 2025 May 4;10(1):26. doi: 10.1186/s41077-025-00355-1.
Simulation-based medical education (SBME) is a critical training tool in healthcare, shaping learners' skills, professional identities, and inclusivity. Leadership demographics in SBME, including age, gender, race/ethnicity, and medical specialties, influence program design and learner outcomes. Artificial intelligence (AI) platforms increasingly generate demographic data, but their biases may perpetuate inequities in representation. This study evaluated the demographic profiles of simulation instructors and heads of simulation labs generated by three AI platforms-ChatGPT, Gemini, and Claude-across nine global locations.
A global cross-sectional study was conducted over 5 days (November 2024). Standardized English prompts were used to generate demographic profiles of simulation instructors and heads of simulation labs from ChatGPT, Gemini, and Claude. Outputs included age, gender, race/ethnicity, and medical specialty data for 2014 instructors and 1880 lab heads. Statistical analyses included ANOVA for continuous variables and chi-square tests for categorical data, with Bonferroni corrections for multiple comparisons: P significant < 0.05.
Significant demographic differences were observed among AI platforms. Claude profiles depicted older heads of simulation labs (mean: 57 years) compared to instructors (mean: 41 years), while ChatGPT and Gemini showed smaller age gaps. Gender representation varied, with ChatGPT and Gemini generating balanced profiles, while Claude showed a male predominance (63.5%) among lab heads. ChatGPT and Gemini outputs reflected greater racial diversity, with up to 24.4% Black and 20.6% Hispanic/Latin representation, while Claude predominantly featured White profiles (47.8%). Specialty preferences also differed, with Claude favoring anesthesiology and surgery, whereas ChatGPT and Gemini offered broader interdisciplinary representation.
AI-generated demographic profiles of SBME leadership reveal biases that may reinforce inequities in healthcare education. ChatGPT and Gemini demonstrated broader diversity in age, gender, and race, while Claude skewed towards older, White, and male profiles, particularly for leadership roles. Addressing these biases through ethical AI development, enhanced AI literacy, and promoting diverse leadership in SBME are essential to fostering equitable and inclusive training environments.
Not applicable. This study exclusively used AI-generated synthetic data.
基于模拟的医学教育（SBME）是医疗保健领域的一种关键培训工具，塑造着学习者的技能、职业身份和包容性。SBME中的领导人口统计学特征，包括年龄、性别、种族/族裔和医学专业，会影响项目设计和学习者的成果。人工智能（AI）平台越来越多地生成人口统计数据，但其偏差可能会使代表性方面的不平等长期存在。本研究评估了由ChatGPT、Gemini和Claude这三个AI平台在全球九个地点生成的模拟教员和模拟实验室负责人的人口统计学特征。
在2024年11月的5天内进行了一项全球横断面研究。使用标准化的英语提示从ChatGPT、Gemini和Claude生成模拟教员和模拟实验室负责人的人口统计学特征。输出结果包括2014名教员和1880名实验室负责人的年龄、性别、种族/族裔和医学专业数据。统计分析包括对连续变量的方差分析和对分类数据的卡方检验，并对多重比较进行Bonferroni校正：P值<0.05为显著。
在人工智能平台之间观察到显著的人口统计学差异。Claude生成的模拟实验室负责人的年龄（平均57岁）比教员（平均41岁）大，而ChatGPT和Gemini显示的年龄差距较小。性别代表性各不相同，ChatGPT和Gemini生成的特征较为均衡，而Claude显示实验室负责人中男性占主导（63.5%）。ChatGPT和Gemini的输出反映出更大的种族多样性，黑人占比高达24.4%，西班牙裔/拉丁裔占比20.6%，而Claude生成的主要是白人特征（4�8%）。专业偏好也有所不同，Claude更倾向于麻醉学和外科，而ChatGPT和Gemini提供了更广泛的跨学科代表性。
人工智能生成的SBME领导层人口统计学特征揭示了可能加剧医学教育不平等的偏差。ChatGPT和Gemini在年龄·性别和种族方面表现出更广泛的多样性，而Claude则倾向于年龄较大、白人、男性的特征，尤其是在领导角色方面。通过符合道德的人工智能开发、提高人工智能素养以及在SBME中促进多元化领导来解决这些偏差，对于营造公平和包容的培训环境至关重要。
不适用。本研究仅使用人工智能生成的合成数据。
Zotero 插件